Skip to content
This repository has been archived by the owner on Oct 11, 2024. It is now read-only.

Commit

Permalink
mm publish workflow (#193)
Browse files Browse the repository at this point in the history 
SUMMARY:
* seed workflow for neuralmagic publish whl and tar workflow

TEST PLAN:
i'll be iterating on this going forward

---------

Co-authored-by: andy-neuma <andy@neuralmagic.com>
  • Loading branch information
@andy-neuma
andy-neuma and andy-neuma authored Apr 17, 2024
1 parent f614283 commit e07da81
Showing 1 changed file with 83 additions and 0 deletions.
83 changes: 83 additions & 0 deletions .github/workflows/nm-publish.yml
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,83 @@
name: nm publish whl
run-name: ${{ github.actor }} publishing whl on branch '${{ github.ref}}'
on:
workflow_dispatch:
inputs:
label:
description: "requested runner label (specifies instance)"
type: string
required: true
timeout:
description: "time limit for run in minutes "
type: string
required: true
gitref:
description: 'git commit hash or branch name'
type: string
required: true
python:
description: "python version, e.g. 3.10.12"
type: string
required: true
whl:
description: "whl to test (variable appears late binding so unusable outside 'download artifact')"
type: string
required: true
tarfile:
description: "source distribution (variable appears late binding so unusable outside 'download artifact')"
type: string
required: true

workflow_call:
inputs:
label:
description: "requested runner label (specifies instance)"
type: string
required: true
timeout:
description: "time limit for run in minutes "
type: string
required: true
gitref:
description: 'git commit hash or branch name'
type: string
required: true
python:
description: "python version, e.g. 3.10.12"
type: string
required: true
whl:
description: "whl to test (variable appears late binding so unusable outside 'download artifact')"
type: string
required: true
tarfile:
description: "source distribution (variable appears late binding so unusable outside 'download artifact')"
type: string
required: true

jobs:

PUBLISH:
runs-on: ${{ inputs.label }}
timeout-minutes: ${{ fromJson(inputs.timeout) }}

steps:

- name: checkout
id: checkout
uses: actions/checkout@v4
with:
fetch-depth: 0
ref: ${{ inputs.gitref }}
submodules: recursive

- name: setenv
id: setenv
uses: ./.github/actions/nm-set-env/

- name: set python
id: set_python
uses: ./.github/actions/nm-set-python/
with:
python: ${{ inputs.python }}
venv: UPLOAD

4 comments on commit e07da81

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

bigger_is_better

Benchmark suite Current: e07da81 Previous: f614283 Ratio
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.865557069328743 prompts/s 3.8610593618713698 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3962.1959960619615 tokens/s 3957.585845918154 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.930654367665678 prompts/s 2.8762319297766785 prompts/s 0.98
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 380.98506779653815 tokens/s 373.9101508709682 tokens/s 0.98
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 13.857039254021684 prompts/s 13.818340120764868 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3561.2590882835725 tokens/s 3551.313411036571 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.0380924243226337 prompts/s 2.0356812748181445 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4176.051377437077 tokens/s 4171.110932102378 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 13.86771808437483 prompts/s 13.83313329191869 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3564.003547684332 tokens/s 3555.115256023103 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 11.406821091551251 prompts/s 11.461714839559455 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1482.8867419016626 tokens/s 1490.0229291427293 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.4640350461686084 prompts/s 0.4640629472911598 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 122.66302410420995 tokens/s 122.67039948694519 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 115.09306571771272 tokens/s 115.0999859401354 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.3718606777712226 prompts/s 3.3727731640942387 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 438.34188811025894 tokens/s 438.46051133225103 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6.642699956070449 prompts/s 6.6455362786143315 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 863.5509942891584 tokens/s 863.9197162198632 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.038251332822462 prompts/s 2.0362376384303142 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4176.376980953224 tokens/s 4172.250921143714 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 5.588186873290385 prompts/s 5.584184632594255 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2612.8126544756524 tokens/s 2610.9413668157695 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.8162228992328795 prompts/s 1.8242133028106884 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 577.0697125885289 tokens/s 579.6085088442419 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 406.66199366037097 tokens/s 408.4571683479375 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 8.20005880279259 prompts/s 8.204311308211423 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4206.630165832598 tokens/s 4208.81170111246 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4.752292582565571 prompts/s 4.761041106221304 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 617.7980357335242 tokens/s 618.9353438087696 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 10.15079109289863 prompts/s 10.188295927248848 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1319.6028420768218 tokens/s 1324.4784705423501 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6.272232110471651 prompts/s 6.284131863935961 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 815.3901743613146 tokens/s 816.937142311675 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 13.880559589163818 prompts/s 13.849374891153944 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3567.303814415101 tokens/s 3559.289347026564 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.0704658410446255 prompts/s 2.0687898741290685 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4242.384508300438 tokens/s 4238.950452090461 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.2920252408249966 prompts/s 2.2931050805793345 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 707.8385150398476 tokens/s 708.1719983517139 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 459.2240651843874 tokens/s 459.05823481789736 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4.121969285160537 prompts/s 4.117774112755088 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1306.2108467745227 tokens/s 1304.88143859096 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 823.0913147379968 tokens/s 822.036736494782 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.9836203489082419 prompts/s 0.9829525462328785 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 290.02373861009147 tokens/s 289.82683476525165 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 212.30461610835494 tokens/s 212.1604775789045 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 8.0296620857472 prompts/s 8.003867020226506 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4119.216649988313 tokens/s 4105.983781376198 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 5.77977525777631 prompts/s 5.774533967974599 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2965.024707239247 tokens/s 2962.3359255709697 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.362522321565149 prompts/s 3.36494277442852 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1065.54969848078 tokens/s 1066.3167157886537 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 768.5963855371709 tokens/s 769.1743224451518 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 5.57256643019758 prompts/s 5.5756421453211775 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2605.5091601031804 tokens/s 2606.94724146637 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.793140509089632 prompts/s 1.7923805285807484 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 233.10826618165214 tokens/s 233.0094687154973 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.4504338196181465 prompts/s 2.451690844780352 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 756.7593083999401 tokens/s 757.1475112907001 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 562.6228722294193 tokens/s 562.8461084601677 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.2565247725546387 prompts/s 0.25619437756826696 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 33.348220432103034 tokens/s 33.305269083874705 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.2241811592405525 prompts/s 2.2254334822374995 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 686.8864534710623 tokens/s 687.2732042077995 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 510.5770957655032 tokens/s 510.8645756931544 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.761187251828045 prompts/s 3.7534636774748145 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3855.2169331237465 tokens/s 3847.300269411685 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.24055182046855503 prompts/s 0.24030119136010175 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 31.271736660912154 tokens/s 31.239154876813227 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6.355078088300283 prompts/s 6.357323118259827 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 826.1601514790368 tokens/s 826.4520053737775 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 7.055617150371873 prompts/s 7.085602937383071 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 917.2302295483435 tokens/s 921.1283818597992 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6.042365267885718 prompts/s 6.031337879721132 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2825.1683046526464 tokens/s 2820.0123390424123 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6.434153756957472 prompts/s 6.4694506264038045 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 836.4399884044714 tokens/s 841.0285814324945 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.3715750233638944 prompts/s 3.3699096999560294 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 438.3047530373063 tokens/s 438.08826099428387 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.002114392968378 prompts/s 1.999974660314367 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4102.332391192206 tokens/s 4097.948078984138 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.7597826159917553 prompts/s 3.7524375975159714 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3853.777181391549 tokens/s 3846.248537453871 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 24.15862864717864 prompts/s 24.209914874388183 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3116.463095486045 tokens/s 3123.0790187960756 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.406773737595247 prompts/s 3.4259820434496215 prompts/s 1.01
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1079.5725297065578 tokens/s 1085.6594497487506 tokens/s 1.01
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 778.7226121220133 tokens/s 783.1201114438304 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.7762320774504231 prompts/s 0.7749833249942054 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 100.910170068555 tokens/s 100.74783224924671 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.8035522685850456 prompts/s 3.8074443941468323 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1460.5640711366575 tokens/s 1462.0586473523838 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6.648644425852843 prompts/s 6.661109719147804 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 864.3237753608695 tokens/s 865.9442634892146 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 24.209753106999734 prompts/s 24.209434267431366 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3123.058150802966 tokens/s 3123.017020498646 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 5.587310909067573 prompts/s 5.603939541478122 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 726.3504181787845 tokens/s 728.512140392156 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.9402589195120397 prompts/s 1.937156376790832 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3975.590526080169 tokens/s 3969.2334160444148 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.9680373444610474 prompts/s 0.9680550720331814 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 288.6461485802469 tokens/s 288.65143452861395 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 210.04474300115805 tokens/s 210.0421358292795 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.9839282882071408 prompts/s 0.9839830038004401 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 290.11453553883615 tokens/s 290.1306686139058 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 212.4005995752755 tokens/s 212.40585114370901 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6.6523115189097 prompts/s 6.662976396622885 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 864.800497458261 tokens/s 866.186931560975 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 19.692425319712775 prompts/s 19.78486715107585 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2560.015291562661 tokens/s 2572.0327296398605 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 10.14914877750025 prompts/s 10.173573114455698 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1319.3893410750325 tokens/s 1322.5645048792408 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.976375764354526 prompts/s 1.9758159876136816 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 256.9288493660884 tokens/s 256.8560783897786 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 13.161879669160863 prompts/s 13.300308666926675 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 855.522178495456 tokens/s 864.5200633502338 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 7.472383657855176 prompts/s 7.453273360982778 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3833.332816479706 tokens/s 3823.529234184165 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 25.803409139406487 prompts/s 25.826332223569324 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1677.2215940614215 tokens/s 1678.711594532006 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 11.531389103291039 prompts/s 11.557698016145565 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1499.080583427835 tokens/s 1502.5007420989234 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 15.850345601758182 prompts/s 15.844759134566926 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4073.5388196518525 tokens/s 4072.1030975837 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.4892914970274247 prompts/s 3.4867424099285627 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3576.5237844531107 tokens/s 3573.910970176777 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.2169688813771815 prompts/s 2.217444800755505 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 684.6591097394437 tokens/s 684.8060863346534 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 508.59335309748496 tokens/s 508.97454000701254 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.4918930762468003 prompts/s 0.4919040196850423 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 130.0270157750792 tokens/s 130.0299085635441 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 121.9862036220315 tokens/s 121.99219688189051 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.014582889277618 prompts/s 1.0140688093709789 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 131.89577560609035 tokens/s 131.82894521822726 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.038787071823669 prompts/s 2.036113249809608 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4177.474710166698 tokens/s 4171.996048859887 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.46608246601441056 prompts/s 0.46608893956115444 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 126.03491324317682 tokens/s 126.03666377653032 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 118.10218967161154 tokens/s 118.10383002519947 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.4822689710749053 prompts/s 3.392540246789395 prompts/s 0.97
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1103.4962142439267 tokens/s 1075.0620788050915 tokens/s 0.97
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 796.3136607421725 tokens/s 775.8558608927518 tokens/s 0.97
{"name": "request_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 5.5887785806066965 prompts/s 5.585521134841847 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2613.089313148467 tokens/s 2611.566261806654 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.24056004084524407 prompts/s 0.24039690868156252 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 31.27280530988173 tokens/s 31.251598128603128 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 23.941627439523007 prompts/s 23.902574254883636 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3088.469939698468 tokens/s 3083.4320788799887 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 28.77622982103914 prompts/s 28.685373432532312 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3712.133646914049 tokens/s 3700.4131727966683 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.731449223638767 prompts/s 2.730121197361893 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 355.08839907303974 tokens/s 354.9157556570461 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 11.420382824426436 prompts/s 11.470490779562892 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1484.6497671754369 tokens/s 1491.163801343176 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 25.754436954975038 prompts/s 25.323428935788236 prompts/s 0.98
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1674.0384020733775 tokens/s 1646.0228808262352 tokens/s 0.98
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.7944812221230908 prompts/s 1.7955875055772352 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 233.28255887600181 tokens/s 233.42637572504057 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.9108234907835872 prompts/s 0.9110758622036463 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 118.40705380186633 tokens/s 118.43986208647402 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.689306276171444 prompts/s 1.6871370010620006 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 550.9458440938182 tokens/s 550.2383624943581 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 383.9815689821374 tokens/s 383.47499276138564 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.9125749922358185 prompts/s 0.9125257291449026 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 118.6347489906564 tokens/s 118.62834478883734 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 9.391510951112915 prompts/s 9.382388249580933 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2413.6183144360193 tokens/s 2411.2737801422995 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.761104538895074 prompts/s 3.7544907839644397 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3855.132152367451 tokens/s 3848.353053563551 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.1798190230855488 prompts/s 3.1840643884454667 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1498.8585736357427 tokens/s 1500.859694523986 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.9839508920530994 prompts/s 0.9839178110374442 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 290.1212003581632 tokens/s 290.1114463104272 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 212.5727507191516 tokens/s 212.42457567694743 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.49189644600910093 prompts/s 0.49190932959507616 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 130.02790653804573 tokens/s 130.03131218516242 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 122.00015653917721 tokens/s 121.99351373957889 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4.043525740341625 prompts/s 4.0438138332352525 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4144.6138838501665 tokens/s 4144.909179066134 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.3518631145960494 prompts/s 2.3469404665437135 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 726.3180461369826 tokens/s 724.7978011478066 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 539.7494489823072 tokens/s 538.588415278273 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.24029821321605013 prompts/s 0.24050293681709098 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 31.238767718086514 tokens/s 31.265381786221827 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.9474319777177606 prompts/s 0.9475788682829257 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 279.3534767366741 tokens/s 279.3967879094483 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 204.5158248167482 tokens/s 204.55069170000758 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 49.02623119211846 prompts/s 48.96805455867365 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3186.7050274877 tokens/s 3182.923546313787 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.376053193027138 prompts/s 3.374610743765939 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 438.88691509352793 tokens/s 438.6993966895721 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.7662324204314981 prompts/s 0.7657568479062911 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 99.61021465609475 tokens/s 99.54839022781783 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 25.20761877041418 prompts/s 24.766332745105206 prompts/s 0.98
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1638.4952200769217 tokens/s 1609.8116284318385 tokens/s 0.98
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 16.829730530608096 prompts/s 16.866089445975987 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2187.8649689790527 tokens/s 2192.591627976878 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 25.793172955729585 prompts/s 25.094879977834303 prompts/s 0.97
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1676.556242122423 tokens/s 1631.1671985592297 tokens/s 0.97
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.7935447069365122 prompts/s 1.7953852896812499 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 233.16081190174657 tokens/s 233.40008765856248 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4.38312644856314 prompts/s 4.38161638314841 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1388.9689402851734 tokens/s 1388.4904156558998 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1001.8395240108808 tokens/s 1001.6228997997828 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.47409127556233777 prompts/s 0.47419619713499056 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 125.32128778214836 tokens/s 125.34902275066341 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 99.24310701771604 tokens/s 99.26507060025803 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 7.470159625721567 prompts/s 7.455363008350998 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3832.1918879951636 tokens/s 3824.601223284062 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 7.471949051605978 prompts/s 7.454831011288954 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3833.1098634738664 tokens/s 3824.3283087912337 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.9126937775274014 prompts/s 0.9128817503542456 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 118.65019107856219 tokens/s 118.67462754605192 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 23.19143238005445 prompts/s 23.659972503065408 prompts/s 1.02
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2991.6947770270244 tokens/s 3052.1364528954373 tokens/s 1.02
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.6954508679943148 prompts/s 3.6998055414826614 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 480.4086128392609 tokens/s 480.974720392746 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 7.056194907735735 prompts/s 7.047655458259762 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3299.1944910609204 tokens/s 3295.201786063935 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 12.251688594885204 prompts/s 12.255799367736572 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1580.4678287401914 tokens/s 1580.9981184380176 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 14.173890992115513 prompts/s 14.173546057606423 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3642.689984973687 tokens/s 3642.6013368048507 tokens/s 1.00

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

smaller_is_better

Benchmark suite Current: e07da81 Previous: f614283 Ratio
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6378.581413000575 ms 6371.257894000337 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 111.62117981332509 ms 110.22614894004315 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 71.45514599960734 ms 70.59756849957921 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 36.29889292579925 ms 36.29656023542237 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 36.97315168296708 ms 37.003949056523695 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 259614.5786724992 ms 255015.42043600057 ms 1.02
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 244308.37738517133 ms 241563.51897034133 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 246049.76843250007 ms 241788.34201849985 ms 1.02
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 70.71990302659997 ms 69.81532707122858 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 66.76681651907927 ms 66.55115569993295 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6069.2304214990145 ms 6038.91504549938 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 122.46475216280184 ms 122.6481674813161 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 92.15208550085663 ms 90.39529849906103 ms 1.02
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 55.099070833348854 ms 54.70650422192142 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 48.127281345213696 ms 47.87876855092492 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 18684.68709550143 ms 18570.783775499876 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1066.796312155333 ms 1069.8053945153758 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 235.2079810007126 ms 240.836296499765 ms 0.98
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 162.1375558564125 ms 161.64414130966222 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 149.2627111229515 ms 149.01588750678897 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3662.992083000063 ms 3693.144686998494 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 142.85702721993098 ms 145.47056394336929 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 96.08639149882947 ms 96.20490799898107 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 24.29278854544752 ms 24.574329407755346 ms 0.99
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 23.695896667542776 ms 23.984857821068697 ms 0.99
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 66557.87866549962 ms 66362.64726900027 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 23176.74328639136 ms 23032.820925857333 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 17623.261793999518 ms 17523.711426500086 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 261.0721530367403 ms 260.76070724560543 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 278.010503890837 ms 277.67840906393417 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2416.305994499453 ms 2395.6918630001383 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 116.5629329613439 ms 116.00974644261198 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 81.06380800018087 ms 81.38328200038814 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 17.91632854260476 ms 17.886947256665138 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 16.1783885536632 ms 16.07437932938582 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 7913.390777499444 ms 7836.637858499671 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 156.60907870132846 ms 156.41282928670262 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 129.36970100054168 ms 133.33034749939543 ms 0.97
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 58.246506549947505 ms 57.948193639127204 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 57.82946412437782 ms 57.730200439855054 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 64696.70536650028 ms 64516.988244999084 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 41947.97561523735 ms 41670.433804386 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 41184.04677449962 ms 41353.95252950002 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 100.50144845859734 ms 99.74073506785071 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 102.81648032746044 ms 101.81170827653953 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6057.023901000321 ms 6043.065258500064 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 119.8196006967252 ms 121.5648558233382 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 80.72478450048948 ms 81.48733600046398 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 39.78841433967801 ms 39.74703724134066 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 39.49186390275661 ms 39.47839664805598 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2135.121108499334 ms 2118.9894785002252 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 136.22967760002211 ms 136.5197102166288 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 103.10689500056469 ms 104.6584610003265 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 15.054929704787915 ms 15.07500200677561 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 13.618156992207627 ms 13.621546843457175 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 12290.266897000947 ms 12121.188299499408 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 198.90342098401015 ms 200.5113167760137 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 172.09408800044912 ms 170.04918400016322 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 95.86945681823846 ms 95.57540617024532 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 101.47341102832368 ms 101.46300535968142 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2071.1908989997028 ms 2048.588919499707 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 117.41245240675198 ms 116.12702218662889 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 85.8170905003135 ms 87.0328774999507 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 11.566048442568711 ms 11.610714418591797 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 11.59489551136406 ms 11.619119963638639 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6139.707543999975 ms 6115.064929499567 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 104.89042469334284 ms 106.53483679331353 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 67.97110799925576 ms 68.27251650065591 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 34.705563867462644 ms 34.64606003196931 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 35.20031126436331 ms 35.15808529023893 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 66015.65591649887 ms 72274.7995009995 ms 0.91
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 29164.48208331334 ms 32685.66015171202 ms 0.89
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 28257.675152000957 ms 31708.220931999676 ms 0.89
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 188.41709116546843 ms 195.2866665236033 ms 0.96
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 199.66467406590337 ms 206.56636920034035 ms 0.97
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 77150.65222300019 ms 77780.31293799996 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 57567.22902227469 ms 58013.02879457739 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 70169.3598439988 ms 71104.39611950004 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 66.24332605701616 ms 67.65889710905799 ms 0.98
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 65.03487843963993 ms 65.26044596054808 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1825.4841359994316 ms 1822.055524499774 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 93.11112377665268 ms 92.95755092335337 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 59.797208499730914 ms 56.33668000064063 ms 1.06
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 12.549763244228975 ms 12.412678296313379 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 11.856144193277148 ms 11.713674675928727 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1924.9286479998773 ms 1922.8771115003838 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 80.66479348000333 ms 79.14218095322818 ms 1.02
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 37.79352500077948 ms 37.81903199978842 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 10.946614239261466 ms 11.032363034803458 ms 0.99
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 11.055925112089627 ms 11.069762455417846 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 5372.168074500223 ms 5468.762963999325 ms 0.98
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 183.6061620000934 ms 184.89295740929083 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 174.95899550158356 ms 175.7307474999834 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 40.48906765663941 ms 41.12904338038068 ms 0.98
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 37.60412458571284 ms 38.23433923721759 ms 0.98
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6058.910562000165 ms 6050.86946100073 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 130.22003160331224 ms 128.14271269004166 ms 1.02
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 86.42561249962455 ms 85.98969000013312 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 40.38834413966342 ms 40.36753486631414 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 40.362877592449095 ms 40.32901228271403 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 14761.370102 ms 14996.088238500306 ms 0.98
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 205.65284574066027 ms 212.75933790932686 ms 0.97
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 165.0134245001027 ms 166.0542159997931 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 119.7739543031504 ms 121.12365628200462 ms 0.99
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 125.81718531554768 ms 127.37200629793627 ms 0.99
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 5240.570841498993 ms 5224.6451695009455 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 91.48198475971488 ms 92.90686182672896 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 62.23678750029649 ms 62.84830799995689 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 37.2797004113452 ms 37.25550891063875 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 31.689187422468247 ms 31.614687938990084 ms 1.00

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

bigger_is_better

Benchmark suite Current: e07da81 Previous: f614283 Ratio
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.0702939191933645 prompts/s 2.0687898741290685 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4242.032240427204 tokens/s 4238.950452090461 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.794609334579142 prompts/s 1.7955875055772352 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 233.29921349528846 tokens/s 233.42637572504057 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.694720730899894 prompts/s 3.6998055414826614 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 480.31369501698623 tokens/s 480.974720392746 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.4508216087429515 prompts/s 2.451690844780352 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 756.8790680227232 tokens/s 757.1475112907001 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 562.6923025566567 tokens/s 562.8461084601677 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.4740911070594224 prompts/s 0.47419619713499056 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 125.32124324008772 tokens/s 125.34902275066341 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 99.24307174443909 tokens/s 99.26507060025803 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 9.387657202414662 prompts/s 9.382388249580933 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2412.627901020568 tokens/s 2411.2737801422995 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 25.80969397756293 prompts/s 25.323428935788236 prompts/s 0.98
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1677.6301085415903 tokens/s 1646.0228808262352 tokens/s 0.98
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 5.582398986275805 prompts/s 5.603939541478122 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 725.7118682158548 tokens/s 728.512140392156 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 23.251052355121523 prompts/s 23.659972503065408 prompts/s 1.02
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2999.3857538106763 tokens/s 3052.1364528954373 tokens/s 1.02
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 11.435125989698314 prompts/s 11.470490779562892 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1486.5663786607809 tokens/s 1491.163801343176 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 8.145339574865245 prompts/s 8.204311308211423 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4178.55920190587 tokens/s 4208.81170111246 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.815413888629922 prompts/s 1.8242133028106884 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 576.8126651103108 tokens/s 579.6085088442419 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 406.4820621475713 tokens/s 408.4571683479375 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.9839734208658726 prompts/s 0.9839830038004401 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 290.1278430537054 tokens/s 290.1306686139058 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 212.4136222737188 tokens/s 212.40585114370901 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.29206255203473 prompts/s 2.2931050805793345 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 707.8500377363789 tokens/s 708.1719983517139 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 458.9259324186018 tokens/s 459.05823481789736 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.372363203932501 prompts/s 3.3727731640942387 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 438.4072165112251 tokens/s 438.46051133225103 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6.330721120854613 prompts/s 6.4694506264038045 prompts/s 1.02
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 822.9937457110997 tokens/s 841.0285814324945 tokens/s 1.02
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 23.902047496230495 prompts/s 23.902574254883636 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3083.3641270137337 tokens/s 3083.4320788799887 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.7596729099885065 prompts/s 3.7524375975159714 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3853.664732738219 tokens/s 3846.248537453871 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.2244495000752393 prompts/s 2.2254334822374995 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 686.9693242765692 tokens/s 687.2732042077995 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 510.5912404492701 tokens/s 510.8645756931544 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.9835887953016751 prompts/s 0.9829525462328785 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 290.0144349240166 tokens/s 289.82683476525165 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 212.29780557791355 tokens/s 212.1604775789045 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6.946141785746538 prompts/s 7.085602937383071 prompts/s 1.02
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 902.9984321470499 tokens/s 921.1283818597992 tokens/s 1.02
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.9759398845127358 prompts/s 1.9758159876136816 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 256.87218498665567 tokens/s 256.8560783897786 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.490199610200046 prompts/s 3.4867424099285627 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3577.454600455047 tokens/s 3573.910970176777 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.7285213712036342 prompts/s 2.730121197361893 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 354.70777825647247 tokens/s 354.9157556570461 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.46593866936991607 prompts/s 0.46608893956115444 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 125.9960287132169 tokens/s 126.03666377653032 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 118.06575256054093 tokens/s 118.10383002519947 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.938731963718114 prompts/s 1.937156376790832 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3972.4617936584154 tokens/s 3969.2334160444148 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.3726989527416027 prompts/s 3.374610743765939 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 438.4508638564084 tokens/s 438.6993966895721 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.9111925284769179 prompts/s 0.9110758622036463 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 118.45502870199934 tokens/s 118.43986208647402 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.24031972409109284 prompts/s 0.24030119136010175 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 31.24156413184207 tokens/s 31.239154876813227 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 7.466838184925605 prompts/s 7.453273360982778 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3830.487988866835 tokens/s 3823.529234184165 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 11.410612520227023 prompts/s 11.461714839559455 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1483.379627629513 tokens/s 1490.0229291427293 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 12.245573715102868 prompts/s 12.255799367736572 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1579.67900924827 tokens/s 1580.9981184380176 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.040871180157846 prompts/s 2.036113249809608 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4181.745048143427 tokens/s 4171.996048859887 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 11.513151198455704 prompts/s 11.557698016145565 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1496.7096557992415 tokens/s 1502.5007420989234 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.7721373150332344 prompts/s 0.7749833249942054 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 100.37785095432046 tokens/s 100.74783224924671 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.2161736937117804 prompts/s 2.217444800755505 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 684.4135345833635 tokens/s 684.8060863346534 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 508.688690061569 tokens/s 508.97454000701254 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.9133957640859882 prompts/s 0.9125257291449026 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 118.74144933117847 tokens/s 118.62834478883734 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6.645078986839318 prompts/s 6.662976396622885 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 863.8602682891113 tokens/s 866.186931560975 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.9897856977679733 prompts/s 1.999974660314367 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4077.070894726577 tokens/s 4097.948078984138 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 13.882529340741351 prompts/s 13.83313329191869 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3567.8100405705272 tokens/s 3555.115256023103 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4.384804196056201 prompts/s 4.38161638314841 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1389.5006016882496 tokens/s 1388.4904156558998 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1002.2785431345266 tokens/s 1001.6228997997828 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 15.743864950539436 prompts/s 15.844759134566926 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4046.173292288635 tokens/s 4072.1030975837 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6.272076435649383 prompts/s 6.284131863935961 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 815.3699366344198 tokens/s 816.937142311675 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 13.918835844529392 prompts/s 13.849374891153944 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3577.1408120440537 tokens/s 3559.289347026564 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6.347700601574241 prompts/s 6.357323118259827 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 825.2010782046513 tokens/s 826.4520053737775 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 25.832241855306044 prompts/s 25.826332223569324 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1679.0957205948928 tokens/s 1678.711594532006 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 28.58215259319803 prompts/s 28.685373432532312 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3687.0976845225455 tokens/s 3700.4131727966683 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.4639685477754497 prompts/s 0.4640629472911598 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 122.64544591896238 tokens/s 122.67039948694519 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 115.07657234291887 tokens/s 115.0999859401354 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.7934982434362643 prompts/s 1.7923805285807484 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 233.15477164671435 tokens/s 233.0094687154973 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.9839289176855154 prompts/s 0.9839178110374442 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 290.1147211426332 tokens/s 290.1114463104272 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 212.39089617159536 tokens/s 212.42457567694743 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.442933520757558 prompts/s 3.4259820434496215 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1091.0312033928626 tokens/s 1085.6594497487506 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 786.9926300418041 tokens/s 783.1201114438304 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 5.589270424670841 prompts/s 5.5756421453211775 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2613.319279759098 tokens/s 2606.94724146637 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 24.2111868214395 prompts/s 24.209914874388183 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3123.2430999656954 tokens/s 3123.0790187960756 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.4918911085408706 prompts/s 0.49190932959507616 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 130.02649563169373 tokens/s 130.03131218516242 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 121.21836551475522 tokens/s 121.99351373957889 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.2564787589274486 prompts/s 0.25619437756826696 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 33.34223866056831 tokens/s 33.305269083874705 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 8.01695091272711 prompts/s 8.003867020226506 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4112.695818229007 tokens/s 4105.983781376198 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.8027468630648054 prompts/s 3.8074443941468323 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - synthetic\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1460.2547954168851 tokens/s 1462.0586473523838 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6.647018386357555 prompts/s 6.661109719147804 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 864.1123902264821 tokens/s 865.9442634892146 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.24037063771646877 prompts/s 0.24039690868156252 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 31.24818290314094 tokens/s 31.251598128603128 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 10.17414044802103 prompts/s 10.188295927248848 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1322.6382582427339 tokens/s 1324.4784705423501 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 19.673891162562082 prompts/s 19.78486715107585 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2557.6058511330707 tokens/s 2572.0327296398605 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 5.7765448860039825 prompts/s 5.774533967974599 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2963.367526520043 tokens/s 2962.3359255709697 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4.1222197377215535 prompts/s 4.117774112755088 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1306.290212686583 tokens/s 1304.88143859096 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 823.1413261071906 tokens/s 822.036736494782 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 7.465860504993817 prompts/s 7.455363008350998 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3829.986439061828 tokens/s 3824.601223284062 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.864286803923941 prompts/s 3.8610593618713698 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3960.89397402204 tokens/s 3957.585845918154 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6.638373450186385 prompts/s 6.6455362786143315 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 862.98854852423 tokens/s 863.9197162198632 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 7.023258858277639 prompts/s 7.047655458259762 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3283.7949117762932 tokens/s 3295.201786063935 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.7949325927483175 prompts/s 1.7953852896812499 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 233.3412370572813 tokens/s 233.40008765856248 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 13.312288380270536 prompts/s 13.300308666926675 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 865.2987447175849 tokens/s 864.5200633502338 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4.746103576217203 prompts/s 4.761041106221304 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 8\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 616.9934649082364 tokens/s 618.9353438087696 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 48.752812859278606 prompts/s 48.96805455867365 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3168.9328358531093 tokens/s 3182.923546313787 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.040018163173438 prompts/s 2.0356812748181445 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4179.997216342375 tokens/s 4171.110932102378 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6.039602516272134 prompts/s 6.031337879721132 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2823.8765525081985 tokens/s 2820.0123390424123 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.996983029226785 prompts/s 4.0438138332352525 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4096.907604957455 tokens/s 4144.909179066134 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.369128846929252 prompts/s 3.36494277442852 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1067.6432403034107 tokens/s 1066.3167157886537 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 770.1469170336597 tokens/s 769.1743224451518 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.6818187212871252 prompts/s 1.6871370010620006 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 548.5038729106595 tokens/s 550.2383624943581 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 382.2616983738315 tokens/s 383.47499276138564 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.4919016140192644 prompts/s 0.4919040196850423 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 130.02927264985235 tokens/s 130.0299085635441 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 121.99160027677758 tokens/s 121.99219688189051 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.947527282813599 prompts/s 0.9475788682829257 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 279.3815777618657 tokens/s 279.3967879094483 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 204.53955611669556 tokens/s 204.55069170000758 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.7248233191891917 prompts/s 2.8762319297766785 prompts/s 1.06
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 354.22703149459494 tokens/s 373.9101508709682 tokens/s 1.06
{"name": "request_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 5.593284961974514 prompts/s 5.584184632594255 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2615.196316820804 tokens/s 2610.9413668157695 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.4714774390692384 prompts/s 3.392540246789395 prompts/s 0.98
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1100.076485666651 tokens/s 1075.0620788050915 tokens/s 0.98
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 793.5982571175696 tokens/s 775.8558608927518 tokens/s 0.98
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 25.38217947629556 prompts/s 25.094879977834303 prompts/s 0.99
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1649.8416659592112 tokens/s 1631.1671985592297 tokens/s 0.99
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 13.8401538622829 prompts/s 13.818340120764868 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3556.9195426067054 tokens/s 3551.313411036571 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.7597034986405204 prompts/s 3.7534636774748145 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3853.6960861065336 tokens/s 3847.300269411685 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.351545361129132 prompts/s 2.3469404665437135 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 726.2199153929728 tokens/s 724.7978011478066 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 539.676524985321 tokens/s 538.588415278273 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.370966750656052 prompts/s 3.3699096999560294 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 16\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 438.22567758528675 tokens/s 438.08826099428387 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 16.79421297617141 prompts/s 16.866089445975987 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 32\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2183.247686902283 tokens/s 2192.591627976878 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.9680488881195148 prompts/s 0.9680550720331814 prompts/s 1.00
{"name": "input_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 288.64959062984985 tokens/s 288.65143452861395 tokens/s 1.00
{"name": "output_throughput", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 210.04079408491822 tokens/s 210.0421358292795 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.24019285095526624 prompts/s 0.24050293681709098 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1,\n \"sparsity\": \"semi_structured_sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 31.22507062418461 tokens/s 31.265381786221827 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 10.15975125243689 prompts/s 10.173573114455698 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 64\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1320.7676628167956 tokens/s 1322.5645048792408 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.180951888709995 prompts/s 3.1840643884454667 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Dense (with dataset)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1499.3925679736753 tokens/s 1500.859694523986 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3.7605749906840873 prompts/s 3.7544907839644397 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 1024,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3854.5893654511897 tokens/s 3848.353053563551 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 24.631297798077004 prompts/s 24.766332745105206 prompts/s 1.01
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 64,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1601.0343568750054 tokens/s 1609.8116284318385 tokens/s 1.01
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 24.10894136258604 prompts/s 24.209434267431366 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 128,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3110.053435773599 tokens/s 3123.017020498646 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2.040529627952652 prompts/s 2.0362376384303142 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2048,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 4181.045207674983 tokens/s 4172.250921143714 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 7.464238181831078 prompts/s 7.454831011288954 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Sparse (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 512,\n \"output-len\": 1,\n \"num-prompts\": 1,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3829.154187279343 tokens/s 3824.3283087912337 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 5.5886457941375856 prompts/s 5.585521134841847 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine throughput - Sparse (with dataset)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"dataset\": \"sharegpt\",\n \"output-len\": 128,\n \"num-prompts\": 1000,\n \"sparsity\": \"sparse_w16a16\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2613.027227506969 tokens/s 2611.566261806654 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.763848850052772 prompts/s 0.7657568479062911 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 99.30035050686037 tokens/s 99.54839022781783 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 0.9128070467790825 prompts/s 0.9128817503542456 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 118.66491608128072 tokens/s 118.67462754605192 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 14.162669819219499 prompts/s 14.173546057606423 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine prefill throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 256,\n \"output-len\": 1,\n \"num-prompts\": 1\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3639.806143539411 tokens/s 3642.6013368048507 tokens/s 1.00
{"name": "request_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1.0143071669797927 prompts/s 1.0140688093709789 prompts/s 1.00
{"name": "token_throughput", "description": "VLLM Engine decode throughput - Dense (synthetic)\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax_model_len - 4096\nbenchmark_throughput {\n \"use-all-available-gpus_\": \"\",\n \"input-len\": 2,\n \"output-len\": 128,\n \"num-prompts\": 4\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 131.85993170737305 tokens/s 131.82894521822726 tokens/s 1.00

This comment was automatically generated by workflow using github-action-benchmark.

@github-actions
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

smaller_is_better

Benchmark suite Current: e07da81 Previous: f614283 Ratio
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2419.0179269999135 ms 2395.6918630001383 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 117.20231991069159 ms 116.00974644261198 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 81.61429250048968 ms 81.38328200038814 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 18.161439199680725 ms 17.886947256665138 ms 1.02
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 16.233541161772678 ms 16.07437932938582 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 5237.471369999184 ms 5224.6451695009455 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 91.50180193314252 ms 92.90686182672896 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 60.09721300142701 ms 62.84830799995689 ms 0.96
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 37.25594788852032 ms 37.25550891063875 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 31.757409114712246 ms 31.614687938990084 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 257308.96987199958 ms 255015.42043600057 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 243518.98619232 ms 241563.51897034133 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 243613.63321599946 ms 241788.34201849985 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 69.78756619298498 ms 69.81532707122858 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 66.90193202142939 ms 66.55115569993295 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2138.947069999631 ms 2118.9894785002252 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 136.90058723338853 ms 136.5197102166288 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 103.70127050009614 ms 104.6584610003265 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 15.110137266517217 ms 15.07500200677561 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 13.690271990730617 ms 13.621546843457175 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6066.9346030008455 ms 6038.91504549938 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 123.10161839067102 ms 122.6481674813161 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 95.20024650191772 ms 90.39529849906103 ms 1.05
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 54.95573334879406 ms 54.70650422192142 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 47.997319712418715 ms 47.87876855092492 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 7860.4375695013005 ms 7836.637858499671 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 156.88288736399167 ms 156.41282928670262 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 131.119418999333 ms 133.33034749939543 ms 0.98
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 57.92863140120204 ms 57.948193639127204 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 57.57763395532975 ms 57.730200439855054 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 3639.5473400007177 ms 3693.144686998494 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 142.48296434999548 ms 145.47056394336929 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 93.17490099965653 ms 96.20490799898107 ms 0.97
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 24.499413221090332 ms 24.574329407755346 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 23.80052844854185 ms 23.984857821068697 ms 0.99
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6134.418095000001 ms 6115.064929499567 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 106.02016567999574 ms 106.53483679331353 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 69.55545349956083 ms 68.27251650065591 ms 1.02
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 34.6953445373128 ms 34.64606003196931 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 35.20752824027514 ms 35.15808529023893 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 12626.318402501056 ms 12121.188299499408 ms 1.04
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 199.22702569068014 ms 200.5113167760137 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 171.93636349929875 ms 170.04918400016322 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 98.15261906506214 ms 95.57540617024532 ms 1.03
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 102.4142414242919 ms 101.46300535968142 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 14772.837518000415 ms 14996.088238500306 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 213.32647224934527 ms 212.75933790932686 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 170.28883050079457 ms 166.0542159997931 ms 1.03
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 119.73412242863623 ms 121.12365628200462 ms 0.99
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 126.07364461852305 ms 127.37200629793627 ms 0.99
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6391.470271500111 ms 6371.257894000337 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 110.91013564660291 ms 110.22614894004315 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 72.0535739992556 ms 70.59756849957921 ms 1.02
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 36.298603910084346 ms 36.29656023542237 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 36.987131992162624 ms 37.003949056523695 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1813.6816339992947 ms 1822.055524499774 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 93.77405370001725 ms 92.95755092335337 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 59.02614100068604 ms 56.33668000064063 ms 1.05
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 12.54288899389369 ms 12.412678296313379 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 11.801870261373644 ms 11.713674675928727 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 63103.95905850055 ms 64516.988244999084 ms 0.98
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 40465.1282516867 ms 41670.433804386 ms 0.97
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 39963.13871999973 ms 41353.95252950002 ms 0.97
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 99.1119378434025 ms 99.74073506785071 ms 0.99
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 101.46212264841162 ms 101.81170827653953 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1923.9120854999783 ms 1922.8771115003838 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 79.35825840001296 ms 79.14218095322818 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 38.7081480012057 ms 37.81903199978842 ms 1.02
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 11.005709969769681 ms 11.032363034803458 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-marlin\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 11.046433172364836 ms 11.069762455417846 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 18350.793265000902 ms 18570.783775499876 ms 0.99
{"name": "mean_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 1004.3674109506488 ms 1069.8053945153758 ms 0.94
{"name": "median_ttft_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 228.4966870029166 ms 240.836296499765 ms 0.95
{"name": "mean_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 161.5440135561111 ms 161.64414130966222 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - 2:4 Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4\nmax-model-len - 4096\nsparsity - semi_structured_sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 148.57129311499128 ms 149.01588750678897 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 66089.68731499954 ms 66362.64726900027 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 22913.870488265333 ms 23032.820925857333 ms 0.99
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 17695.353991999582 ms 17523.711426500086 ms 1.01
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 260.45060158512024 ms 260.76070724560543 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 276.9664923178263 ms 277.67840906393417 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 77670.21108049902 ms 77780.31293799996 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 58117.25211760533 ms 58013.02879457739 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 71017.82879199981 ms 71104.39611950004 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 68.47808421234329 ms 67.65889710905799 ms 1.01
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 65.81628570147745 ms 65.26044596054808 ms 1.01
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 2078.5573924995333 ms 2048.588919499707 ms 1.01
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 117.74490705339606 ms 116.12702218662889 ms 1.01
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 86.20008799971401 ms 87.0328774999507 ms 0.99
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 11.61648378924836 ms 11.610714418591797 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"150,0.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 11.644324777421152 ms 11.619119963638639 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6062.8847400003 ms 6050.86946100073 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 128.54808192007718 ms 128.14271269004166 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 90.95952099960414 ms 85.98969000013312 ms 1.06
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 40.43786095935808 ms 40.36753486631414 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - teknium/OpenHermes-2.5-Mistral-7B\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 40.41590427764934 ms 40.32901228271403 ms 1.00
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 67057.54034400161 ms 72274.7995009995 ms 0.93
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 29466.208567894686 ms 32685.66015171202 ms 0.90
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 28570.35936400098 ms 31708.220931999676 ms 0.90
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 189.36645500562108 ms 195.2866665236033 ms 0.97
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"1500,5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 201.19039318388218 ms 206.56636920034035 ms 0.97
{"name": "median_request_latency", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 5490.021190002153 ms 5468.762963999325 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 181.62115946935955 ms 184.89295740929083 ms 0.98
{"name": "median_ttft_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 170.90898900096363 ms 175.7307474999834 ms 0.97
{"name": "mean_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 40.47174379320329 ms 41.12904338038068 ms 0.98
{"name": "median_tpot_ms", "description": "VLLM Serving - Sparse\nmodel - neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50\nmax-model-len - 4096\nsparsity - sparse_w16a16\nbenchmark_serving {\n \"nr-qps-pair_\": \"750,2.5\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 37.77759011427127 ms 38.23433923721759 ms 0.99
{"name": "median_request_latency", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 6053.833416999623 ms 6043.065258500064 ms 1.00
{"name": "mean_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 121.70561902992631 ms 121.5648558233382 ms 1.00
{"name": "median_ttft_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 81.82817649958452 ms 81.48733600046398 ms 1.00
{"name": "mean_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 39.81906862449224 ms 39.74703724134066 ms 1.00
{"name": "median_tpot_ms", "description": "VLLM Serving - Dense\nmodel - NousResearch/Llama-2-7b-chat-hf\nmax-model-len - 4096\nsparsity - None\nbenchmark_serving {\n \"nr-qps-pair_\": \"300,1\",\n \"dataset\": \"sharegpt\"\n}", "gpu_description": "NVIDIA A10G x 1", "vllm_version": "0.2.0", "python_version": "3.10.12 (main, Mar 7 2024, 18:39:53) [GCC 9.4.0]", "torch_version": "2.2.1+cu121"} 39.45704398079755 ms 39.47839664805598 ms 1.00

This comment was automatically generated by workflow using github-action-benchmark.

Please sign in to comment.