This repository has been archived by the owner on Oct 11, 2024. It is now read-only.
forked from vllm-project/vllm
-
Notifications
You must be signed in to change notification settings - Fork 10
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Benchmark : Update benchmark configs for Nightly (#126)
SUMMARY: Update the benchmark configs such that the Nightly runs the following models, - `7b Mistral` * Base : teknium/OpenHermes-2.5-Mistral-7B * GPTQ : TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ * Marlin : neuralmagic/OpenHermes-2.5-Mistral-7B-marlin * Sparse : neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50 * Sparse 2:4 : neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4 - `Llama 7b fp16` * NousResearch/Llama-2-7b-chat-hf #fp16 - Update benchmark_serving num_prompts and qps pairs. - Minor update to the benchmark_throughput prefill and decode cases. TEST PLAN: Manual testing --------- Co-authored-by: Varun Sundar Rabindranath <varun@neuralmagic.com>
- Loading branch information
1 parent
b8c95c3
commit ac8f242
Showing
9 changed files
with
316 additions
and
83 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,3 +1,5 @@ | ||
| neuralmagic/benchmarks/configs/benchmark_serving.json | ||
| neuralmagic/benchmarks/configs/benchmark_throughput.json | ||
| neuralmagic/benchmarks/configs/benchmark_throughput_decode.json | ||
| neuralmagic/benchmarks/configs/benchmark_throughput_prefill.json | ||
| neuralmagic/benchmarks/configs/benchmark_remote_push.json |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,24 +1,26 @@ | ||
| { | ||
| "configs": [ | ||
| { | ||
| "description": "Benchmark vllm engine throughput - synthetic", | ||
| "description": "VLLM Engine throughput - synthetic", | ||
| "models": [ | ||
| "NousResearch/Llama-2-7b-chat-hf" | ||
| ], | ||
| "max_model_lens" : [4096], | ||
| "max_model_lens": [ | ||
| 4096 | ||
| ], | ||
| "script_name": "benchmark_throughput", | ||
| "script_args": { | ||
| "input-len" : [ | ||
| 256 | ||
| ], | ||
| "input-len": [ | ||
| 256 | ||
| ], | ||
| "output-len": [ | ||
| 128 | ||
| ], | ||
| "num-prompts": [ | ||
| 1000 | ||
| ], | ||
| "use-all-available-gpus_" : [] | ||
| "use-all-available-gpus_": [] | ||
| } | ||
| } | ||
| ] | ||
| } | ||
| } |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,97 +1,79 @@ | ||
| { | ||
| "configs": [ | ||
| { | ||
| "description": "VLLM Engine throughput (with dataset)", | ||
| "description": "VLLM Engine throughput - Dense (with dataset)", | ||
| "models": [ | ||
| "facebook/opt-125m", | ||
| "TinyLlama/TinyLlama-1.1B-Chat-v1.0", | ||
| "mistralai/Mistral-7B-Instruct-v0.2", | ||
| "teknium/OpenHermes-2.5-Mistral-7B", | ||
| "neuralmagic/OpenHermes-2.5-Mistral-7B-marlin", | ||
| "TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ", | ||
| "NousResearch/Llama-2-7b-chat-hf" | ||
| ], | ||
| "max_model_lens" : [4096], | ||
| "max_model_lens": [ | ||
| 4096 | ||
| ], | ||
| "script_name": "benchmark_throughput", | ||
| "script_args": { | ||
| "backend": [ | ||
| "vllm" | ||
| ], | ||
| "dataset": [ | ||
| "sharegpt" | ||
| ], | ||
| "output-len": [ | ||
| 128 | ||
| ], | ||
| "n": [ | ||
| 1 | ||
| ], | ||
| "num-prompts": [ | ||
| 1000 | ||
| ], | ||
| "seed": [ | ||
| 0 | ||
| ], | ||
| "dtype": [ | ||
| "auto" | ||
| ], | ||
| "use-all-available-gpus_" : [] | ||
| "use-all-available-gpus_": [] | ||
| } | ||
| }, | ||
| { | ||
| "description": "VLLM Engine prefill throughput (synthetic)", | ||
| "description": "VLLM Engine throughput - Sparse (with dataset)", | ||
| "models": [ | ||
| "facebook/opt-125m", | ||
| "TinyLlama/TinyLlama-1.1B-Chat-v1.0", | ||
| "mistralai/Mistral-7B-Instruct-v0.2", | ||
| "NousResearch/Llama-2-7b-chat-hf" | ||
| "neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50" | ||
| ], | ||
| "max_model_lens": [ | ||
| 4096 | ||
| ], | ||
| "max_model_lens" : [4096], | ||
| "script_name": "benchmark_throughput", | ||
| "script_args": { | ||
| "input-len": [ | ||
| 1, | ||
| 16, | ||
| 32, | ||
| 64, | ||
| 128, | ||
| 256, | ||
| 512, | ||
| 1024 | ||
| "dataset": [ | ||
| "sharegpt" | ||
| ], | ||
| "output-len": [ | ||
| 1 | ||
| 128 | ||
| ], | ||
| "num-prompts": [ | ||
| 1 | ||
| 1000 | ||
| ], | ||
| "sparsity": [ | ||
| "sparse_w16a16" | ||
| ], | ||
| "use-all-available-gpus_" : [] | ||
| "use-all-available-gpus_": [] | ||
| } | ||
| }, | ||
| { | ||
| "description": "VLLM Engine decode throughput (synthetic)", | ||
| "description": "VLLM Engine throughput - 2:4 Sparse (with dataset)", | ||
| "models": [ | ||
| "facebook/opt-125m", | ||
| "TinyLlama/TinyLlama-1.1B-Chat-v1.0", | ||
| "mistralai/Mistral-7B-Instruct-v0.2", | ||
| "NousResearch/Llama-2-7b-chat-hf" | ||
| "neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4" | ||
| ], | ||
| "max_model_lens": [ | ||
| 4096 | ||
| ], | ||
| "max_model_lens" : [4096], | ||
| "script_name": "benchmark_throughput", | ||
| "script_args": { | ||
| "input-len": [ | ||
| 2 | ||
| "dataset": [ | ||
| "sharegpt" | ||
| ], | ||
| "output-len": [ | ||
| 128 | ||
| ], | ||
| "num-prompts": [ | ||
| 1, | ||
| 4, | ||
| 8, | ||
| 16, | ||
| 32, | ||
| 64 | ||
| 1000 | ||
| ], | ||
| "sparsity": [ | ||
| "semi_structured_sparse_w16a16" | ||
| ], | ||
| "use-all-available-gpus_" : [] | ||
| "use-all-available-gpus_": [] | ||
| } | ||
| } | ||
| ] | ||
| } | ||
| } |
94 changes: 94 additions & 0 deletions
94
neuralmagic/benchmarks/configs/benchmark_throughput_decode.json
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,94 @@ | ||
| { | ||
| "configs": [ | ||
| { | ||
| "description": "VLLM Engine decode throughput - Dense (synthetic)", | ||
| "models": [ | ||
| "teknium/OpenHermes-2.5-Mistral-7B", | ||
| "neuralmagic/OpenHermes-2.5-Mistral-7B-marlin", | ||
| "TheBloke/OpenHermes-2.5-Mistral-7B-GPTQ", | ||
| "NousResearch/Llama-2-7b-chat-hf" | ||
| ], | ||
| "max_model_lens": [ | ||
| 4096 | ||
| ], | ||
| "script_name": "benchmark_throughput", | ||
| "script_args": { | ||
| "input-len": [ | ||
| 2 | ||
| ], | ||
| "output-len": [ | ||
| 128 | ||
| ], | ||
| "num-prompts": [ | ||
| 1, | ||
| 4, | ||
| 8, | ||
| 16, | ||
| 32, | ||
| 64 | ||
| ], | ||
| "use-all-available-gpus_": [] | ||
| } | ||
| }, | ||
| { | ||
| "description": "VLLM Engine decode throughput - Sparse (synthetic)", | ||
| "models": [ | ||
| "neuralmagic/OpenHermes-2.5-Mistral-7B-pruned50" | ||
| ], | ||
| "max_model_lens": [ | ||
| 4096 | ||
| ], | ||
| "script_name": "benchmark_throughput", | ||
| "script_args": { | ||
| "input-len": [ | ||
| 2 | ||
| ], | ||
| "output-len": [ | ||
| 128 | ||
| ], | ||
| "num-prompts": [ | ||
| 1, | ||
| 4, | ||
| 8, | ||
| 16, | ||
| 32, | ||
| 64 | ||
| ], | ||
| "sparsity": [ | ||
| "sparse_w16a16" | ||
| ], | ||
| "use-all-available-gpus_": [] | ||
| } | ||
| }, | ||
| { | ||
| "description": "VLLM Engine decode throughput - 2:4 Sparse (synthetic)", | ||
| "models": [ | ||
| "neuralmagic/OpenHermes-2.5-Mistral-7B-pruned2.4" | ||
| ], | ||
| "max_model_lens": [ | ||
| 4096 | ||
| ], | ||
| "script_name": "benchmark_throughput", | ||
| "script_args": { | ||
| "input-len": [ | ||
| 2 | ||
| ], | ||
| "output-len": [ | ||
| 128 | ||
| ], | ||
| "num-prompts": [ | ||
| 1, | ||
| 4, | ||
| 8, | ||
| 16, | ||
| 32, | ||
| 64 | ||
| ], | ||
| "sparsity": [ | ||
| "semi_structured_sparse_w16a16" | ||
| ], | ||
| "use-all-available-gpus_": [] | ||
| } | ||
| } | ||
| ] | ||
| } |
Oops, something went wrong.