Commits

Breadcrumbs

History for

onf01134d

User selector

All users

Datepicker

All time

Commit History

Commits on Jul 15, 2024

• [CPU] Allow deepspeed.comm.inference_all_reduce in torch.compile graph (#5604)

  Show description for ec6cbb3
  

  3 peopleauthored
  ec6cbb3

Commits on Jul 1, 2024

• update xpu fusedadam opbuilder for pytorch 2.3 (#5702)

  Show description for e392296
  

  baodii

  and

  loadams

  authored
  e392296

Commits on Jun 28, 2024

• Change source of CPUAdam for xpu accelerator (#5703)

  Show description for 4b8a4a0
  Liangliang-Ma

  authored
  4b8a4a0

Commits on Jun 26, 2024

• [CPU] add fp16 support to shm inference_all_reduce (#5669)

  Show description for 19da95f
  

  3 peopleauthored
  19da95f

Commits on Jun 24, 2024

• Fixed Windows inference build. (#5609)

  Show description for b3767d0
  

  4 peopleauthored
  b3767d0

Commits on Jun 12, 2024

• [CPU] SHM based allreduce improvement for small message size (#5571)

  Show description for eda5075
  

  3 peopleauthored
  eda5075

Commits on Jun 5, 2024

• Add Compressedbackend for Onebit optimizers (#5473)

  Show description for 11a62a0
  

  Liangliang-Ma

  and

  tjruwase

  authored
  11a62a0

Commits on May 31, 2024

• Fixed the Windows build. (#5596)

  Show description for e7dd28a
  

  3 peopleauthored
  e7dd28a

Commits on May 28, 2024

• Rename files in fp_quantize op from quantize.* to fp_quantize.* (#5577)

  Show description for bf66acd
  loadams

  authored
  bf66acd

Commits on May 20, 2024

• CPUAdam fp16 and bf16 support (#5409)

  Show description for 69af361
  

  BacharL

  and

  tjruwase

  authored
  69af361

Commits on May 17, 2024

• Rocm warp size fix (#5402)

  Show description for 76c9c69
  

  rraminen

  and

  loadams

  authored
  76c9c69

• rocblas -> hipblas changes for ROCm (#5401)

  Show description for d3dd8e7
  

  rraminen

  and

  loadams

  authored
  d3dd8e7

Commits on Apr 22, 2024

• 64bit indexing fused adam (#5187)

  Show description for 9b6ef9e
  

  3 peopleauthored
  9b6ef9e

Commits on Apr 19, 2024

• Selective dequantization (#5375)

  Show description for c632ea0
  

  3 peopleauthored
  c632ea0

Commits on Apr 17, 2024

• [CPU] Support SHM based inference_all_reduce in TorchBackend (#5391)

  Show description for b22706a
  

  3 peopleauthored
  b22706a

Commits on Apr 10, 2024

• Optimize the fp-dequantizer to get high memory-BW utilization (#5373)

  Show description for a8b8215
  

  RezaYazdaniAminabadi

  and

  sfc-gh-reyazda

  authored
  a8b8215

Commits on Apr 4, 2024

• CPU SHM based inference_all_reduce improve (#5320)

  Show description for 731fd68
  

  3 peopleauthored
  731fd68

• FP [6,8,12] quantizer op (#5336)

  Show description for 3fbd01c
  

  jeffra

  and

  sfc-gh-reyazda

  authored
  3fbd01c

Commits on Feb 12, 2024

• [NPU]ZeRO-Infinity feature compatibility (#5077)

  Show description for 0d9cfa0
  

  3 peopleauthored
  0d9cfa0

Commits on Jan 10, 2024

• Fix confusing width in simd_load (#4714)

  Show description for f517903
  

  yzhblind

  and

  loadams

  authored
  f517903

Commits on Jan 8, 2024

• [Fix] Fix cpu inference UT failure (#4430)

  Show description for d8d865f
  

  10 peopleauthored
  d8d865f

Commits on Jan 5, 2024

• [NPU]Add ZeRO-Infinity feature for NPU (#4809)

  Show description for b596963
  

  misstek

  and

  tjruwase

  authored
  b596963

• [XPU] XPU accelerator support for Intel GPU device (#4547)

  Show description for f4f3131
  

  7 peopleauthored
  f4f3131

Commits on Jan 2, 2024

• [NPU] Fix npu offload bug (#4883)

  Show description for d873ce6
  

  4 peopleauthored
  d873ce6

Commits on Nov 14, 2023

• [CANN] Support cpu offload optimizer for Ascend NPU (#4568)

  Show description for c1ba6a1
  

  hipudding

  and

  tjruwase

  authored
  c1ba6a1

Commits on Nov 9, 2023

• Fix the openfold training. (#4657)

  Show description for 1d1a20c
  

  cctry

  and

  conglongli

  authored
  1d1a20c

Commits on Nov 3, 2023

• DeepSpeed-FastGen (#4604)

  Show description for 38b41df
  

  6 peopleauthored
  38b41df

Commits on Oct 26, 2023

• fix multiple definition while building evoformer (#4556)

  Show description for 8f168c2
  

  fecet

  and

  conglongli

  authored
  8f168c2

Commits on Oct 23, 2023

• Add missing RocBlas include (#4557)

  loadams

  authored
  869629c

Commits on Oct 19, 2023

• Switch from HIP_PLATFORM_HCC to HIP_PLATFORM_AMD (#4539)

  Show description for c7724c6
  loadams

  authored
  c7724c6

• Added rocblas header (#4538)

  Show description for 3e4a587
  

  rraminen

  and

  loadams

  authored
  3e4a587

• [Bug fix] Add rope_theta for llama config (#4480)

  Show description for beed962
  

  3 peopleauthored
  beed962

Commits on Oct 12, 2023

• [CCLBackend] update API (#4378)

  Show description for 4fc181b
  

  Liangliang-Ma

  and

  tjruwase

  authored
  4fc181b

Commits on Oct 11, 2023

• adding 8bit dequantization kernel for asym fine-grained block quantization in zero-inference (#4450)

  Show description for 6c86ff3
  

  3 peopleauthored
  6c86ff3

Commits on Oct 5, 2023

• feat: add Lion optimizer (#4331)

  Show description for 8e64c3b
  

  enneamer

  and

  loadams

  authored
  8e64c3b

Pagination

PreviousNext