Commits

Breadcrumbs

History for

onmaster

User selector

All users

All time

Commit History

Commits on Apr 27, 2025

• Fix compile error for nv_bloat162 (#7248)

  Show description for ee492c3
  

  loscrossos

  and

  loadams

  authored
  ee492c3

  Copy full SHA for ee492c3

Commits on Nov 4, 2024

• [Bug Fix] Support threads_per_head < 64 for wavefront size of 64 (#6622)

  Show description for 2b41d62
  

  jagadish-amd

  and

  loadams

  authored
  2b41d62

  Copy full SHA for 2b41d62

Commits on Oct 9, 2024

• Rearrange inference OPS and stop using builder.load (#5490)

  Show description for 645639b
  

  5 peopleauthored
  645639b

  Copy full SHA for 645639b

Commits on Jun 24, 2024

• Fixed Windows inference build. (#5609)

  Show description for b3767d0
  

  4 peopleauthored
  b3767d0

  Copy full SHA for b3767d0

Commits on May 17, 2024

• Rocm warp size fix (#5402)

  Show description for 76c9c69
  

  rraminen

  and

  loadams

  authored
  76c9c69

  Copy full SHA for 76c9c69

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

  Show description for d3dd8e7
  

  rraminen

  and

  loadams

  authored
  d3dd8e7

  Copy full SHA for d3dd8e7

Commits on Nov 3, 2023

• DeepSpeed-FastGen (#4604)

  Show description for 38b41df
  

  6 peopleauthored
  38b41df

  Copy full SHA for 38b41df

Commits on Oct 19, 2023

• Switch from HIP_PLATFORM_HCC to HIP_PLATFORM_AMD (#4539)

  Show description for c7724c6
  loadams

  authored
  c7724c6

  Copy full SHA for c7724c6

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

  Show description for beed962
  

  3 peopleauthored
  beed962

  Copy full SHA for beed962

Commits on Sep 19, 2023

• Add the policy to run llama model from the official repo (#4313)

  Show description for 468882f
  

  3 peopleauthored
  468882f

  Copy full SHA for 468882f

Commits on Sep 14, 2023

• Fix a bug in the implementation of dequantization for inference (#3433)

  Show description for 9bf7778
  

  3 peopleauthored
  9bf7778

  Copy full SHA for 9bf7778

Commits on Jul 6, 2023

• Avoid deprecation warnings in `CHECK_CUDA` (#3854)

  Show description for 9aeba94
  

  3 peopleauthored
  9aeba94

  Copy full SHA for 9aeba94

Commits on Jun 23, 2023

• Fix Bug in transform.cu (#3534)

  Show description for aebdfb3
  

  2 peopleauthored and

  jeffra

  committed
  aebdfb3

  Copy full SHA for aebdfb3

Commits on Jun 13, 2023

• Fix output transpose dimension bugs (#3747)

  loadams

  authored
  cd911f9

  Copy full SHA for cd911f9

• Include cublas error details when getting cublas handle fails (#3695)

  Show description for 46bb08c
  

  jli

  and

  loadams

  authored
  46bb08c

  Copy full SHA for 46bb08c

Commits on May 18, 2023

• fix_typo (#3559)

  inkcherry

  authored
  d1c3c0d

  Copy full SHA for d1c3c0d

Commits on May 16, 2023

• [ROCm] Hip headers fix (#3532)

  Show description for 5147b90
  

  rraminen

  and

  loadams

  authored
  5147b90

  Copy full SHA for 5147b90

Commits on May 8, 2023

• fix spelling error (#3482)

  Show description for f1fab90
  

  luliyucoordinate

  and

  loadams

  authored
  f1fab90

  Copy full SHA for f1fab90

Commits on May 4, 2023

• Hybrid Engine Refactor and Llama Inference Support (#3425)

  Show description for 0a61d5d
  

  cmikeh2

  and

  jeffra

  authored
  0a61d5d

  Copy full SHA for 0a61d5d

Commits on Apr 22, 2023

• add bf16 cuda kernel support (#3092)

  Show description for 297cd9e
  

  dc3671

  and

  tjruwase

  authored
  297cd9e

  Copy full SHA for 297cd9e

Commits on Apr 21, 2023

• Fix formatting (#3343)

  Show description for 4a3ca4e
  mrwyattii

  authored
  4a3ca4e

  Copy full SHA for 4a3ca4e

• Fix missing scale attributes for GPTJ (#3256)

  Show description for 145c3a7
  

  3 peopleauthored
  145c3a7

  Copy full SHA for 145c3a7

Commits on Apr 11, 2023

• DeepSpeed Chat (#3186)

  Show description for 47f9f13
  

  9 peopleauthored
  47f9f13

  Copy full SHA for 47f9f13

Commits on Apr 5, 2023

• Optimize Softmax Kernel (#3112)

  Show description for e73de8c
  molly-smith

  authored
  e73de8c

  Copy full SHA for e73de8c

Commits on Mar 31, 2023

• Update DeepSpeed copyright license to Apache 2.0 (#3111)

  Show description for b361c72
  

  mrwyattii

  and

  jeffra

  authored
  b361c72

  Copy full SHA for b361c72

Commits on Mar 22, 2023

• Softmax Scheduling Cleanup (#3046)

  Show description for 1286e37
  

  cmikeh2

  and

  jeffra

  authored
  1286e37

  Copy full SHA for 1286e37

• fix return prev key and value , added strides to from_blob (#2828)

  Show description for 871c8a3
  

  4 peopleauthored
  871c8a3

  Copy full SHA for 871c8a3

Commits on Mar 21, 2023

• [docs] add MCR-DL paper to readme/docs (#3066)

  Quentin-Anthony

  authored
  b38b303

  Copy full SHA for b38b303

Commits on Feb 27, 2023

• add missing license info to top of all source code (#2889)

  Show description for da84e60
  

  4 peopleauthored
  da84e60

  Copy full SHA for da84e60

Commits on Feb 16, 2023

• Fix gpt-Neox rotary embedding implementation (#2782)

  Show description for 5b7413a
  

  3 peopleauthored
  5b7413a

  Copy full SHA for 5b7413a

• Port Reza's INT8-quantization fix to container architecture (#2725)

  Show description for fd1449c
  

  5 peopleauthored
  fd1449c

  Copy full SHA for fd1449c

Commits on Feb 3, 2023

• Reset KV-cache at the beginning of text-generation (#2669)

  Show description for 9f41ffe
  

  3 peopleauthored
  9f41ffe

  Copy full SHA for 9f41ffe

Commits on Jan 26, 2023

• Fix softmax backward (#2709)

  Show description for 0b06e0c
  

  3 peopleauthored
  0b06e0c

  Copy full SHA for 0b06e0c

Commits on Dec 13, 2022

• Move layer norm to new schedule (#2590)

  Show description for 3a3dfe6
  

  lokoppakmsft

  and

  cmikeh2

  authored
  3a3dfe6

  Copy full SHA for 3a3dfe6

Commits on Dec 6, 2022

• Fix quantized-inference & Add generic support of checkpoint loading (#2547)

  Show description for 35b350b
  

  RezaYazdaniAminabadi

  and

  mrwyattii

  authored
  35b350b

  Copy full SHA for 35b350b