Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Jul 6, 2023

• remove mpi_cxx from multi-gpu build for now (NVIDIA#705)

  dwyatte

  committedJul 6, 2023
  f8e42aa

Commits on Jun 29, 2023

• Support size_per_head=112 (NVIDIA#660)

  dskhudia

  committedJun 29, 2023
  7777ff1

Commits on Jun 26, 2023

• fix: swap tensor bug (NVIDIA#683)

  PerkzZheng

  committedJun 26, 2023
  eb9b81b

Commits on Jun 21, 2023

• [bugfix] Fix 2-shot All Reduce correctness issue (indexing bug). (NVIDIA#672)

  rkindi

  committedJun 21, 2023
  1cf9b51

Commits on May 1, 2023

• Fix/gpt early stop (NVIDIA#584)

  byshiue

  committedMay 1, 2023
  c6e8f60

Commits on Apr 24, 2023

• perf(bloom): improve performance of huggingface_bloom_convert.py, decrease the time cost and the mem using (NVIDIA#568)

  

  Yangruipis

  and

  r.yang

  committedApr 24, 2023
  19b2956

• [Enhancement]create huggingface_gptneox_convert.py (NVIDIA#569)

  zhang-ge-hao

  committedApr 24, 2023
  3460e20

Commits on Apr 20, 2023

• Update unfused_attention_kernels.cu

  byshiue

  committedApr 20, 2023
  d7ccf83

Commits on Apr 19, 2023

• fix overflow in softmax_kernel when process long seqlen and big batch_size (NVIDIA#524)

  zhangxin81

  committedApr 19, 2023
  adb21c3

Commits on Apr 18, 2023

• Update cublasMMWrapper.cc

  byshiue

  committedApr 18, 2023
  c6ba315

• Update cublasMMWrapper.cc

  byshiue

  committedApr 18, 2023
  a6ef7af

• [Enhancement]add pytorch backend support for gptneox (NVIDIA#550)

  zhang-ge-hao

  committedApr 18, 2023
  169b8df

Commits on Apr 17, 2023

• Update T5DecodingWeight.cc

  byshiue

  committedApr 17, 2023
  0c12805

Commits on Apr 6, 2023

• fix: fix bug of gpt buffer and gpt gemm overflow

  byshiue

  committedApr 6, 2023
  4402759

Commits on Mar 29, 2023

• Update gpt_guide.md (NVIDIA#529)

  Ying1123

  committedMar 29, 2023
  bc4139e

Commits on Mar 17, 2023

• fix: gpt tensor shapes inconsistency (NVIDIA#505)

  zhang-ge-hao

  committedMar 17, 2023
  e838426

Commits on Mar 14, 2023

• fix: change int of some kernels to int64_t to prevent overflow

  byshiue

  committedMar 14, 2023
  bb94e2d

Commits on Mar 10, 2023

• Update beam_search_topk_kernels.cu

  byshiue

  committedMar 10, 2023
  72d3dce

Commits on Mar 6, 2023

• Merge branch 'main' of https://github.com/NVIDIA/FasterTransformer into main

  byshiue

  committedMar 6, 2023
  303e052

• fix: fix custom_ar_kernel threshold

  byshiue

  committedMar 6, 2023
  f7a4418

Commits on Feb 28, 2023

• fix: scale input bug inside AddBiasResidualLayerNorm kernel (NVIDIA#466)

  zobinHuang

  committedFeb 28, 2023
  8bf96eb

Commits on Feb 27, 2023

• [fix] fix integer overflow error (NVIDIA#469)

  luliyucoordinate

  committedFeb 27, 2023
  5b9a2ce

Commits on Feb 24, 2023

• Update SelfAttentionFP8Layer.cc

  byshiue

  committedFeb 24, 2023
  38847d0

Commits on Feb 21, 2023

• Feat/perf opt (NVIDIA#462)

  byshiue

  committedFeb 21, 2023
  fba7567

• Fix Japanese Hugging face GPT conversion (NVIDIA#460)

  noppayut

  committedFeb 21, 2023
  43ea4f3

Commits on Feb 20, 2023

• fix: link for the gpt-j slim weights (NVIDIA#458)

  f0rmiga

  committedFeb 20, 2023
  a04b6ce

Commits on Feb 15, 2023

• Feat/cuda 12 support (NVIDIA#450)

  byshiue

  committedFeb 15, 2023
  35989aa

Commits on Feb 14, 2023

• update t5 model conversion script to numpy mode (NVIDIA#448)

  

  Qing Lan

  committedFeb 14, 2023
  2144e29

Commits on Feb 13, 2023

• [Improvement] accelerate T5 model conversion and fix bloom model on multi-process (NVIDIA#447)

  

  Qing Lan

  committedFeb 13, 2023
  9b6d718

Commits on Feb 10, 2023

• place multi-processing init to main method (NVIDIA#443)

  

  Qing Lan

  committedFeb 10, 2023
  7634698

Commits on Feb 8, 2023

• support gated_activation on flant5 series model (NVIDIA#441)

  lanking520

  committedFeb 8, 2023
  fff55e7

• Add Layernorm option for Wenet trt (NVIDIA#434)

  

  yuekaizhang

  and

  root

  committedFeb 8, 2023
  3357ef6

Commits on Feb 6, 2023

• Update: set fmha, and context_qk_half_accumulation as optional

  byshiue

  committedFeb 6, 2023
  7921074

• fix: malloc host opt

  byshiue

  committedFeb 6, 2023
  90cc5a7

• fix: fix perf issue of small model due to calling fmtstr

  byshiue

  committedFeb 6, 2023
  652e61e

Pagination

PreviousNext