Commits

Breadcrumbs

History for

onmaster

User selector

All users

All time

Commit History

Commits on Apr 24, 2024

• Add nccl_allocator for zero-copy user buffer (#1796)

  Show description for 6038fc1
  Aidyn-A

  authored
  6038fc1

  Copy full SHA for 6038fc1

Commits on Feb 7, 2024

• Add GPUDirect Storage (#1774)

  Show description for 5b67cd5
  Aidyn-A

  authored
  5b67cd5

  Copy full SHA for 5b67cd5

Commits on Nov 14, 2023

• A fused `apply_rotary_pos_emb` implementation for Megatron-Core (#1746)

  Show description for 08f7402
  yaox12

  authored
  08f7402

  Copy full SHA for 08f7402

Commits on Sep 30, 2023

• Add hysteresis support for AMP gradient scale update (#1733)

  Show description for 6a77872
  

  4 peopleauthored
  6a77872

  Copy full SHA for 6a77872

Commits on Sep 4, 2023

• [contrib.xentropy] bfloat16 support (#1720)

  Show description for 730376f
  

  crcrpar

  and

  IvanYashchuk

  authored
  730376f

  Copy full SHA for 730376f

Commits on Jul 25, 2023

• Fast CUDA NHWC Group Norm (#1695)

  Show description for 50ac842
  

  alpha0422

  and

  crcrpar

  authored
  50ac842

  Copy full SHA for 50ac842

Commits on Jun 12, 2023

• Add `pyproject.toml` (#1669)

  Show description for 2d8302a
  crcrpar

  authored
  2d8302a

  Copy full SHA for 2d8302a

Commits on Mar 7, 2023

• [FMHA] Remove zero fill for softmax output (#1565)

  Show description for b1c7600
  vasunvidia

  authored
  b1c7600

  Copy full SHA for b1c7600

Commits on Feb 9, 2023

• decrease warnings & remove append_nvcc_threads from setup.py (#1581)

  Show description for 0145d69
  crcrpar

  authored
  0145d69

  Copy full SHA for 0145d69

Commits on Jan 26, 2023

• bump cutlass to v2.11.0 for contrib/multihead_attn (#1570)

  Show description for 536e549
  crcrpar

  authored
  536e549

  Copy full SHA for 536e549

Commits on Nov 22, 2022

• Update megatron fused softmax follow megatron-lm (#1539)

  Show description for abeca58
  yaoyu-33

  authored
  abeca58

  Copy full SHA for abeca58

Commits on Nov 14, 2022

• [README] Add a list of install options (#1492)

  Show description for a9950b1
  crcrpar

  authored
  a9950b1

  Copy full SHA for a9950b1

Commits on Oct 6, 2022

• check `CUDA_HOME` before using it (#1499)

  Show description for b213772
  crcrpar

  authored
  b213772

  Copy full SHA for b213772

• Update for CUDA 11.8 (#1502)

  Show description for 5f1a73e
  

  3 peopleauthored
  5f1a73e

  Copy full SHA for 5f1a73e

Commits on Sep 14, 2022

• Skip test if related contrib modules are not found (#1483)

  Show description for 5ba7af1
  crcrpar

  authored
  5ba7af1

  Copy full SHA for 5ba7af1

Commits on Sep 8, 2022

• [contrib] cuDNN based GBN (#1476)

  Show description for 7a12ee6
  

  crcrpar

  and

  nvcastet

  authored
  7a12ee6

  Copy full SHA for 7a12ee6

Commits on Aug 5, 2022

• Add a generic fused softmax (#1440)

  Show description for cd0a1f1
  yidong72

  authored
  cd0a1f1

  Copy full SHA for cd0a1f1

Commits on Aug 1, 2022

• add customized fused op index mulitiplication (#1438)

  Show description for a4f4258
  

  BaoHhhhhhan

  and

  Han Bao

  authored
  a4f4258

  Copy full SHA for a4f4258

Commits on Jul 7, 2022

• Remove `pyprof` and `reparameterization` (#1404)

  Show description for 8a7a332
  crcrpar

  authored
  8a7a332

  Copy full SHA for 8a7a332

Commits on Apr 21, 2022

• Give Some Extensions Version Guard in Build&Runtime (#1358)

  Show description for f9305e7
  crcrpar

  authored
  f9305e7

  Copy full SHA for f9305e7

Commits on Apr 20, 2022

• Merge pull request #1340 from NVIDIA/peer_memory

  Show description for fed20d2
  thorjohnsen

  authored
  fed20d2

  Copy full SHA for fed20d2

Commits on Apr 19, 2022

• [submodule update] Bump cudnn-frontend to v0.6.1 (#1353)

  Show description for d89f5e6
  crcrpar

  authored
  d89f5e6

  Copy full SHA for d89f5e6

Commits on Apr 5, 2022

• Rename nccl_p2p extension to nccl_p2p_cuda

  thorjohnsen

  committed
  d8db8c1

  Copy full SHA for d8db8c1

• Rename peer_memory extension to peer_memory_cuda

  thorjohnsen

  committed
  6e7e2d9

  Copy full SHA for 6e7e2d9

Commits on Mar 30, 2022

• Conv-Bias-ReLU fusion (#1332)

  Show description for 23cfb57
  

  3 peopleauthored
  23cfb57

  Copy full SHA for 23cfb57

Commits on Mar 25, 2022

• Add bottleneck block

  thorjohnsen

  committed
  3ade5b2

  Copy full SHA for 3ade5b2

Commits on Mar 24, 2022

• Add CUDA Focal Loss Implementation (#1337)

  Show description for 28f8539
  

  crcrpar

  and

  alpha0422

  authored
  28f8539

  Copy full SHA for 28f8539

Commits on Mar 23, 2022

• Peer memory halo exchange

  thorjohnsen

  committed
  40a0e02

  Copy full SHA for 40a0e02

Commits on Feb 27, 2022

• build fused grad accum w/ wgrad only if cuda>10 (#1312)

  crcrpar

  authored
  47c269b

  Copy full SHA for 47c269b

Commits on Feb 26, 2022

• [transformer] Fuse grad accumulation with wgrad (#1297)

  Show description for ddc0803
  

  crcrpar

  and

  erhoo82

  authored
  ddc0803

  Copy full SHA for ddc0803

Commits on Feb 10, 2022

• 8.6 requires CUDA 11.1 (#1289)

  crcrpar

  authored
  e1aa1fc

  Copy full SHA for e1aa1fc

Commits on Feb 1, 2022

• Add the permutation related support as the extension for asp lib. (#1194)

  Show description for 89edb81
  ChongyuNVIDIA

  authored
  89edb81

  Copy full SHA for 89edb81

Commits on Jan 19, 2022

• pass flags to transducer joint kernel (#1273)

  crcrpar

  authored
  c4e85f7

  Copy full SHA for c4e85f7

Commits on Jan 13, 2022

• support new path to CUDAGeneratorImpl.h (#1267)

  shintaro-iwasaki

  authored
  b2fdf9c

  Copy full SHA for b2fdf9c

Commits on Dec 16, 2021

• version guard (#1253)

  crcrpar

  authored
  e847382

  Copy full SHA for e847382