Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Dec 27, 2025

• [PyTorch] Fuse permute+pad and unpermute+unpad ops for FP8 optimization (#1921)

  Show description for 5ba01fa
  

  3 peopleauthored
  5ba01fa

  Copy full SHA for 5ba01fa

Commits on Dec 22, 2025

• Fix ptxas compilation on sm103 for triton kernels (#2539)

  Show description for 97a09c2
  tdophung

  authored
  97a09c2

  Copy full SHA for 97a09c2

Commits on Dec 20, 2025

• [PyTorch][NVFP4][MOE] NVFP4 Grouped Quantize with Hadamard Transform (#2411)

  Show description for eb8e792
  

  3 peopleauthored
  eb8e792

  Copy full SHA for eb8e792

• [JAX] Remove unused TE DPA module dtype which fixes cuDNN backend detection to properly use input dtypes (#2485)

  Show description for 47902e9
  

  jberchtold-nvidia

  and

  pre-commit-ci[bot]

  authored
  47902e9

  Copy full SHA for 47902e9

Commits on Dec 19, 2025

• [JAX] Handle meshs set with jax.set_mesh (#2532)

  Show description for d46d5db
  

  jberchtold-nvidia

  and

  pre-commit-ci[bot]

  authored
  d46d5db

  Copy full SHA for d46d5db

• [PyTorch] Make sure Float8Tensor.contiguous supports autograd (#2533)

  Show description for 6fd6209
  

  3 peopleauthored
  6fd6209

  Copy full SHA for 6fd6209

Commits on Dec 18, 2025

• ci: Use whitelisted sha for `get-release` (#2531)

  Show description for 3e69397
  ko3n1g

  authored
  3e69397

  Copy full SHA for 3e69397

• Fix meta device check failure when passing torch.device objects (#2519)

  Show description for 14ddb43
  

  3 peopleauthored
  14ddb43

  Copy full SHA for 14ddb43

Commits on Dec 17, 2025

• [JAX] Add tutorial for integrating TE/JAX quantization into an existing framework (#2423)

  Show description for 442513c
  

  jberchtold-nvidia

  and

  pre-commit-ci[bot]

  authored
  442513c

  Copy full SHA for 442513c

• Add ccache support to TE and use it in GitHub actions (#2444)

  Show description for 5c2f2ff
  ptrendx

  authored
  5c2f2ff

  Copy full SHA for 5c2f2ff

• Reset cache logic of weight workspace for NVFP4TensorStorage (#2524)

  Show description for dbd0197
  jinhangchoi

  authored
  dbd0197

  Copy full SHA for dbd0197

Commits on Dec 16, 2025

• Remove test skip logic for GEMM-AR tests (#2516)

  Show description for eac8af6
  vcherepanov-nv

  authored
  eac8af6

  Copy full SHA for eac8af6

Commits on Dec 15, 2025

• [PyTorch debug] Fix test for debug tools (#2507)

  Show description for 2886cbc
  

  pggPL

  and

  timmoon10

  authored
  2886cbc

  Copy full SHA for 2886cbc

• Check calling convention for amax switch. (#2506)

  Show description for b215116
  kwyss-nvidia

  authored
  b215116

  Copy full SHA for b215116

• fix ce loss calculation when some tokens are ignored (#2476)

  Show description for 36f2dfd
  

  yashaswikarnati

  and

  pre-commit-ci[bot]

  authored
  36f2dfd

  Copy full SHA for 36f2dfd

Commits on Dec 12, 2025

• [PyTorch] Add triton requirement (#2490)

  Show description for 8c9f7c2
  

  ksivaman

  and

  tdophung

  authored
  8c9f7c2

  Copy full SHA for 8c9f7c2

Commits on Dec 11, 2025

• [JAX] Unset NVTE_FUSED_RING_ATTENTION_USE_SCAN by default (#2503)

  Show description for 887a4fc
  

  KshitijLakhani

  and

  pre-commit-ci[bot]

  authored
  887a4fc

  Copy full SHA for 887a4fc

• [PyTorch] Convert sample tuple to list in cudagraph input reuse (#2426)

  Show description for 5035232
  

  buptzyb

  and

  ksivaman

  authored
  5035232

  Copy full SHA for 5035232

• [PyTorch] Update RNG global states in tracker set_states (#2501)

  Show description for 811e090
  

  buptzyb

  and

  ksivaman

  authored
  811e090

  Copy full SHA for 811e090

• Add separate RNG states for column-wise quantization with Stochastic Rounding (#2487)

  Show description for a5694f2
  negvet

  authored
  a5694f2

  Copy full SHA for a5694f2

Commits on Dec 10, 2025

• [PyTorch] Add THD support for max_logit/MuonClip (#2480)

  Show description for 93c5c65
  cyanguwa

  authored
  93c5c65

  Copy full SHA for 93c5c65

• [PyTorch Debug] Add nvdlfw-inspect to dependencies (#2173)

  Show description for e411547
  

  pggPL

  and

  pre-commit-ci[bot]

  authored
  e411547

  Copy full SHA for e411547

• [JAX] Make softmax_type in FFI optional (#2491)

  Show description for 5afbb0e
  

  jberchtold-nvidia

  and

  pre-commit-ci[bot]

  authored
  5afbb0e

  Copy full SHA for 5afbb0e

Commits on Dec 9, 2025

• Jax primitives for permutation on single GPU (#2473)

  Show description for 46c6ef3
  

  3 peopleauthored
  46c6ef3

  Copy full SHA for 46c6ef3

• Fix the sm120 compilation with CUDA 12 (#2482)

  Show description for dbaa02d
  ptrendx

  authored
  dbaa02d

  Copy full SHA for dbaa02d

• [PyTorch] Change order of args in another permutation triton kernel (#2488)

  Show description for e05f87e
  tdophung

  authored
  e05f87e

  Copy full SHA for e05f87e

• Fix runtime lib loading logic (#2297)

  Show description for 8ef3a33
  ksivaman

  authored
  8ef3a33

  Copy full SHA for 8ef3a33

Commits on Dec 8, 2025

• [Pytorch][Bug]MXFP8 Split tensor Bug fix (#2427)

  Show description for c09411d
  

  3 peopleauthored
  c09411d

  Copy full SHA for c09411d

• Changed VERSION to 2.12.0.dev0

  Show description for fd91bae
  ptrendx

  committed
  fd91bae

  Copy full SHA for fd91bae

Commits on Dec 6, 2025

• [JAX] Add CP + THD + AG + Striped>1 + SWA support (#2379)

  Show description for fd0cd12
  

  4 peopleauthored
  fd0cd12

  Copy full SHA for fd0cd12

Commits on Dec 5, 2025

• Fix bugs from refactoring C++ tensor class (#2481)

  Show description for f0572aa
  timmoon10

  authored
  f0572aa

  Copy full SHA for f0572aa

• [JAX] Enable TE/JAX test timings in CI (#2475)

  Show description for 50be029
  jberchtold-nvidia

  authored
  50be029

  Copy full SHA for 50be029

Commits on Dec 4, 2025

• [Core] Fix inconsistent logic in C++ tensor class (#2330)

  Show description for 6182206
  

  3 peopleauthored
  6182206

  Copy full SHA for 6182206

Commits on Dec 2, 2025

• Add primary weighs fp8 support for mxfp8 (#2055)

  Show description for d126cdd
  

  3 peopleauthored
  d126cdd

  Copy full SHA for d126cdd

• [JAX] Make test_layer.py tolerances stricter (#2306)

  Show description for cc42a57
  jberchtold-nvidia

  authored
  cc42a57

  Copy full SHA for cc42a57

Pagination

Previous

Next