Skip to content

Pull requests: pytorch/ao

New pull request New
312 Open 3,200 Closed
312 Open 3,200 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

[fp8 training] support linear op overrides in Float8TrainingWeightWrapperTensor CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. float8 module: training quantize_ api training flow
#4325 opened Apr 23, 2026 by danielvegamyhre Contributor Loading… FP8 Rowwise Training
4
fix(fp8_fa3): remove broken import and fix setup_fp8_fa3 signature
#4324 opened Apr 23, 2026 by Anai-Guo Loading…
1
2
nvfp4 gptq: prep for bmm and grouped_mm CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4323 opened Apr 23, 2026 by vkuzo Contributor Loading…
2
install debug expert token counters on nvfp4 moe test script CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: not user facing Use this tag if you don't want this PR to show up in release notes
#4322 opened Apr 23, 2026 by vkuzo Contributor Loading…
2
docs: fix broken quantize_ link in first_quantization_example topic: documentation Use this tag if this PR adds or improves documentation
#4317 opened Apr 23, 2026 by Anai-Guo Loading…
1
6
docs(finetuning): replace TorchTune QAT section with Unsloth
#4312 opened Apr 21, 2026 by Anai-Guo Loading…
3 tasks
1
2
[PT2E][Docs] Calibration is a must for both static and dynamic quantization CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: not user facing Use this tag if you don't want this PR to show up in release notes module: pt2e_quant pt2 export quantization (prepare_pt2e, convert_pt2e, quantizer) topic: documentation Use this tag if this PR adds or improves documentation
#4307 opened Apr 21, 2026 by Xia-Weiwen Collaborator Loading…
7
docs(qat): update supported configs list, fix unclickable link, add blog ref
#4305 opened Apr 21, 2026 by Anai-Guo Loading…
1
2
[test][xpu] Adjust the test sequence for Intel GPU CI ciflow/xpu label used to trigger xpu CI jobs CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: not user facing Use this tag if you don't want this PR to show up in release notes topic: for developers Use this tag if this PR is mainly developer facing xpu Intel XPU related features
#4298 opened Apr 19, 2026 by zxd1997066 Contributor Loading…
4
fix(utils): propagate non_blocking in TorchAOBaseTensor._to_copy and _get_to_kwargs CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4297 opened Apr 19, 2026 by Dev-next-gen Loading…
4
add FSDP and TP tests for Float8BlockwiseLinear CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: training quantize_ api training flow
#4295 opened Apr 17, 2026 by iamzainhuda Contributor Loading… FP8 Blockwise Training
8
Move collect_producer_nodes to graph_utils.py CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4294 opened Apr 17, 2026 by tom-arm Contributor Loading…
1
[mxfp8 moe training] fuse dynamic per-group padding into cutedsl 2d mxfp8 quant kernel CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4293 opened Apr 17, 2026 by MagellaX Contributor Loading…
1
6
basic enablement for mxfp8 and mxfp4 inference on AMD MI350x CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: inference quantize_ api inference flow
#4290 opened Apr 16, 2026 by vkuzo Contributor Loading…
4
[AARCH64] Enable MKLDNN Backend for Int8DynamicActivationInt8WeightConfig() on ARM CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: inference quantize_ api inference flow
#4281 opened Apr 15, 2026 by agrawal-aka Contributor Loading…
2
Add Sparse2x4HIPSPARSELTFloat8Tensor (#4277) CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. fb-exported meta-exported module: rocm
#4277 opened Apr 14, 2026 by bbeckca Contributor Loading…
9
[ROCm] Add MXFP8 training support for gfx950 (MI355X) ciflow/rocm-mi300 CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: rocm
#4275 opened Apr 14, 2026 by indianspeedster • Draft MXFP8 Training
13
[xpu][mx] Fix NaN scale propagation in RCEIL triton kernel CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4271 opened Apr 13, 2026 by ugolowic Contributor Loading…
3
[optim] Add GrokAdamW optimizer with low-bit quantization support CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4270 opened Apr 13, 2026 by vaibhavhariram Loading…
1
3
Add torch.uint16, torch.uint32 CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4269 opened Apr 12, 2026 by Freed-Wu Loading…
1
Add reduce_range to avoid overflow in int8 tensor CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: not user facing Use this tag if you don't want this PR to show up in release notes
#4266 opened Apr 10, 2026 by cyxlily Contributor Loading…
32
Support 32x32 scaling for weights in MXFP8 weight quantization kernel CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: training quantize_ api training flow moe mx
#4254 opened Apr 9, 2026 by alexsamardzic Collaborator Loading… MXFP8 Training
1
@alexsamardzic
34
[nvfp4 training] add autograd support for NVFP4 emulated grouped GEMM CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4252 opened Apr 9, 2026 by roycho96 Contributor Loading…
1
[xpu][mx] Enable mx matmul tests on xpu CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4251 opened Apr 8, 2026 by ugolowic Contributor • Draft
1
[ROCm] Switch CI to allowlist of validated tests CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: not user facing Use this tag if you don't want this PR to show up in release notes topic: rocm
#4248 opened Apr 7, 2026 by brucechanglongxu Contributor Loading…
12
ProTip! Find all pull requests that aren't related to any open issues with -linked:issue.