Skip to content

Pull requests: pytorch/ao

New pull request New
69 Open 720 Closed
69 Open 720 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
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

[low-bit optim] Fix load state dict when device is different CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1021 opened Oct 5, 2024 by gau-nernst Loading…
4
Add generic fake quantized linear for QAT CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1020 opened Oct 4, 2024 by andrewor14 Loading…
1
Make module swap the main QAT flow again CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1019 opened Oct 4, 2024 by andrewor14 Loading…
1
Add quantized embedding kernels to torchao CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. fb-exported
#1018 opened Oct 4, 2024 by metascroy Loading…
4
Dynamic Float8 benchmarking llama CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1017 opened Oct 4, 2024 by jainapurva Loading…
8
[not for land] float8 training -> quantize_ CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1016 opened Oct 4, 2024 by vkuzo Loading…
18
Re-run regression tests with CUDA_LAUNCH_BLOCKING CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1009 opened Oct 4, 2024 by malfet • Draft
1
Temporary Fix: Skip TestAffineQuantizedTensorParallel on H100 CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1001 opened Oct 3, 2024 by jainapurva • Draft
4
Enable ROCM in CI ciflow/rocm CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: rocm
#999 opened Oct 3, 2024 by msaroufim Loading…
2
Kleidi 4b blockwise gemv prototype CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#997 opened Oct 2, 2024 by digantdesai • Draft
1
Subclass API (#966) CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. fb-exported
#995 opened Oct 2, 2024 by metascroy Loading…
6
manual expert from fbcode CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#993 opened Oct 2, 2024 by y-sq • Draft
1
add "_gemm_input_role" to dunder slots CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#984 opened Oct 1, 2024 by crcrpar Loading…
2
[wip] SpinQuant CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#983 opened Oct 1, 2024 by tobiasvanderwerff Loading…
4 of 6 tasks
8
[autoquant] Fix the autoquant multi_head_attention torch_function dispatch CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#977 opened Sep 30, 2024 by IvanKobzarev Loading…
2
cleaned up and tested tp support CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#976 opened Sep 30, 2024 by debajyotidatta Loading…
6
Update base.h unit to unsigned int
#962 opened Sep 27, 2024 by EnragedAntelope Loading…
4
Introduce lowbit quantized linear MPS kernels CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. fb-exported
#954 opened Sep 26, 2024 by manuelcandales Loading…
49
float8 training axiswise scaling support with per-gemm-argument configuration CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#940 opened Sep 24, 2024 by vkuzo Loading…
17
add axiswise scaling to Float8Linear CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#920 opened Sep 23, 2024 by vkuzo Loading…
11
add axiswise granularity to Float8Tensor CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#919 opened Sep 23, 2024 by vkuzo Loading…
11
[wip] fp8 + 24sparse benchmarking CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#915 opened Sep 22, 2024 by jcaip • Draft
1
[float8] fuse abs/max with torch.linalg.vector_norm CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#905 opened Sep 19, 2024 by weifengpy • Draft
4
float8 profile script: add activation checkpointing CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#892 opened Sep 16, 2024 by vkuzo Loading…
1
Add blocksparse_int_addmm. Eliminate unnecessary contiguous calls which leads to performance increase. CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#891 opened Sep 16, 2024 by pearu • Draft
1
ProTip! no:milestone will show everything without a milestone.