Skip to content

Pull requests: pytorch/ao

New pull request New
110 Open 1,180 Closed
110 Open 1,180 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

Add CUTLASS-based row-wise scaled sparse FP8 kernel CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. float8 sparsity topic: new feature Use this tag if this PR adds a new feature
#1671 opened Feb 5, 2025 by alexsamardzic • Draft
2
Support power of 2 scaling factors in float8 training CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. topic: new feature Use this tag if this PR adds a new feature
#1670 opened Feb 5, 2025 by danielvegamyhre Loading…
12
Feat/blockwise fp8 quant CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1668 opened Feb 5, 2025 by Degnel Loading…
3
Support mixed MX element dtype in mx_mm function and MXLinear. CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. topic: improvement Use this tag if this PR is an improvement (doesn't fit into any of the other categories)
#1667 opened Feb 5, 2025 by balancap Loading…
10
Add boiler plate code to Tensor subclass CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. topic: improvement Use this tag if this PR is an improvement (doesn't fit into any of the other categories)
#1663 opened Feb 4, 2025 by jainapurva Loading…
1
Add mx_fp4_kernel CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. mx topic: new feature Use this tag if this PR adds a new feature
#1661 opened Feb 4, 2025 by drisspg Loading…
1
Tensor parallel support for fpx dtype CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1660 opened Feb 4, 2025 by jainapurva • Draft
1
Attempt to switch everything to cmake CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1659 opened Feb 4, 2025 by drisspg Loading…
1
Tensor parallel support for uintx dtype CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. topic: improvement Use this tag if this PR is an improvement (doesn't fit into any of the other categories)
#1656 opened Feb 3, 2025 by jainapurva • Draft
1
draft ukernel selection logic CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1652 opened Feb 3, 2025 by metascroy • Draft
22
[Fix]: Fallback to KleidiAI channelwise kernel groupsize isnt suitable CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1647 opened Jan 31, 2025 by ng-05 Loading…
5
Change TORCH_LIBRARY to TORCH_LIBRARY_FRAGMENT CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1645 opened Jan 30, 2025 by metascroy Loading…
2
Tests jeanschmidt/NVIDIA_IMEX_CHANNELS changes CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1644 opened Jan 30, 2025 by jeanschmidt Loading…
1
mx: add ceil and RNE rounding modes to the cast from fp32 to e8m0 CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. topic: improvement Use this tag if this PR is an improvement (doesn't fit into any of the other categories)
#1643 opened Jan 30, 2025 by vkuzo Loading…
2
Add mx_fp8_bf16 kernel CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. mx topic: new feature Use this tag if this PR adds a new feature
#1637 opened Jan 29, 2025 by drisspg Loading…
7
Create separate float8 tensor subclass CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1636 opened Jan 29, 2025 by jainapurva • Draft
3
Add boilerplate code CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. topic: for developers Use this tag if this PR is mainly developer facing topic: improvement Use this tag if this PR is an improvement (doesn't fit into any of the other categories)
#1635 opened Jan 29, 2025 by jainapurva • Draft
7
Update to cutlass 3.8 | wait for tag to land CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. mx topic: not user facing Use this tag if you don't want this PR to show up in release notes
#1634 opened Jan 28, 2025 by drisspg Loading…
1
Fix broken nav link in README
#1633 opened Jan 28, 2025 by ByronHsu Loading…
2
[not for land] hook up MX to CUDA 12.8 cuBLAS MX gemm CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. topic: not user facing Use this tag if you don't want this PR to show up in release notes
#1625 opened Jan 27, 2025 by vkuzo Loading…
4
convert dora fusion test from pytest to unittest CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. topic: not user facing Use this tag if you don't want this PR to show up in release notes
#1622 opened Jan 26, 2025 by osbm Loading…
5
Move tensor_flatten/unflatten from AQT -> TensorAOBaseClass CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. topic: for developers Use this tag if this PR is mainly developer facing
#1615 opened Jan 24, 2025 by jainapurva • Draft
2
Add H100 to CI for regression CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. topic: for developers Use this tag if this PR is mainly developer facing topic: not user facing Use this tag if you don't want this PR to show up in release notes
#1614 opened Jan 24, 2025 by jainapurva • Draft
1
[WIP] speed up CodebookQuantizedTensor inference CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1607 opened Jan 23, 2025 by DerekLiu35 Loading…
4 tasks
1
Move Hqq quantization to subclass CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#1604 opened Jan 23, 2025 by jainapurva • Draft
1
ProTip! Find all pull requests that aren't related to any open issues with -linked:issue.