Skip to content

Issues: pytorch/ao

[RFC] torchao Contributor Guide
#391 opened Jun 18, 2024 by jerryzh168
Open 16
low precision training upcoming feature tracker
#556 opened Jul 30, 2024 by vkuzo
Open 2
Multibackend tracker
#1082 opened Oct 15, 2024 by msaroufim
Open
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

218 Open 215 Closed
218 Open 215 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
Assignee
Filter by who’s assigned
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Issues list

Failed Dispatch of Custom CUDA OP in TorchAO
#1981 opened Mar 30, 2025 by leslie-fang-intel
Failed to save the static quantized model
#1950 opened Mar 25, 2025 by yiliu30
cast to mxfp8 across dim1 should be performant float8
#1945 opened Mar 24, 2025 by vkuzo
Torchao import time
#1944 opened Mar 24, 2025 by felipemello1
1
[Bug] FSDP2 FP8 compatibility problem with nn.Linear layers (GPU count > out_features) float8
#1938 opened Mar 24, 2025 by HIT-cwh
4
FSDP2 + CPU Offload + AdamW8bit issue
#1931 opened Mar 21, 2025 by psinger
1
Torchao's CPU overhead counteracts the performance benefit of using quantization kernel.
#1930 opened Mar 21, 2025 by LuFinch
11
fp8 quantization with FSDP2 error
#1929 opened Mar 20, 2025 by happynear
2
Does torchao support FP8 Grouped GEMM? float8
#1928 opened Mar 20, 2025 by zigzagcai
5
Accelerate activation sparsity with activation compression good first issue Good for newcomers
#1920 opened Mar 18, 2025 by jcaip
placeholder for tracking saturated cast logic to float8_e4m3fn
#1912 opened Mar 17, 2025 by vkuzo
@vkuzo
Move Experimental to TorchAO prototype
#1899 opened Mar 14, 2025 by drisspg
11
Can FP8 GEMM be enabled via module hooks instead of module swapping?
#1887 opened Mar 14, 2025 by zigzagcai
7
Torchao opt resuming from ckpt requires weights_only=False optimizer
#1885 opened Mar 13, 2025 by felipemello1
2
TorchAo autoquant + hugging face breaks on Load autoquant
#1819 opened Mar 3, 2025 by drisspg
@jerryzh168
1
Status of prototype features rfc topic: deprecation Use this tag if this PR deprecates a feature tracker
#1807 opened Mar 1, 2025 by msaroufim
6 of 11 tasks
5
What kind of layers are optimized by torchao on a RTX 4090? performance question Further information is requested
#1805 opened Mar 1, 2025 by naiveen
1
New Pytorch Triton breaks custom cast kernel MX mx
#1799 opened Feb 28, 2025 by drisspg
2
An error was encountered setting torch._dynamo.decorators.mark_unbacked quantize question Further information is requested triaged
#1790 opened Feb 27, 2025 by songh11
3
mx cast to mxfp8 across dim0 and dim1 should be performant mx triaged
#1788 opened Feb 26, 2025 by vkuzo
@eellison
2
error in using torchao and torch compile on rtx 4090 autoquant bug Something isn't working performance quantize
#1775 opened Feb 25, 2025 by zhangvia
@jerryzh168
13
torch.compile cast to mxfp8 with blocked scales should be performant
#1773 opened Feb 24, 2025 by vkuzo
MX single node performance tracker
#1768 opened Feb 24, 2025 by vkuzo
@vkuzo
The cpython structures(e.g. Py_buffer) are excluded, compiling failed (e.g.ao/torchao/csrc/cuda/tensor_core_tiled_layout/tensor_core_tiled_layout.cu) build
#1767 opened Feb 24, 2025 by codereba
3
[QST] About NaNs generated during FP16->FP8 quantization high priority triage review
#1766 opened Feb 23, 2025 by alexsamardzic
1
8
ProTip! Follow long discussions with comments:>50.