Skip to content

Issues: pytorch/torchtitan

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

71 Open 187 Closed
71 Open 187 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

[Checkpointing] fails out if checkpoint folder does not exist when using keep_latest_k bug Something isn't working module: checkpoint
#911 opened Mar 2, 2025 by lessw2020
[Checkpointing] Set reasonable default for keep_latest_k checkpoints and avoid overflowing user drives...currently set to 'infinite' checkpoints module: checkpoint
#910 opened Mar 1, 2025 by lessw2020
[Checkpointing] Using keep_latest_k setting results in failure when using external mounted drive module: checkpoint
#909 opened Mar 1, 2025 by lessw2020
[Possible PR discuss] Will a PR of training HF model be welcomed? community help wanted huggingface integration
#903 opened Feb 28, 2025 by junjzhang
3
Question about triton in deepseek implementtion question Further information is requested
#902 opened Feb 28, 2025 by zqwenn
@kwen2501
better handling of tensorwise float8 recipe in configuration
#901 opened Feb 27, 2025 by vkuzo
@vkuzo
1
Moving train.py to torchtitan submodule makes run_train.sh failed with "Can not find module"
#897 opened Feb 27, 2025 by jianiw25
1
[Slurm support] New file layout breaks slurm launching...cannot find module 'torchtitan.blah'
#890 opened Feb 26, 2025 by lessw2020
3
[Feature request] Integrate DeepGEMM
#889 opened Feb 26, 2025 by lessw2020
@lessw2020
2
dcp.load fails on checkpoints prior to AdamW refactor module: checkpoint
#886 opened Feb 25, 2025 by eminorhan
@fegin
1
Possible to integrate DeepEP?
#885 opened Feb 25, 2025 by ericxsun
4
[Evaluation] Minimal support for downstream tasks enhancement New feature or request
#883 opened Feb 24, 2025 by K-H-Ismail
3
[Float8] Rowwise with AsyncTP runs at roughly same perf as vanilla TP bug Something isn't working module: float8
#866 opened Feb 20, 2025 by lessw2020
@danielvegamyhre
1
[Float8] Float8 rowwise with vanilla TP encounters NaN around 80 iters in... bug Something isn't working module: float8
#865 opened Feb 20, 2025 by lessw2020
@danielvegamyhre
5
[Float8] Unable to run asyncTP + Float8 row with 'full' AC active, leading dims mismatch bug Something isn't working module: float8
#864 opened Feb 20, 2025 by lessw2020
1
@danielvegamyhre
34
How to define Custom Communication Operations for Custom Operators in Distributed Settings module: dtensor question Further information is requested
#852 opened Feb 17, 2025 by Doraemonzzz
@tianyu-l
3
"Universal" Checkpointing module: checkpoint question Further information is requested
#850 opened Feb 17, 2025 by jeromeku
4
Mitigation to HuggingFace Trainer enhancement New feature or request huggingface integration
#824 opened Feb 6, 2025 by huyiwen
7
HSDP causes loss instability module: fsdp question Further information is requested
#813 opened Jan 31, 2025 by apkumar
@weifengpy @mori360
3
debug model training hangs on NVIDIA B200 with >1 GPU bug Something isn't working module: c10d
#810 opened Jan 28, 2025 by vkuzo
3
unable to run 8b llama bug Something isn't working
#807 opened Jan 27, 2025 by asahni04
1
Gradient Scaling With Pipeline Parallelism module: pipelining question Further information is requested
#803 opened Jan 24, 2025 by windsornguyen
2
[Bug] Unexpected performance drop with float8 training + compiling only nn.Linear layers + using selective per op AC bug Something isn't working
#786 opened Jan 10, 2025 by danielvegamyhre
@tianyu-l
4
Why use RowwiseParallel for nn.Embedding instead of ColwiseParallel? question Further information is requested
#785 opened Jan 10, 2025 by corey-lambda
@tianyu-l
5
BUG: early_step_in_backward with pipeline parallelism and len(model_parts) > 1 bug Something isn't working
#777 opened Jan 7, 2025 by cassanof
@H-Huang @mori360
2
ProTip! Exclude everything labeled bug with -label:bug.