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

51 Open 147 Closed
51 Open 147 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

Checkpoint conversion
#758 opened Dec 20, 2024 by MaxiBoether
3
[question]can't disable CP for specific (unsupported) SDPA op context_parallel enhancement New feature or request
#757 opened Dec 20, 2024 by FindDefinition
@fegin @XilunWu
1
Any plans to support DPO training?
#756 opened Dec 20, 2024 by xs1997zju
1
JobConfig does not support typing enhancement New feature or request
#753 opened Dec 18, 2024 by greeneggsandyaml
1
Model init with HuggingFace model
#743 opened Dec 16, 2024 by neeldani
@mori360
4
Low bit Optimizers & FA-3
#742 opened Dec 16, 2024 by asahni04
4
using fsdp2 wrapper Flux(text to image) model , gradient is inconsistent with fsdp1 question Further information is requested
#734 opened Dec 13, 2024 by yanmj0601
9
Issue: Loss Discrepancy Between FSDP1 and FSDP2 with AdamW Optimizer question Further information is requested
#724 opened Dec 9, 2024 by Teng-xu
1
Context parallelism understanding context_parallel question Further information is requested
#723 opened Dec 9, 2024 by jinsong-mao
4
First Shard Group Save and Load Checkpoint for HSDP question Further information is requested
#709 opened Nov 29, 2024 by qsh-zh
4
Vote on new features in Discussions
#694 opened Nov 23, 2024 by tianyu-l
16
[rfc] torchtitan release practices release_blocking Issues that are blocking the milestone / release completion
#688 opened Nov 22, 2024 by tianyu-l torchtitan v1.0.0 release
torch.compile(sync_float8_amax_and_scale_history) not working with triton latest main bug Something isn't working
#681 opened Nov 19, 2024 by goldhuang
2
[Parallelism] Implement vocabulary parallelism enhancement New feature or request
#680 opened Nov 15, 2024 by casper-hansen
1
Any suggestion for Llama-3.1-70b(128k seq len) deploy mesh with torchtian? enhancement New feature or request question Further information is requested
#678 opened Nov 15, 2024 by medivh-xp
8
Very low wps with H200 Gpus question Further information is requested
#676 opened Nov 13, 2024 by aniltrkkn
7
Questions about FSDP2 support and memory usage. question Further information is requested
#658 opened Oct 29, 2024 by tangjiasheng
6
meta device issue with float8 delayed scale bug Something isn't working
#654 opened Oct 25, 2024 by weifengpy torchtitan v1.0.0 release
8
torch.distributed.breakpoint(rank=1) hangs because of --local-ranks-filter 0 documentation Improvements or additions to documentation
#652 opened Oct 25, 2024 by weifengpy
[Multimodal] Adding OBELICS DataLoader enhancement New feature or request
#650 opened Oct 24, 2024 by TJ-Solergibert
1
8
[Config] Make FSDP reshard_after_forward: bool configurable enhancement New feature or request
#644 opened Oct 22, 2024 by awgu
What is the expected inference steps after I apply torchao in training? question Further information is requested
#638 opened Oct 21, 2024 by goldhuang
3
add H100 in CI better_engineering Repo code quality improvements integration test Adding integration tests
#632 opened Oct 18, 2024 by tianyu-l
1
create a note on torchtitan official release documentation Improvements or additions to documentation release_blocking Issues that are blocking the milestone / release completion
#631 opened Oct 18, 2024 by tianyu-l torchtitan v1.0.0 release
@XilunWu @H-Huang
@lessw2020 @tianyu-l
Non-DP runs default to float32 precision enhancement New feature or request
#630 opened Oct 18, 2024 by carmocca
1
1
ProTip! no:milestone will show everything without a milestone.