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

46 Open 88 Closed
46 Open 88 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

Process got stuck when trying to optimize different groups of parameters using different types of data
#584 opened Sep 18, 2024 by Yangyi-Chen
3
Support INT8 mixed-precision training from torchao?
#578 opened Sep 14, 2024 by gau-nernst
DDP (replicate) + TP? question Further information is requested
#577 opened Sep 13, 2024 by yzs981130
2
Wrong train_state.step when resuming from checkpoint for the second time bug Something isn't working
#571 opened Sep 8, 2024 by LeoXinhaoLee
@wz337
1
Pipeline Parallelism + FSDP question Further information is requested
#562 opened Aug 29, 2024 by jeromeku
1
Fail-safe and partial redundancy for HSDP on unreliable compute enhancement New feature or request
#561 opened Aug 27, 2024 by evkogs
@wconstab
5
Adjust MFU to account for FP8
#560 opened Aug 23, 2024 by lessw2020
1
PP UX/training confusion re: loss = -1. (need to better document or add auto logging of last rank loss?)
#550 opened Aug 21, 2024 by lessw2020
2D whole model compile fails at embedding layer bug Something isn't working
#534 opened Aug 20, 2024 by tianyu-l
@bdhirsh @tianyu-l
2
[rfc] getting rid of seed-checkpoint for Pipeline Parallelism enhancement New feature or request
#514 opened Aug 10, 2024 by wconstab
4
train llama3 error
#502 opened Aug 5, 2024 by starstream
2
[Request] Decouple profiler profile_freq from memory snapshot frequency enhancement New feature or request
#475 opened Jul 23, 2024 by awgu
Only half of parameters are saved when applied PP bug Something isn't working
#474 opened Jul 22, 2024 by dmammfl
7
[FP8 options] Float8Linear vs TransformerEngine question Further information is requested
#462 opened Jul 16, 2024 by yundai424
7
Question about custom cuda operators for tensor parallelism question Further information is requested
#434 opened Jun 28, 2024 by vermouth1992
5
Question about Pipeline parallelism question Further information is requested
#431 opened Jun 27, 2024 by vermouth1992
@wconstab @H-Huang
5
LoRA fine-tuning weights explosion in FSDP training
#421 opened Jun 24, 2024 by MinghaoYan
@weifengpy
12
Llama models with custom configurations and uploading to Hugging Face enhancement New feature or request
#420 opened Jun 24, 2024 by bkchang
1
DataLoader state is empty for different ranks ? question Further information is requested
#409 opened Jun 17, 2024 by ahatamiz
1
Some testing from me
#407 opened Jun 17, 2024 by ad8e
3
How to use nsys? enhancement New feature or request
#399 opened Jun 13, 2024 by vedantroy
1
benchmark perf numbers on H100 GPUs and update performance.md documentation Improvements or additions to documentation
#394 opened Jun 12, 2024 by tianyu-l torchtitan release 1.0
Add torchdata to requirements after release better_engineering Repo code quality improvements
#351 opened May 21, 2024 by gokulavasan
numerical difference for SDPA between non-dtensor vs dtensor, when math attention and fp16 are used bug Something isn't working
#317 opened May 8, 2024 by tianyu-l
1
freqs_cis in llama model should be a non-persistent buffer bug Something isn't working
#316 opened May 8, 2024 by tianyu-l
ProTip! Updated in the last three days: updated:>2024-09-17.