Skip to content

Issues: hpcaitech/ColossalAI

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

395 Open 1,239 Closed
395 Open 1,239 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

[FEATURE]: How to skip a custom node from generating strategies in colossal-auto? enhancement New feature or request
#5983 opened Aug 8, 2024 by robotsp
[fp8] support amp
#5974 opened Aug 7, 2024 by ver217
@ver217
[fp8] support hybrid parallel plugin
#5972 opened Aug 7, 2024 by wangbluo
@wangbluo
qwen2 fp8 forward/backward
#5971 opened Aug 7, 2024 by wangbluo
[BUG]: Hang on startup bug Something isn't working
#5969 opened Aug 6, 2024 by rob-hen
1 task done
2
support moe
#5954 opened Jul 30, 2024 by flybird11111
@flybird11111
[BUG]: Pytest with a specific config failed after PR #5868 bug Something isn't working shardformer
#8639 opened Jul 29, 2024 by GuangyaoZhang
1 task done
@Edenzzzz
[FEATURE]: Request updates for pretraining roberta enhancement New feature or request
#8638 opened Jul 29, 2024 by jiahuanluo
[Feature]: support FP8 communication in Gemini
#8633 opened Jul 26, 2024 by BurkeHulk
@BurkeHulk
[BUG]: Pipeline Parallelism fails when input shape varies bug Something isn't working shardformer
#8630 opened Jul 25, 2024 by GuangyaoZhang
1 task done
1
[BUG]: pip install . error: identifier "__hsub" is undefined bug Something isn't working
#5929 opened Jul 19, 2024 by jtmer
1 task done
4
[Feature]: support FP8 communication in FSDP
#5927 opened Jul 19, 2024 by BurkeHulk
1
@BurkeHulk
[Feature]: support FP8 communication in DDP
#5926 opened Jul 19, 2024 by BurkeHulk
1
@BurkeHulk
[BUG]: Shardformer FP8 communication training accuracy degradation bug Something isn't working
#5920 opened Jul 18, 2024 by GuangyaoZhang
1 task done
[Feature] support shardformer
#5916 opened Jul 17, 2024 by GuangyaoZhang
@GuangyaoZhang
[BUG]: Low_Level_Zero plugin crashes with LoRA bug Something isn't working
#5909 opened Jul 15, 2024 by Fallqs
1 task done
14
[PROPOSAL]: Does the LowLevelZero Plugin Support Lora, This Code Is Confusing enhancement New feature or request
#5908 opened Jul 15, 2024 by YeAnbang
1 task
1
[BUG]: run opt inference but failed with No module named 'energonai' bug Something isn't working
#5906 opened Jul 13, 2024 by munger1985
1 task done
Whether to support the training acceleration of the StableDiffusion3 algorithm model？ enhancement New feature or request
#5900 opened Jul 10, 2024 by tensorflowt
1
training issue
#5890 opened Jul 5, 2024 by MaleekaA
1
[BUG]: ColossalChat train sft is skipped with opt-1.3b model bug Something isn't working
#5865 opened Jun 27, 2024 by smash1999
1 task done
7
[BUG]: Colossal AI failed to load ChatGLM2 bug Something isn't working
#5861 opened Jun 26, 2024 by hiprince
1 task done
@isky-cd
2
Use gemini plugin and LowLevelZero to run llama2_7b. In the pulgin in gemini, set the policy to static, shard_param_frac, offload_optim_frac, and offload_param_frac to 0.0, making gemini equal to zero2, and set stage to 2 in LowLevelZero. Using bf16 for training, and comparing the two plugins, we found that the GPU memory usage of gemini is higher than that of LowLevelZero. Why is this? In principle, gemini should save more GPU memory
#5830 opened Jun 18, 2024 by JJGSBGQ
2
[FEATURE]: LoRA with sharded model enhancement New feature or request
#5826 opened Jun 17, 2024 by KaiLv69
[BUG]: Shardformer failure with torch 2.3 bug Something isn't working
#5757 opened May 27, 2024 by Edenzzzz
1 task done
@ver217
ProTip! Follow long discussions with comments:>50.