Skip to content

Issues: microsoft/DeepSpeed

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

998 Open 1,660 Closed
998 Open 1,660 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

run evaluate_cogagent_demo.py error from sat import mpu, get_args, get_tokenizer error bug Something isn't working inference
#5803 opened Jul 27, 2024 by gyjlll
1
How to silence warning when import deepseed without change the source code? enhancement New feature or request
#5801 opened Jul 25, 2024 by sean-wade
@loadams
[BUG]RuntimeError: disagreement between rank0 and rank1: rank0: bug Something isn't working compression
#5799 opened Jul 24, 2024 by yiyepiaoling0715
1
[BUG] Expert parallel hangs at the last MoE layer bug Something isn't working training
#5794 opened Jul 23, 2024 by JessePrince
2
[BUG] Excessive CPU and GPU Memory Usage with Multi-GPU Inference Using DeepSpeed bug Something isn't working inference
#5793 opened Jul 23, 2024 by gawain000000
2
[BUG] pipline engine's training stucked when zero=1 bug Something isn't working compression
#5792 opened Jul 23, 2024 by janelu9
2
[REQUEST]How to set Ulysses in deepspeed config json? enhancement New feature or request
#5787 opened Jul 22, 2024 by xs1997zju
does deepspeed support pure bf16 training? enhancement New feature or request
#5784 opened Jul 21, 2024 by hjc3613
how to set "training_step" during training? bug Something isn't working training
#5779 opened Jul 17, 2024 by qwerfdsadad
[BUG] Universal checkpoint conversion - "Cannot find layer_01* files in there" bug Something isn't working training
#5776 opened Jul 17, 2024 by exnx
@xylian86
1
[BUG] Multi-node fine-tuning with thunderbolt bug Something isn't working training
#5766 opened Jul 11, 2024 by Raywang0211
2
[BUG] Multi-gpu stuck when the computation graph is not complete for wach process. bug Something isn't working training
#5762 opened Jul 10, 2024 by gary-young
[BUG] I can't run fp8 with pipeline parallel bug Something isn't working training
#5760 opened Jul 10, 2024 by exnx
2
In distributed training, in order to continue training, an error occurred when loading model checkpoints after saving them. training
#5754 opened Jul 7, 2024 by WhaleSpring
[BUG] Learning rate scheduler and optimizer logical issue bug Something isn't working training
#5731 opened Jul 5, 2024 by zhourunlong
lr scheduler defined in config cannot be overwritten by lr scheduler defined in code and pass to deepspeed.initialize [BUG] bug Something isn't working training
#5726 opened Jul 5, 2024 by xiyang-aads-lilly
[BUG] ImportError: /home/nlp/.cache/torch_extensions/py310_cu121/cpu_adam/cpu_adam.so: cannot open shared object file: No such file or directory bug Something isn't working training
#5723 opened Jul 4, 2024 by PhysicianHOYA
@loadams
1
[REQUEST] Asynchronous Checkpointing enhancement New feature or request
#5721 opened Jul 2, 2024 by zaptrem
1
Issue with LoRA Tuning on llama3-70b using PEFT and TRL's SFTTrainer training
#5719 opened Jul 2, 2024 by yutanozaki1
Different seeds are giving the exact same loss on Zero 1,2 and 3 during multi gpu training [BUG] bug Something isn't working training
#5717 opened Jul 2, 2024 by selenerkan
[REQUEST] Does Universal Checkpoint supports for MoE Checkpoint？ enhancement New feature or request
#5716 opened Jul 2, 2024 by tiggerwu
@tohtana
3
[BUG] localhost: Permission denied, please try again. with single node and multi-gpus with --autotuning run bug Something isn't working training
#5709 opened Jul 1, 2024 by Looong01
[BUG] 1-bit LAMB not compatible with bf16 bug Something isn't working training
#5708 opened Jun 28, 2024 by catid
on Activation Checkpointing bug Something isn't working training
#5704 opened Jun 28, 2024 by ChaunceyWang
6
[BUG] Mixed-precision: fp16 will cast input_ids into torch.cuda.HalfTensor instead of Long or Int.
#5701 opened Jun 28, 2024 by zhaoyang02
1
ProTip! Add no:assignee to see everything that’s not assigned.