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

959 Open 1,629 Closed
959 Open 1,629 Closed
Author
Filter by author
Label
Filter by label
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Milestones
Filter by milestone
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

[BUG] 4-bit quantized models would repeatedly generate the same tokens when bf16.enabled is true bug Something isn't working compression
#5636 opened Jun 10, 2024 by Atry
1
Deepspeed stage 3 hanging after 1st validation sample bug Something isn't working training
#5635 opened Jun 10, 2024 by AceMcAwesome77
[BUG] RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! bug Something isn't working training
#5634 opened Jun 10, 2024 by fahadh4ilyas
1
[BUG] is_zero_init_model is always False when I'm using zero_init! bug Something isn't working training
#5631 opened Jun 8, 2024 by CHNRyan
3
[BUG] RuntimeError encountered when generating tokens from a DeepSpeedHybridEngine initialized with 4-bit quantization. bug Something isn't working deepspeed-chat Related to DeepSpeed-Chat
#5630 opened Jun 8, 2024 by Atry
2
[BUG] 1: error: must run as root and 2: raise RuntimeError("Ninja is required to load C++ extensions") bug Something isn't working training
#5627 opened Jun 7, 2024 by YangBrooksHan
[BUG] RuntimeError: Error building extension 'fused_adam' Loading extension module fused_adam bug Something isn't working compression
#5623 opened Jun 6, 2024 by JinQiangWang2021
[REQUEST] Moving a trainable model with an optimiser between GPU and CPU enhancement New feature or request
#5620 opened Jun 5, 2024 by kfertakis
[BUG] Pipeline Dataloader Samler: shuffle=False bug Something isn't working training
#5619 opened Jun 5, 2024 by Coobiw
[BUG] ZeRO optimizer with MoE Expert Parallelism bug Something isn't working training
#5618 opened Jun 5, 2024 by Jack47
@jomayeri
1
[HELP] ZeRO3 partition parameters after fully load to each GPU! bug Something isn't working training
#5617 opened Jun 5, 2024 by CHNRyan
@jomayeri
3
nv-ds-chat CI test failure ci-failure
#5616 opened Jun 5, 2024 by github-actions bot
# [REQUEST] Upstream modifications of PaRO enhancement New feature or request
#5607 opened Jun 3, 2024 by youshaox
[BUG] cannot import name '_get_socket_with_port' from 'torch.distributed.elastic.agent.server.api' bug Something isn't working training
#5603 opened Jun 3, 2024 by fahadh4ilyas
3
[BUG] Zero3 causes AttributeError: 'NoneType' object has no attribute 'numel' in continual training bug Something isn't working training
#5602 opened Jun 3, 2024 by thkimYonsei
2
[BUG] CUDA OOM error when Hugging Face ignore_mismatched_sizes is enabled bug Something isn't working training
#5599 opened May 31, 2024 by matthewclso
[BUG] M1 Mac has an issue with hostname -I not being a valid command bug Something isn't working build Improvements to the build and testing systems.
#5597 opened May 31, 2024 by AbhinavMir
@loadams
2
[REQUEST] Supporting custom generation loop (outlines, LMQL, guidance) in DeepSpeedHybridEngine enhancement New feature or request
#5595 opened May 31, 2024 by Atry
[BUG] RepeatingLoader may be invalid in the pipe stages neither the fist nor last bug Something isn't working training
#5593 opened May 31, 2024 by janelu9
[BUG] DeepSpeed is loads the whole model to every GPUs instead of partitioning bug Something isn't working inference
#5592 opened May 31, 2024 by yunoJ
@jomayeri
1
different setting for same (num_gpus * batch_size * grad_accum_steps) output different loss and gradient norm bug Something isn't working training
#5583 opened May 29, 2024 by SeunghyunSEO
1
Data Loading for DeepSpeed Ulysses and Data Parallelism bug Something isn't working training
#5582 opened May 29, 2024 by zijian-hu
[BUG] deepspeed amp seems to convert all input to specific dtype bug Something isn't working training
#5580 opened May 29, 2024 by rangehow
[BUG] fp6 can‘t load qwen1.5-34b-chat bug Something isn't working inference
#5579 opened May 29, 2024 by pointerhacker
[BUG]I found that the parameters of model will be fully transferred to the VRAM of each process. Is this abnormal in my understanding? bug Something isn't working deepspeed-chat Related to DeepSpeed-Chat
#5575 opened May 28, 2024 by tiandazhao
5
ProTip! Adding no:label will show everything without a label.