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

978 Open 1,876 Closed
978 Open 1,876 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

[BUG] using deepspeed slower inference time bug Something isn't working inference
#6818 opened Dec 4, 2024 by williamlin0518
deepspeed installation problem
#6817 opened Dec 4, 2024 by sagie-dekel
1
[QST] MoE auxiliary loss
#6816 opened Dec 4, 2024 by osayamenja
[BUG] DeepSpeed accuracy issue for torch.compile if activation checkpoint function not compiler disabled bug Something isn't working training
#6811 opened Dec 1, 2024 by NirSonnenschein
[BUG] Enabling drop_tokens in MoE layer causes inference to hang bug Something isn't working inference
#6809 opened Nov 29, 2024 by Shamauk
[Questions] Why Ulysess need all2all for QKV, but RingAttention just need KV under context parallel ?
#6808 opened Nov 29, 2024 by elevenxiang
multiple runs on same machine, with ctrl+c, all runs are killed
#6807 opened Nov 29, 2024 by ysyyork
1
[BUG] Getting "SymIntArrayRef expected to contain only concrete integers" error when > 1 GPU bug Something isn't working training
#6806 opened Nov 28, 2024 by rileyhun
[BUG] deepspeed inference for llama3.1 70b for 2 node, each node with 2 gpu bug Something isn't working inference
#6805 opened Nov 28, 2024 by rastinrastinii
[BUG] Unnecessary memory copy in paramater partition in ZeRO3 bug Something isn't working training
#6804 opened Nov 28, 2024 by yingtongxiong
3
Use DS4Sci_EvoformerAttention and torch.util.checkpoint.checkpoint at the same time during training
#6802 opened Nov 28, 2024 by cbyzju
nv-torch-nightly-v100 CI test failure ci-failure
#6801 opened Nov 28, 2024 by github-actions bot
nv-ds-chat CI test failure ci-failure
#6800 opened Nov 28, 2024 by github-actions bot
deepspeed setup for requiring grads on the input (explainability) without huge increase in memory over all gpus
#6798 opened Nov 27, 2024 by GonyRosenman
Question about using Autotuner with ZeRO and tensor parallelism
#6796 opened Nov 27, 2024 by rlanday
Does ZeRO++ Work on AMD GPU Mi200?
#6795 opened Nov 27, 2024 by unavailableun
@hwchen2017
2
AssertionError: no sync context manager is incompatible with gradientpartitioning logic of ZeRo stage 3
#6793 opened Nov 26, 2024 by 66RomanReigns
13
nv-nightly CI test failure ci-failure
#6790 opened Nov 26, 2024 by github-actions bot
[REQUEST] Let ZeRO-offload use CPU and GPU parallelly enhancement New feature or request
#6778 opened Nov 23, 2024 by fzyzcjy
[BUG] [Fix Suggestion] Uneven head sequence parallelism bug Something isn't working training
#6774 opened Nov 21, 2024 by Eugene29
3
[BUG] [Fix-Suggested] ZeRO Stage 3 Overwrites Module ID Attribute Causing Incorrect Expert Placement on GPUs bug Something isn't working training
#6772 opened Nov 20, 2024 by traincheck-team
@tjruwase
1
[BUG] [Fix-Suggested] Checkpoint Inconsistency When Freezing Model Parameters Before deepspeed.initialize
#6771 opened Nov 20, 2024 by traincheck-team
[BUG] [Fix-Suggested] KeyError in stage_1_and_2.py Due to Optimizer-Model Parameter Mismatch
#6770 opened Nov 20, 2024 by traincheck-team
1
grad is None bug Something isn't working training
#6768 opened Nov 20, 2024 by suanflower
4
[BUG] clip_grad_norm for zero_optimization mode is not working bug Something isn't working training
#6767 opened Nov 20, 2024 by chengmengli06
3
ProTip! Adding no:label will show everything without a label.