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

482 Open 570 Closed
482 Open 570 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] GPT-j int8 requires more memory than float16 bug Something isn't working inference
#2467 opened Nov 2, 2022 by zelcookie
[BUG] GPT-j memory overflows sharply if you use kernel inject bug Something isn't working inference
#2466 opened Nov 2, 2022 by zelcookie
@cmikeh2
1
[REQUEST] Does DeepSpeed support ONNX export and Triton Inference Server? enhancement New feature or request
#2465 opened Nov 2, 2022 by stevezheng23
[BUG]Call torch.cuda.synchronize each time to reduce gradients in ZeRO Stage2 enhancement New feature or request training user-question Questions about DeepSpeed.
#2463 opened Nov 1, 2022 by li-yi-dong
@samyam
3
[BUG] stage 3 cannot load the checkpoint when optimizer is not configured bug Something isn't working inference
#2449 opened Oct 27, 2022 by rohitgr7
@ShijieZZZZ
7
[Question] MoE: Why exp_counts have to be moved to CPU? question Further information is requested training
#2444 opened Oct 25, 2022 by clumsy
[BUG] MP-sharded checkpoint loading does not work for models except BLOOM bug Something isn't working inference
#2442 opened Oct 24, 2022 by pai4451
@RezaYazdaniAminabadi
1
What's the use of synchronization in activation checkpointing? question Further information is requested training
#2439 opened Oct 21, 2022 by afcruzs
@samyam
1
[QUESTION] Pipeline explaination needed question Further information is requested training
#2435 opened Oct 20, 2022 by bruno-darochac
@ShadenSmith
[REQUEST] Hey, Microsoft...Could you PLEASE Support Your Own OS? enhancement New feature or request
#2427 opened Oct 15, 2022 by d8ahazard
20
Inference Megatron GPT2 ERROR. enhancement New feature or request inference user-question Questions about DeepSpeed.
#2421 opened Oct 13, 2022 by cdj0311
@awan-10
2
DeepSpeed Megatron [BUG] bug Something isn't working training
#2417 opened Oct 12, 2022 by koliaok
@GuanhuaWang
1
Feature request: Ability to disable autocast locally enhancement New feature or request training
#2400 opened Oct 7, 2022 by gahdritz
[Question] How to preshard a model for tensor parallism enhancement New feature or request inference
#2379 opened Sep 29, 2022 by lanking520
@RezaYazdaniAminabadi
6
Schema for ds_config.json enhancement New feature or request training
#2355 opened Sep 24, 2022 by timoklimmer
1
[REQUEST] BF16 mixed precision => grad accum in fp32 enhancement New feature or request training
#2352 opened Sep 23, 2022 by stas00
@tjruwase @jomayeri
1
Minibatch gradients aren't readily accessible in zero optimization level 1 bug Something isn't working training
#2329 opened Sep 17, 2022 by bcui19
@loadams
1
hi ，when the ZeroQuant inference will be released?
#2326 opened Sep 15, 2022 by xk503775229
@yaozhewei
2
[REQUEST] An option to only save the model state_dict when save_checkpoint(), and how to manually save & load the model state_dict when using ZERO3 enhancement New feature or request training
#2304 opened Sep 8, 2022 by BlinkDL
DeepSpeed still gives CUDA-out-of-memory issue bug Something isn't working training
#2302 opened Sep 7, 2022 by buttercutter
27
questuion : how to inference Int8 models (GPT) supported through ZeroQuant technology ? enhancement New feature or request inference
#2301 opened Sep 7, 2022 by xk503775229
@cli99 @yaozhewei
9
[Question] why are overlap and contiguous grads meaningless in stage 1 and are ignored training user-question Questions about DeepSpeed.
#2295 opened Sep 6, 2022 by woolpeeker
[REQUEST] individual tensor .grad getter/setter for flattened tensors enhancement New feature or request training
#2290 opened Sep 2, 2022 by stas00
get_fp32_state_dict_from_zero_checkpoint with fixed weights in pipeline enhancement New feature or request training
#2288 opened Sep 2, 2022 by CodyLDA
2
[BUG] Wrong output for batch input for opt model inference. bug Something isn't working inference
#2281 opened Sep 1, 2022 by yingapple
@molly-smith
6
ProTip! Follow long discussions with comments:>50.