Skip to content

Issues: microsoft/Megatron-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

126 Open 56 Closed
126 Open 56 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

Model conversion problem
#449 opened Sep 26, 2024 by yuanzhiyong1999
1
Async allreduce for tensor-parallel
#447 opened Sep 23, 2024 by drcanchi
[TRACKER] Customer support related PR tracker for Intel devices
#446 opened Sep 20, 2024 by delock
7 of 12 tasks
how to calcuate the training throughput
#444 opened Sep 12, 2024 by bigtree2020
llama3 and llama3.1 support
#443 opened Sep 10, 2024 by fmiao2372
1
[Bug] Missing weight gradients from LinearWithGradAccumulationAndAsyncCommunication when Zero Bubble Pipeline Parallelism Is disabled
#442 opened Sep 1, 2024 by mksit
Optimizer problem when using finetune_llama.sh
#440 opened Aug 28, 2024 by Kaiizx
3
zero3 The checkpoint being loaded used a DP world size of 8 but the current world size is 16. Automatic adjustment of ZeRO's optimizer state partitioning with a new world size is not currently supported.
#439 opened Aug 26, 2024 by ArtificialZeng
Why pretrain_llama_distributed.sh use pretrain_gpt.py ?
#437 opened Aug 22, 2024 by BrucePeng92
[bug]: ipex install breaks non xpu devices
#435 opened Aug 8, 2024 by saforem2
2
A tutorial to help you finetune LLama-2-7b using this repository full of garbarge code with ZeRO2/3 enabled.
#430 opened Jul 25, 2024 by LLMChild
1
How to resume training between GPTModel() checkpoint and GPTModelPipe() checkpoint?
#405 opened Jun 27, 2024 by tiggerwu
MOE TFLOPS calculation
#398 opened Jun 5, 2024 by yingzhao27
why moe can not use zero3
#397 opened Jun 4, 2024 by kuangdao
Inquiry on Sequence Parallel Support for VocabParallelEmbedding
#389 opened May 18, 2024 by qinxiangyujiayou
about the optimizer param group
#387 opened May 17, 2024 by L-hongbin
屎山代码DeepSpeed
#386 opened May 11, 2024 by ControllableGeneration
3
Sequence Parallel is incompatible with Rotary Positional Embedding
#385 opened May 9, 2024 by anogkongda
4
Spurious all gather performance drop.
#384 opened Apr 29, 2024 by etiennemlb
Call for Conversion from Huggingface to Megads with MoE
#381 opened Apr 24, 2024 by ControllableGeneration
Expert deepcopy raises PickleError
#380 opened Apr 23, 2024 by sxontheway
AttributeError: 'Namespace' object has no attribute 'deepspeed_config_dict'. Did you mean: 'deepspeed_config'? && batch = next(self.data_iterator)
#379 opened Apr 20, 2024 by hi20240217
2
Assertion failure when there are more than 255 tokenized data files (assert num_datasets < 255 in blendable_dataset.py)
#377 opened Apr 17, 2024 by Jeronymous
Pipeline parallelism + CPU offload?
#369 opened Mar 21, 2024 by webber26232
[BUG] Problems with Mixture-of-Experts (MoE)
#367 opened Mar 16, 2024 by nikit-srivastava
1
ProTip! Find all open issues with in progress development work with linked:pr.