-
Notifications
You must be signed in to change notification settings - Fork 360
Issues: NVIDIA/TransformerEngine
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.
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
Author
Label
Projects
Milestones
Assignee
Sort
Issues list
Support simulating FP8 on older hardware
enhancement
New feature or request
#71
opened Feb 10, 2023 by
zplizzi
PyTorch 2.2.0 NVFuser deprecation is incompatible with TransformerEngine.
bug
Something isn't working
#666
opened Feb 13, 2024 by
sirutBuasai
What is context parallelism
documentation
Improvements or additions to documentation
#546
opened Nov 30, 2023 by
li-yi-dong
FP8 Unable to achieve the expected FLOPS indicator in 4090
performance
#606
opened Jan 16, 2024 by
kaijun924
AttributeError: module 'transformer_engine' has no attribute 'pytorch'
#1014
opened Jul 15, 2024 by
Lzhang-hub
[Feature Request] Compressed Tiled Matrix Multiply for Autoregressive LLM Inference
#223
opened May 17, 2023 by
veritas9872
Selective Activation Checkpointing with LayerNormMLP
enhancement
New feature or request
performance
#623
opened Jan 22, 2024 by
denizokt
[Feature Request] Please support pytorch and Jax FP8 type without custom data type
#391
opened Aug 20, 2023 by
MoFHeka
[Feature Request] Integration of DiT components into TransformerEngine.
#900
opened Jun 10, 2024 by
okotaku
FusedAdam optimizer doesn't have
set_to_none
keyword argument
#1453
opened Feb 4, 2025 by
MaciejBalaNV
How about the torch.compile in TransformerEngine ?
question
Further information is requested
#1241
opened Oct 11, 2024 by
south-ocean
qwen1.5-0.5B failed to save model with huggingface transformers
#1482
opened Feb 13, 2025 by
xinpengzz
Expected margin of error versus a typical pytorch implementation
#407
opened Aug 29, 2023 by
152334H
Previous Next
ProTip!
Adding no:label will show everything without a label.