Skip to content

Issues: Lightning-AI/lightning-thunder

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

305 Open 368 Closed
305 Open 368 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

fusion_type="dataflow" can lead to invalid trace fusion logic
#1858 opened Mar 7, 2025 by kshitij12345
Expected NumberType mismatch in pad meta
#1857 opened Mar 7, 2025 by kshitij12345
[reporting] Utility to save failing repros for a FXReport enhancement New feature or request reporting
#1856 opened Mar 7, 2025 by kshitij12345
1
[reporting] Saved repro may have incorrect indentation. reporting
#1855 opened Mar 7, 2025 by kshitij12345
[reporting] Calling run_repro leads to jit error - Encountered exception AssertionError: attribute nodes of Graph object out of sync reporting
#1854 opened Mar 7, 2025 by kshitij12345
[Benchmark nvFusion Symbol]: wall time number is different between online fd execution and the saved script execution benchmarking nvfuser
#1852 opened Mar 7, 2025 by kiya00
1
test_dynamo.py::test_thundercompiler_optim_step broken dynamo thunderfx for things that could be applicable to the dynamo+thunder frontend
#1821 opened Mar 1, 2025 by t-vi
Feed fp16/bf16 friendly values to test cases enhancement New feature or request testing
#1803 opened Feb 25, 2025 by riccardofelluga
@riccardofelluga
cuDNN SDPA executor does not support batched broadcast across attention heads mask cudnn
#1799 opened Feb 25, 2025 by IvanYashchuk
Sharing memory pools in CUDAGraphs has non-trivial constraints for memory reuse cudagraphs
#1792 opened Feb 24, 2025 by ali-alshaar7
@ali-alshaar7
2
A mask conversion for cuDNN execution of scaled dot product attention should be fused cudnn enhancement New feature or request sdpa
#1789 opened Feb 24, 2025 by IvanYashchuk
[Reporting] Automated bisection to identify significant changes in performance reporting
#1781 opened Feb 18, 2025 by mruberry
@kiya00
_interpret_call could reuse names in a trace when it's called in the lookaside of torch.autograd.Function autograd
#1776 opened Feb 18, 2025 by crcrpar
1
@crcrpar
1
Infinite loop in inplace functionalization bug Something isn't working in-place
#1770 opened Feb 17, 2025 by beverlylytle
3
Taking requires_grad propagation seriously in Thunder autograd design required enhancement New feature or request
#1768 opened Feb 17, 2025 by IvanYashchuk
High memory consumption without dataflow-based fusion fusion logic thunderfx for things that could be applicable to the dynamo+thunder frontend
#1762 opened Feb 11, 2025 by riccardofelluga
@riccardofelluga
1
Apply set_execution_file to specific traces such as forward/backward execution traces developer efficiency enhancement New feature or request
#1760 opened Feb 11, 2025 by crcrpar
1
Support packing multiple sequences with Flash Attention without cross-contamination enhancement New feature or request high priority nemo Issues needed to support NVIDIA NeMo models. thunderfx for things that could be applicable to the dynamo+thunder frontend
#1758 opened Feb 10, 2025 by IvanYashchuk
@kevinstephano
6
more general inplace support (index_copy_ in litgpt fails to trace) enhancement New feature or request in-place
#1743 opened Feb 4, 2025 by ali-alshaar7
@ali-alshaar7
1
Thunder slower than eager for PEFT LoRA configs with small input sizes peft
#1738 opened Feb 4, 2025 by riccardofelluga
@riccardofelluga
6
thunderfx produces results with incorrect requires_grad autograd
#1733 opened Feb 1, 2025 by jjsjann123
10
Grad Transform generates inconsistent saved_for_backward between forward and backward trace. autograd
#1732 opened Feb 1, 2025 by jjsjann123
4
dividing a float16 tensor by a python float is inaccurate with nvfuser numerical accuracy nvfuser
#1724 opened Jan 30, 2025 by beverlylytle
1
Run LitGPT benchmarking with a custom Attention implementation priority. benchmarking
#1714 opened Jan 29, 2025 by wprazuch
1
@riccardofelluga
6
Input upcast is missing in Thunder's implementation of torch.nn.functional.rms_norm operators
#1713 opened Jan 29, 2025 by IvanYashchuk
@riccardofelluga
8
ProTip! Mix and match filters to narrow down what you’re looking for.