Skip to content

Issues: Lightning-AI/lightning-thunder

Beta
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

326 Open 403 Closed
326 Open 403 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

Can JIT support a mode that skips cache lookup and prologue checks? constraints & caching enhancement New feature or request profile guided optimization
#2042 opened May 6, 2025 by kiya00
1
Grads for Qwen/Qwen2.5-7B-Instruct on CUDA are not close to reference results with large relative difference cuda numerical accuracy testing
#2040 opened May 6, 2025 by IvanYashchuk
2
"default transform" for executors enhancement New feature or request
#2039 opened May 6, 2025 by riccardofelluga
@riccardofelluga
NotImplementedError with no further explanation when trying to run Hugging Face implementation of Phi 3 model huggingface For supporting HF models interpreter
#2038 opened May 6, 2025 by IvanYashchuk
[ThunderFX] RuntimeError: Advanced indexing currently only supports zero or one-dimensional integer tensors, but found a tensor with dtype thunder.dtypes.bool8 and 3 dimensions
#2036 opened May 5, 2025 by kiya00
1
1
recipe with hybrid FSDP and DDP enhancement New feature or request lightning-l1 recipe thunder.compile
#2033 opened May 4, 2025 by t-vi
1
Add activation checkpointing support to functional TE executor enhancement New feature or request TransformerEngine
#2027 opened May 2, 2025 by riccardofelluga
@riccardofelluga
[profile-guided optimization] The input tensors for the subgraph are newly allocated each time, which leads to OOM, whereas ThunderFX works. profile guided optimization thunderfx for things that could be applicable to the dynamo+thunder frontend
#2026 opened May 2, 2025 by kiya00
1
[profile-guided optimization] The input tensor values are lost for the FX subgraphs profile guided optimization thunderfx for things that could be applicable to the dynamo+thunder frontend
#2025 opened May 2, 2025 by kiya00
1
Automated Hugging Face transformers coverage testing hf-transformers lightning-l1 program-coverage Requests for model and program coverage
#2023 opened May 1, 2025 by t-vi
2
quickstart performance monitoring and analysis huggingface For supporting HF models lightning-l1 performance
#2022 opened May 1, 2025 by t-vi
Fix inconsistency in prologue for test_jit_general.py::test_litgpt_variants_kvcache and enable check_trace in test interpreter lightning-l1
#2009 opened Apr 30, 2025 by t-vi
HF "deepseek-ai/deepseek-v3" TransformerLayer fails to run with thunder.jit and thunderfx huggingface For supporting HF models
#2005 opened Apr 29, 2025 by kshitij12345
@kshitij12345
fx_report should have better error message when the input callable is thunder compiled function reporting
#2002 opened Apr 28, 2025 by kiya00
@kiya00
frexp backward returns garbage values for float64 inputs of larger shapes autograd operators testing
#1991 opened Apr 25, 2025 by khushi-411
@IvanYashchuk
Test dtype consistency against torch enhancement New feature or request operators testing
#1983 opened Apr 24, 2025 by riccardofelluga
@IvanYashchuk
2
Hybrid FSDP and DDP does not work (composable transforms) distributed transforms
#1980 opened Apr 22, 2025 by t-vi
1
@t-vi
ThunderFX reporting: The e2e performance result shows Torch is faster, but the saved FXgraph script says Thunder is faster reporting
#1968 opened Apr 15, 2025 by kiya00
@kiya00
KeyError when running HF microsoft/GRIN-MoE model autograd huggingface For supporting HF models tracing time
#1961 opened Apr 14, 2025 by riccardofelluga
@riccardofelluga
9
thunder.jit does not allow changing the model structure (None->Buffer) in forward and leads to incorrect output if doing so
#1951 opened Apr 9, 2025 by kshitij12345
1
[thunderfx] Problem with buffer updated during the forward - RuntimeError: Trying to backward through the graph a second time (or directly access saved tensors after they have already been freed). huggingface For supporting HF models
#1950 opened Apr 9, 2025 by kshitij12345
3
Support user-side CUDA graph capturing cudagraphs
#1945 opened Apr 5, 2025 by crcrpar
1
implement torch.nn.interpolate mode=bilinear etc. good first issue Good for newcomers operators program-coverage Requests for model and program coverage
#1941 opened Apr 4, 2025 by t-vi
Pytest command line option to enable testing complex dtypes
#1940 opened Apr 3, 2025 by beverlylytle
[reporting] check_consistency to use double precision computation as reference and verify that thunder result is closer. enhancement New feature or request reporting
#1935 opened Apr 3, 2025 by kshitij12345
1
ProTip! Type g i on any issue or pull request to go back to the issue listing page.