GitHub / Lightning-AI/lightning-thunder issues and pull requests
#2497 - Fix test_parse_resnet18 test for train=True, dtype=float32, executor=nvfuser
Issue -
State: open - Opened by beverlylytle 2 days ago
#2494 - restore type condition on training resnet test
Pull Request -
State: open - Opened by beverlylytle 2 days ago
#2493 - Bump the gha-updates group with 4 updates
Pull Request -
State: open - Opened by dependabot[bot] 2 days ago
Labels: dependencies, github_actions
#2492 - Bump liger-kernel from 0.4.0 to 0.6.2
Pull Request -
State: open - Opened by dependabot[bot] 2 days ago
Labels: dependencies, python
#2489 - Bump pytest-xdist from 3.7.0 to 3.8.0
Pull Request -
State: closed - Opened by dependabot[bot] 2 days ago
Labels: dependencies, python
#2488 - Bump bitsandbytes from 0.46.1 to 0.47.0
Pull Request -
State: closed - Opened by dependabot[bot] 2 days ago
Labels: dependencies, python
#2485 - Replace linear checker for TEv2
Pull Request -
State: open - Opened by riccardofelluga 5 days ago
#2483 - convert transformer-engine to lit CI
Pull Request -
State: closed - Opened by Borda 5 days ago
Labels: ci
#2481 - Implement `thunder.executors.custom_op_ex._override_custom_op_forward`
Pull Request -
State: open - Opened by crcrpar 6 days ago
#2480 - Reflect `cd.is_grad_enabled` to PyTorch after everything is done
Pull Request -
State: open - Opened by shino16 6 days ago
#2479 - Ensure autograd is enabled before connecting Thunder-compiled fn to autograd
Pull Request -
State: open - Opened by shino16 6 days ago
#2478 - Add MOE TP example
Pull Request -
State: open - Opened by kshitij12345 6 days ago
#2476 - [TE] Explore supporting newer recipes in TE like Float8BlockScaling
Issue -
State: open - Opened by kshitij12345 6 days ago
Labels: enhancement, TransformerEngine
#2475 - Skip TE test on SM120+ as Float8BlockScaling is currently unsupported in thunder
Pull Request -
State: open - Opened by kshitij12345 6 days ago
#2474 - Treat `set_grad_enabled` faithfully before connecting to autograd
Pull Request -
State: open - Opened by shino16 7 days ago
#2473 - Restore resnet18 test
Pull Request -
State: open - Opened by beverlylytle 8 days ago
#2472 - Support the `trtllm autodeploy` flashinfer KV-Cached attention in Thunder
Issue -
State: open - Opened by kiya00 8 days ago
Labels: enhancement
#2471 - fix imports with Ruff's I
Pull Request -
State: open - Opened by Borda 8 days ago
#2468 - fix `F403` undefined-local-with-import-star
Pull Request -
State: open - Opened by Borda 8 days ago
#2467 - fix `F601` multi-value-repeated-key-literal
Pull Request -
State: open - Opened by Borda 8 days ago
#2466 - fix `F822` undefined-export
Pull Request -
State: closed - Opened by Borda 8 days ago
Labels: install, code health
#2465 - fix typo of "extracts"
Pull Request -
State: closed - Opened by crcrpar 8 days ago
#2463 - Break ref cycle related to CompileStatistics object
Pull Request -
State: open - Opened by kshitij12345 9 days ago
#2462 - [pre-commit.ci] pre-commit suggestions
Pull Request -
State: closed - Opened by pre-commit-ci[bot] 9 days ago
- 2 comments
#2459 - Break ref cycle in interpreter.py:fn_
Pull Request -
State: open - Opened by kshitij12345 9 days ago
#2458 - Memory Leak when running thunder.jit with nn.Module
Issue -
State: closed - Opened by kshitij12345 9 days ago
- 1 comment
#2455 - fix/update linting configuration
Pull Request -
State: closed - Opened by Borda 11 days ago
#2451 - require PyTorch 2.7
Pull Request -
State: closed - Opened by t-vi 12 days ago
- 1 comment
Labels: dependencies
#2450 - [WIP] Add Llama4 MoE implementation to test_networks
Pull Request -
State: open - Opened by kshitij12345 12 days ago
#2449 - Fix memory leak due to CompileData being in a cycle
Pull Request -
State: closed - Opened by kshitij12345 12 days ago
#2448 - -> '
Pull Request -
State: open - Opened by crcrpar 12 days ago
#2447 - CI: fix slicing for PyTorch nightly
Pull Request -
State: closed - Opened by kshitij12345 13 days ago
#2443 - try to fix te ci
Pull Request -
State: closed - Opened by t-vi 14 days ago
- 1 comment
Labels: ci
#2442 - Remove outdated comments in autodiff
Pull Request -
State: closed - Opened by beverlylytle 14 days ago
#2441 - Remove `op_name_to_fn: dict[str, Callable]` from test_cudnn_executor
Pull Request -
State: closed - Opened by crcrpar 14 days ago
#2440 - [TE] Disable TE CI as it is already failing
Pull Request -
State: closed - Opened by kshitij12345 15 days ago
- 2 comments
Labels: ci
#2439 - [TE] CI failing with version `GLIBCXX_3.4.32' not found
Issue -
State: closed - Opened by kshitij12345 15 days ago
- 2 comments
Labels: TransformerEngine
#2438 - Connect TEv2 states to ThunderModule
Issue -
State: open - Opened by riccardofelluga 15 days ago
Labels: enhancement, TransformerEngine
#2437 - Allow user to specify the process group for TEv2
Issue -
State: open - Opened by riccardofelluga 15 days ago
Labels: enhancement, distributed, TransformerEngine
#2435 - Fix nvfuserex scatter translation to use compile time scatter dim.
Pull Request -
State: open - Opened by jjsjann123 16 days ago
#2434 - [pre-commit.ci] pre-commit suggestions
Pull Request -
State: open - Opened by pre-commit-ci[bot] 16 days ago
#2433 - replace True with true in pyproject.toml
Pull Request -
State: open - Opened by kshitij12345 16 days ago
#2432 - Thunder needs to support dynamic shape
Issue -
State: open - Opened by kiya00 16 days ago
Labels: enhancement
#2431 - Add scatter support in nvfuserex
Pull Request -
State: open - Opened by jjsjann123 19 days ago
#2431 - Add scatter support in nvfuserex
Pull Request -
State: open - Opened by jjsjann123 19 days ago
#2430 - [DTensor] Add prims.take for DTensor
Pull Request -
State: open - Opened by kshitij12345 19 days ago
#2430 - [DTensor] Add prims.take for DTensor
Pull Request -
State: open - Opened by kshitij12345 19 days ago
#2429 - thunderfx: handle output node with no example_value
Pull Request -
State: open - Opened by kshitij12345 20 days ago
- 1 comment
#2429 - thunderfx: handle output node with no example_value
Pull Request -
State: open - Opened by kshitij12345 20 days ago
#2428 - Revert "skip - failing dtensor test"
Pull Request -
State: open - Opened by wujingyue 20 days ago
- 1 comment
#2427 - remove setitem_ output manipulation
Pull Request -
State: closed - Opened by beverlylytle 20 days ago
- 1 comment
#2427 - remove setitem_ output manipulation
Pull Request -
State: open - Opened by beverlylytle 20 days ago
#2426 - bitsandbytes: update _bitsandbytes_available
Pull Request -
State: closed - Opened by kshitij12345 21 days ago
#2425 - skip - failing dtensor test
Pull Request -
State: open - Opened by kshitij12345 21 days ago
#2425 - skip - failing dtensor test
Pull Request -
State: closed - Opened by kshitij12345 21 days ago
#2424 - [reporting tool] Fixes import error in report script
Pull Request -
State: open - Opened by kiya00 21 days ago
#2424 - [reporting tool] Fixes import error in report script
Pull Request -
State: closed - Opened by kiya00 21 days ago
- 1 comment
#2423 - [DTensor] Update creation of nvFuser.DeviceMesh
Pull Request -
State: open - Opened by kshitij12345 21 days ago
#2422 - DTensor: support linear
Pull Request -
State: open - Opened by kshitij12345 21 days ago
#2421 - [docs] Add `thunderfx` to dynamo/index.rst
Pull Request -
State: open - Opened by crcrpar 21 days ago
#2421 - [docs] Add `thunderfx` to dynamo/index.rst
Pull Request -
State: closed - Opened by crcrpar 21 days ago
Labels: documentation
#2420 - thunderfx splitter fails to handle `SymInt` FX nodes
Issue -
State: open - Opened by crcrpar 22 days ago
#2419 - thunderfx: Avoid failure when `example_value` does not have attr of `grad_fn`
Pull Request -
State: open - Opened by crcrpar 22 days ago
#2419 - thunderfx: Avoid failure when `example_value` does not have attr of `grad_fn`
Pull Request -
State: closed - Opened by crcrpar 22 days ago
- 1 comment
#2418 - nvfuserex: return cumsum result in `int64` when input is int/bool and result dtypes is not specified
Pull Request -
State: closed - Opened by crcrpar 22 days ago
#2418 - nvfuserex: return cumsum result in `int64` when input is int/bool and result dtypes is not specified
Pull Request -
State: open - Opened by crcrpar 22 days ago
#2417 - fix typo: "NotImplementedErrror" -> "NotImplementedError"
Pull Request -
State: closed - Opened by crcrpar 22 days ago
#2416 - copying over inference benchmark script
Pull Request -
State: closed - Opened by jjsjann123 23 days ago
- 6 comments
#2415 - ci: reinstall correct torch dependencies
Pull Request -
State: closed - Opened by Borda 23 days ago
- 1 comment
Labels: dependencies, ci
#2415 - ci: reinstall correct torch dependencies
Pull Request -
State: closed - Opened by Borda 23 days ago
- 1 comment
Labels: dependencies, ci
#2414 - re-enable cuda-python nb
Pull Request -
State: open - Opened by kshitij12345 23 days ago
#2414 - re-enable cuda-python nb
Pull Request -
State: closed - Opened by kshitij12345 23 days ago
- 1 comment
Labels: dependencies
#2413 - chore: bump Torch 2.8
Pull Request -
State: closed - Opened by Borda 23 days ago
Labels: ci
#2413 - chore: bump Torch 2.8
Pull Request -
State: closed - Opened by Borda 23 days ago
Labels: ci
#2412 - [DTensor] Add a test with opinfo
Pull Request -
State: open - Opened by kshitij12345 23 days ago
Labels: DTensor
#2411 - Relax tolerance for apex xentropy for float16
Pull Request -
State: closed - Opened by beverlylytle 23 days ago
#2411 - Relax tolerance for apex xentropy for float16
Pull Request -
State: open - Opened by beverlylytle 23 days ago
#2410 - deps: pin `cuda-python >=12.0, <13.0.0`
Pull Request -
State: closed - Opened by Borda 23 days ago
Labels: dependencies, ci
#2410 - deps: pin `cuda-python >=12.0, <13.0.0`
Pull Request -
State: closed - Opened by Borda 23 days ago
Labels: dependencies, ci
#2409 - [do not review] cudnn-frontend `rms_norm`
Pull Request -
State: closed - Opened by crcrpar 25 days ago
Labels: install
#2409 - [do not review] cudnn-frontend `rms_norm`
Pull Request -
State: open - Opened by crcrpar 25 days ago
#2408 - docker: build images for Torch 2.8
Pull Request -
State: open - Opened by Borda 26 days ago
#2408 - docker: build images for Torch 2.8
Pull Request -
State: closed - Opened by Borda 26 days ago
- 4 comments
Labels: docker, ci
#2407 - Move amax and scale later in TEv2 grad transform
Issue -
State: open - Opened by riccardofelluga 26 days ago
Labels: enhancement, TransformerEngine
#2406 - Fused amax and scale update for TEv2
Issue -
State: open - Opened by riccardofelluga 26 days ago
Labels: enhancement, TransformerEngine
#2405 - `TORCH_NCCL_AVOID_RECORD_STREAMS` is the default
Issue -
State: open - Opened by crcrpar 27 days ago
#2404 - TEv2 Add multi-gpu support and tests
Pull Request -
State: open - Opened by riccardofelluga 27 days ago
- 1 comment
#2404 - TEv2 Add multi-gpu support and tests
Pull Request -
State: open - Opened by riccardofelluga 27 days ago
#2403 - Implement `thunder.torch.custom_op._register_custom_op`
Pull Request -
State: open - Opened by crcrpar 27 days ago
Labels: documentation
#2402 - Add cudnn-frontend based backward of layer_norm
Pull Request -
State: closed - Opened by crcrpar 27 days ago
#2402 - Add cudnn-frontend based backward of layer_norm
Pull Request -
State: open - Opened by crcrpar 27 days ago
#2401 - Add TEv2 Transform reset
Pull Request -
State: closed - Opened by riccardofelluga 28 days ago
#2401 - Add TEv2 Transform reset
Pull Request -
State: closed - Opened by riccardofelluga 28 days ago
#2400 - [minor] fix type annotation
Pull Request -
State: closed - Opened by kshitij12345 28 days ago
#2399 - ThunderFX: Modify GraphModule in-place
Pull Request -
State: open - Opened by shino16 28 days ago
#2398 - [thunderfx] Bug when no_grad region is split between inductor and thunder
Issue -
State: open - Opened by kshitij12345 29 days ago
- 3 comments
Labels: thunderfx
#2397 - Return the updated `inp` from `seteitem_`
Pull Request -
State: closed - Opened by crcrpar 29 days ago
- 1 comment
Labels: operators, in-place
#2396 - [WIP]
Pull Request -
State: open - Opened by beverlylytle 30 days ago
#2396 - Don't replace unused variables with None
Pull Request -
State: open - Opened by beverlylytle 30 days ago
- 10 comments
Labels: optimization passes
#2395 - Simplify the argsort support
Pull Request -
State: closed - Opened by wujingyue about 1 month ago
- 4 comments
Labels: operators