Lightning-AI/lightning-thunder 0.2.6

Thunder 0.2.6 PyTorch Conference Edition

Repository: Lightning-AI/lightning-thunder

Tag: 0.2.6

Published: 2025-10-22T04:53:39Z

Prerelease: no

Release notes:

What's Changed

• bump release to dev version post 0.2.5 by @t-vi in https://github.com/Lightning-AI/lightning-thunder/pull/2513
• torch.cumsum api change by @jjsjann123 in https://github.com/Lightning-AI/lightning-thunder/pull/2507
• DTensor: support linear by @kshitij12345 in https://github.com/Lightning-AI/lightning-thunder/pull/2422
• TEv2 as default TE executor by @riccardofelluga in https://github.com/Lightning-AI/lightning-thunder/pull/2510
• Create Symbol with is_prim=True in _register_custom_op by @crcrpar in https://github.com/Lightning-AI/lightning-thunder/pull/2516
• [pre-commit.ci] pre-commit suggestions by @pre-commit-ci[bot] in https://github.com/Lightning-AI/lightning-thunder/pull/2515
• Use underlying class's __new__ in MutableMappingWrapper.__new__ by @t-vi in https://github.com/Lightning-AI/lightning-thunder/pull/2514
• fix: working README.md example, support nvfuser for torch==2.8 by @lianakoleva in https://github.com/Lightning-AI/lightning-thunder/pull/2525
• Add KaelanDt as codeowner by @t-vi in https://github.com/Lightning-AI/lightning-thunder/pull/2540
• add ci skips by @t-vi in https://github.com/Lightning-AI/lightning-thunder/pull/2547
• TE: Fix cudnn.h not found by @kshitij12345 in https://github.com/Lightning-AI/lightning-thunder/pull/2536
• feat: _register_custom_op supports List[torch.Tensor] by @lianakoleva in https://github.com/Lightning-AI/lightning-thunder/pull/2529
• feat: provide guidance when registering custom op by @lianakoleva in https://github.com/Lightning-AI/lightning-thunder/pull/2530
• fix: string -> f-string where intended by @lianakoleva in https://github.com/Lightning-AI/lightning-thunder/pull/2528
• Initial TE NVFP4 recipe support by @riccardofelluga in https://github.com/Lightning-AI/lightning-thunder/pull/2523
• Add DTensor prim and torch symbol for exp by @kshitij12345 in https://github.com/Lightning-AI/lightning-thunder/pull/2496
• [DTensor] Add prim and torch sym for neg and reciprocal by @kshitij12345 in https://github.com/Lightning-AI/lightning-thunder/pull/2552
• Bump pytest from 8.3.5 to 8.4.2 by @dependabot[bot] in https://github.com/Lightning-AI/lightning-thunder/pull/2567
• Bump bitsandbytes from 0.47.0 to 0.48.0 by @dependabot[bot] in https://github.com/Lightning-AI/lightning-thunder/pull/2565
• Bump diffusers from 0.34.0 to 0.35.1 by @dependabot[bot] in https://github.com/Lightning-AI/lightning-thunder/pull/2564
• switch CI to non-interruptible by @t-vi in https://github.com/Lightning-AI/lightning-thunder/pull/2554
• tests: clean up xfails in VJP tests by @aobolensk in https://github.com/Lightning-AI/lightning-thunder/pull/2578
• fix output dtype for nvfuserex cumsum by @jjsjann123 in https://github.com/Lightning-AI/lightning-thunder/pull/2580
• Use signature (*args, **kwargs) when signature is unavailable by @shino16 in https://github.com/Lightning-AI/lightning-thunder/pull/2542
• fix: call torch.cuda.is_available() in available_devices by @aobolensk in https://github.com/Lightning-AI/lightning-thunder/pull/2602
• Add uint64 to thunder->torch dtype map by @crcrpar in https://github.com/Lightning-AI/lightning-thunder/pull/2519
• Enable direct bindings in Thunder by @rdspring1 in https://github.com/Lightning-AI/lightning-thunder/pull/2502
• move to nvfuser-cu128-torch28 by @t-vi in https://github.com/Lightning-AI/lightning-thunder/pull/2604
• [DTensor] Add torch symbol and prim for _grouped_mm by @kshitij12345 in https://github.com/Lightning-AI/lightning-thunder/pull/2503
• [DTensor] Add prim and torch symbol for add by @kshitij12345 in https://github.com/Lightning-AI/lightning-thunder/pull/2581
• MoE TensorParallel with Eager by @kshitij12345 in https://github.com/Lightning-AI/lightning-thunder/pull/2582
• fix: missing 'import torch' in README.md by @aobolensk in https://github.com/Lightning-AI/lightning-thunder/pull/2608
• Remove E741 from ruff lint ignore rules by @tpremrud in https://github.com/Lightning-AI/lightning-thunder/pull/2601
• Disallow custom_op that mutates arguments by @crcrpar in https://github.com/Lightning-AI/lightning-thunder/pull/2603
• Disabled TF32 on Amper+ devices to stabilize numeric accuracy by @mattteochen in https://github.com/Lightning-AI/lightning-thunder/pull/2579
• Propagate rounding_mode in div_ by @beverlylytle in https://github.com/Lightning-AI/lightning-thunder/pull/2614
• Refactor quantization.py to use TSP by @tejapulagam in https://github.com/Lightning-AI/lightning-thunder/pull/2522
• [DTensor] Add test with parallelize_module by @kshitij12345 in https://github.com/Lightning-AI/lightning-thunder/pull/2598
• Inference benchmark of "meta-llama/Llama-4-Maverick-17B-128E" by @crcrpar in https://github.com/Lightning-AI/lightning-thunder/pull/2487
• Remove outdated scenarios from inference benchmark by @crcrpar in https://github.com/Lightning-AI/lightning-thunder/pull/2619
• Add float4_e2m1fn_x2 to lcdtype_to_nvdtype_map by @crcrpar in https://github.com/Lightning-AI/lightning-thunder/pull/2532
• avoid torch.float4_e2m1fn_x2 in _get_min_and_val by @crcrpar in https://github.com/Lightning-AI/lightning-thunder/pull/2533
• [benchmark_inference] Fix replacing the MoE by @kshitij12345 in https://github.com/Lightning-AI/lightning-thunder/pull/2620
• Fix FSDP NB by @kshitij12345 in https://github.com/Lightning-AI/lightning-thunder/pull/2629
• Fixes function name is not defined when using DebugTransform by @kiya00 in https://github.com/Lightning-AI/lightning-thunder/pull/2617
• Enable MoE TP with thunderfx by @kshitij12345 in https://github.com/Lightning-AI/lightning-thunder/pull/2611
• Add DIV_EXACT prim by @beverlylytle in https://github.com/Lightning-AI/lightning-thunder/pull/2626
• try getting version from nvfuser_direct first by @crcrpar in https://github.com/Lightning-AI/lightning-thunder/pull/2623
• Fix hf example and benchmark run on CPU by @aobolensk in https://github.com/Lightning-AI/lightning-thunder/pull/2583
• getattr should always be taken from the class and then bound by @t-vi in https://github.com/Lightning-AI/lightning-thunder/pull/2584
• Update benchmark_inference.py to support TP with thunderfx by @kshitij12345 in https://github.com/Lightning-AI/lightning-thunder/pull/2625
• Add TE's NVFP4 recipe to the test suite by @riccardofelluga in…