nvjullin 7ea22e42d5 [Misc] Add override for allreduce fusion thresholds (#23639) ...

Signed-off-by: Julien Lin <jullin@nvidia.com>

2025-08-26 15:53:04 +00:00

..

__init__.py

[torch.compile] avoid Dynamo guard evaluation overhead (#7898)

2024-08-28 16:10:12 -07:00

activation_quant_fusion.py

[Misc] Add SPDX-FileCopyrightText (#19100)

2025-06-03 11:20:17 -07:00

backends.py

Remove graph_pool as member of VllmBackend and argument to CUDAGraphWrapper (#23385)

2025-08-25 19:34:15 -07:00

base_static_graph.py

Remove graph_pool as member of VllmBackend and argument to CUDAGraphWrapper (#23385)

2025-08-25 19:34:15 -07:00

collective_fusion.py

[Misc] Add override for allreduce fusion thresholds (#23639)

2025-08-26 15:53:04 +00:00

compiler_interface.py

Add test case for compiling multiple graphs (#21044)

2025-07-23 11:00:47 -07:00

counter.py

[Bugfix] VLLM_V1 supports passing other compilation levels (#19340)

2025-07-29 09:35:58 -04:00

cuda_graph.py

Remove graph_pool as member of VllmBackend and argument to CUDAGraphWrapper (#23385)

2025-08-25 19:34:15 -07:00

cuda_piecewise_backend.py

[Core] Allow full cudagraph with separate attention routines and orthogonal to compilation, add support for FA2 and FlashInfer (#20059)

2025-08-15 10:01:39 -04:00

decorators.py

[torch.compile] Support conditional torch.compile per module (#22269)

2025-08-20 16:52:59 +00:00

fix_functionalization.py

[Misc] Add SPDX-FileCopyrightText (#19100)

2025-06-03 11:20:17 -07:00

fusion_attn.py

[NVIDIA][torch.compile] Support Flashinfer TRTLLM FP8-q/kv NVFP4-out Attention Kernel (#22703)

2025-08-22 22:09:05 +00:00

fusion.py

fix incompatibililty with non cuda platform for nvfp4 (#23478)

2025-08-24 15:35:41 -07:00

fx_utils.py

[torch.compile][ROCm] Fuse quantization onto attention using a torch.compile pass (#16756)

2025-06-12 08:31:04 -07:00

inductor_pass.py

[BugFix] Update AttnFusionPass cache key (#21947)

2025-08-01 07:11:29 -07:00

monitor.py

[Core] Allow full cudagraph with separate attention routines and orthogonal to compilation, add support for FA2 and FlashInfer (#20059)

2025-08-15 10:01:39 -04:00

multi_output_match.py

[Misc] Add SPDX-FileCopyrightText (#19100)

2025-06-03 11:20:17 -07:00

noop_elimination.py

[torch.compile][ROCm] Fuse quantization onto attention using a torch.compile pass (#16756)

2025-06-12 08:31:04 -07:00

pass_manager.py

[ROCm][Bugfix] Compilation passes fix (#22202)

2025-08-04 19:12:28 -07:00

sequence_parallelism.py

[Feature] Add async tensor parallelism for scaled mm (#20155)

2025-07-30 17:23:41 -04:00

torch25_custom_graph_pass.py

[Misc] Add SPDX-FileCopyrightText (#19100)

2025-06-03 11:20:17 -07:00

vllm_inductor_pass.py

[Bugfix] Upgrade depyf to 0.19 and streamline custom pass logging (#20777)

2025-07-11 00:15:22 -07:00

wrapper.py

[Core] Allow full cudagraph with separate attention routines and orthogonal to compilation, add support for FA2 and FlashInfer (#20059)

2025-08-15 10:01:39 -04:00