Amir Samani 28579b55fa add stream to cuda graph catpure ...

Signed-off-by: Amir Samani <asamani@nvidia.com>

2025-11-21 14:23:45 -08:00

..

__init__.py

…

activation_quant_fusion.py

…

backends.py

[torch.compile] caching of config fields should be opt-out by default (#26468)

2025-11-19 06:13:54 -08:00

base_static_graph.py

…

caching.py

Fixes bench (#29058)

2025-11-20 21:21:54 +00:00

collective_fusion.py

[PERF] Allreduce fusion. Support torch native matching. Tuning of the thresholds (#24248)

2025-11-10 18:33:11 -05:00

compiler_interface.py

[Doc]: fix typos in various files (#28567)

2025-11-15 19:27:50 +00:00

counter.py

…

cuda_graph.py

add stream to cuda graph catpure

2025-11-21 14:23:45 -08:00

decorators.py

[Doc]: fix typos in various files (#28567)

2025-11-15 19:27:50 +00:00

fix_functionalization.py

[Kernel][Perf] fuse QK Norm and RoPE into one cuda kernel for Qwen Model (#27165)

2025-11-11 12:00:31 -05:00

fusion_attn.py

[Chore]:Extract math and argparse utilities to separate modules (#27188)

2025-10-26 04:03:32 -07:00

fusion.py

[Kernel][Perf] fuse QK Norm and RoPE into one cuda kernel for Qwen Model (#27165)

2025-11-11 12:00:31 -05:00

fx_utils.py

…

inductor_pass.py

…

matcher_utils.py

[kernel][perf] support uncontiguous input for rms_norm kernel (#28103)

2025-11-20 19:39:09 -08:00

monitor.py

[Log] Optimize Startup Log (#26740)

2025-10-24 19:27:04 -04:00

noop_elimination.py

…

partition_rules.py

remove resolve_op_overloads and use splitting_ops directly (#28081)

2025-11-08 01:13:13 +00:00

pass_manager.py

[torch.compile] caching of config fields should be opt-out by default (#26468)

2025-11-19 06:13:54 -08:00

piecewise_backend.py

…

post_cleanup.py

…

qk_norm_rope_fusion.py

[Kernel][Perf] fuse QK Norm and RoPE into one cuda kernel for Qwen Model (#27165)

2025-11-11 12:00:31 -05:00

sequence_parallelism.py

[compile] Enable sequence parallelism matching w/o custom ops enabled (#27126)

2025-11-15 11:46:12 +00:00

torch25_custom_graph_pass.py

…

vllm_inductor_pass.py

…

wrapper.py

Avoid bytecode hook and simplify TorchCompileWrapperWithCustomDipatch (#25110)

2025-11-14 14:11:10 -08:00