Skip to content

Fix Qwen3.5 AOTI kernel packaging on CUDA #22758

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

Open
Gasoonjia wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Fix Qwen3.5 AOTI kernel packaging on CUDA #22758

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 7 additions & 5 deletions examples/models/qwen3_5_moe/export.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1138,9 +1138,11 @@ def _export_cuda(model, config, args):
"enable_dynamic_shape": True,
"get_mutable_buffer_metadata": _mutable_buffer_metadata_json(model),
}
# Avoid PyTorch 2.13's fused cast/reduction compile-time autotune path. Keep
# max autotuning enabled so prefill GEMMs use the CUDA backend's Triton-only
# GEMM lowering instead of emitting libtorch fallback kernels.
# Keep compile-time autotuning enabled so AOTI embeds the selected Triton
# kernels in the shared object. PyTorch 2.14's lazy JIT path otherwise leaves
# runtime references to temporary kernel files that disappear after export.
# Max autotuning is also needed so prefill GEMMs use the CUDA backend's
# Triton-only GEMM lowering instead of emitting libtorch fallback kernels.
et_prog = to_edge_transform_and_lower(
{"decode": decode_ep, "prefill": prefill_ep},
partitioner={
Expand All @@ -1151,7 +1153,7 @@ def _export_cuda(model, config, args):
CompileSpec("low_memory_mode", b"ON"),
CompileSpec("emulate_precision_casts", b"OFF"),
CompileSpec("max_autotune", b"ON"),
CompileSpec("autotune_at_compile_time", b"OFF"),
CompileSpec("autotune_at_compile_time", b"ON"),
]
)
],
Expand All @@ -1162,7 +1164,7 @@ def _export_cuda(model, config, args):
CompileSpec("low_memory_mode", b"ON"),
CompileSpec("emulate_precision_casts", b"OFF"),
CompileSpec("max_autotune", b"ON"),
CompileSpec("autotune_at_compile_time", b"OFF"),
CompileSpec("autotune_at_compile_time", b"ON"),
]
)
],
Expand Down
Loading