[Bugfix] Capture prefill kernels for mixed FULL graphs (#58275)

Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com>
Co-authored-by: OpenAI Codex <codex@openai.com>
This commit is contained in:
Andreas Karatzas
2026-09-24 02:05:07 -07:00
committed by GitHub
co-authored by OpenAI Codex
parent bbc4ddeca9
commit 23110a0c9e
4 changed files with 117 additions and 16 deletions
@@ -57,6 +57,7 @@ class SpeculatorCudaGraphManager(CudaGraphManager):
attn_groups,
kv_cache_config,
full_cudagraph=desc.cg_mode == CUDAGraphMode.FULL,
max_query_len=desc.max_query_len or desc.uniform_token_count,
)
return lambda cg_mode: forward_fn(