Add DeepSeek V3/R1 shared experts fusion (#4918)

This commit is contained in:
Xiaoyu Zhang
2025-04-04 16:59:29 +08:00
committed by GitHub
parent 6ff9c6a5e7
commit 924ca7c92c
14 changed files with 536 additions and 36 deletions

View File

@@ -157,6 +157,8 @@ class ModelRunner:
"flashinfer_mla_disable_ragged": server_args.flashinfer_mla_disable_ragged,
"debug_tensor_dump_output_folder": server_args.debug_tensor_dump_output_folder,
"debug_tensor_dump_inject": server_args.debug_tensor_dump_inject,
"n_share_experts_fusion": server_args.n_share_experts_fusion,
"disable_shared_experts_fusion": server_args.disable_shared_experts_fusion,
}
)