[6/N] MoE Refactor: Cleanup MoE-related configs (#8849)

2025-08-14 21:14:53 -07:00
parent 584e1ab2d0
commit 295895120d
69 changed files with 956 additions and 1037 deletions
--- a/python/sglang/srt/model_executor/model_runner.py
+++ b/python/sglang/srt/model_executor/model_runner.py
@@ -60,7 +60,6 @@ from sglang.srt.layers.dp_attention import (
    initialize_dp_attention,
 )
 from sglang.srt.layers.logits_processor import LogitsProcessorOutput
-from sglang.srt.layers.moe.utils import DeepEPMode, MoeA2ABackend
 from sglang.srt.layers.quantization import (
    deep_gemm_wrapper,
    monkey_patch_isinstance_for_vllm_base_layer,
@@ -219,8 +218,6 @@ class ModelRunner:
                # TODO it is indeed not a "server args"
                "use_mla_backend": self.use_mla_backend,
                "speculative_algorithm": self.spec_algorithm,
-                "moe_a2a_backend": MoeA2ABackend(server_args.moe_a2a_backend),
-                "deepep_mode": DeepEPMode(server_args.deepep_mode),
            }
        )