From d6eca0c13037bc6a76e074cfa345150785016e7c Mon Sep 17 00:00:00 2001 From: Sage Moore Date: Mon, 2 Jun 2025 18:03:21 +0000 Subject: [PATCH] remove modular kernel Signed-off-by: Sage Moore --- vllm/model_executor/layers/fused_moe/modular_kernel.py | 4 ---- 1 file changed, 4 deletions(-) diff --git a/vllm/model_executor/layers/fused_moe/modular_kernel.py b/vllm/model_executor/layers/fused_moe/modular_kernel.py index c0e89ff6c40dc..7d3ddf8f14c4d 100644 --- a/vllm/model_executor/layers/fused_moe/modular_kernel.py +++ b/vllm/model_executor/layers/fused_moe/modular_kernel.py @@ -3,7 +3,6 @@ from abc import ABC, abstractmethod from typing import Optional import torch -from vllm.v1.worker.ubatching import get_current_ubatch_context, dump_ubatching_state # # This file defines a set of base classes used to make MoE kernels more modular. @@ -359,9 +358,6 @@ class FusedMoEModularKernel(torch.nn.Module): expert_num_tokens=expert_num_tokens, ) - # if (ubatch_ctx := get_current_ubatch_context()) is not None: - # print("in modular moe3, ubatch:", ubatch_ctx.id, self.fused_experts) - self.prepare_finalize.finalize(output, fused_out, topk_weights, topk_ids, apply_router_weight_on_input)