[Feature] Marlin MoE backend supports DeepseekV3 (#2962)

Co-authored-by: K11OntheBoat <“ruianmaidanglao@163.com”>

fastdeploy/model_executor/layers/moe/fused_moe_marlin_backend.py

@@ -22,10 +22,28 @@ from fastdeploy.distributed.communication import tensor_model_parallel_all_reduc
 from fastdeploy.model_executor.ops.gpu import (
     MoeWna16MarlinGemmApi,
     tritonmoe_preprocess_func,
+    noaux_tc,
 )
 
 from ..quantization.quant_base import QuantMethodBase
 
+def get_moe_scores(gating_output: paddle.Tensor, n_group, topk_group, top_k,
+                   routed_scaling_factor,
+                   e_score_correction_bias) -> paddle.Tensor:
+    """
+    compute moe scores using e_score_correction_bias.
+    """
+    scores = paddle.nn.functional.sigmoid(gating_output)
+    scores_with_bias = scores + e_score_correction_bias.unsqueeze(0)
+    scores = noaux_tc(
+        scores,
+        scores_with_bias,
+        n_group,
+        topk_group,
+        top_k,
+        routed_scaling_factor,
+    )
+    return scores
 
 def gptq_marlin_moe_repack(
     b_q_weight: paddle.Tensor,
@@ -205,14 +223,23 @@ class MarlinWeightOnlyMoEMethod(QuantMethodBase):
         moe_intermediate_size = layer.moe_intermediate_size
         hidden_size = layer.hidden_size
         num_experts = layer.num_experts
+        topk_method = layer.topk_method
 
-        topk_ids, topk_weights = fastdeploy.model_executor.ops.gpu.moe_topk_select(
-            gate_out,
-            layer.gate_correction_bias,
-            top_k,
-            True,  # apply_norm_weight,
-            False,
-        )
+        if topk_method == "noaux_tc":
+            gate_out = get_moe_scores(gate_out, layer.n_group,
+                            layer.topk_group, layer.top_k,
+                            layer.routed_scaling_factor,
+                            layer.gate_correction_bias)
+
+            topk_weights, topk_ids = paddle.topk(gate_out, k=layer.top_k, axis=-1, sorted=False)
+        else:
+            topk_ids, topk_weights = fastdeploy.model_executor.ops.gpu.moe_topk_select(
+                gate_out,
+                layer.gate_correction_bias,
+                top_k,
+                True,  # apply_norm_weight,
+                False,
+            )
 
         block_size_m = 64
 
@@ -291,7 +318,7 @@ class MarlinWeightOnlyMoEMethod(QuantMethodBase):
         ffn_out.reshape_([token_num, -1, hidden_size])
         ffn_out = ffn_out.sum(axis=1)
 
-        if layer.tp_size > 1:
+        if layer.reduce_results and layer.tp_size > 1:
             tensor_model_parallel_all_reduce(ffn_out)
 
         return ffn_out