Embedding parallel by attn_tp (#7623)

2025-07-04 23:21:56 -07:00
parent 8e9fb43d82
commit ef8a29c429
1 changed files with 1 additions and 1 deletions
@@ -1930,7 +1930,7 @@ class DeepseekV2Model(nn.Module):
        self.embed_tokens = VocabParallelEmbedding(
            config.vocab_size,
            config.hidden_size,
-            enable_tp=not global_server_args_dict["enable_dp_attention"],
+            use_attn_tp_group=True,
        )
        self.alt_stream = torch.cuda.Stream() if _is_cuda else None
        self.layers = nn.ModuleList(