Revert "Embedding parallel by attn_tp (#7623)" (#7880)

b6b6268c · Yineng Zhang · GitHub · 08702321 · b6b6268c
Unverified Commit b6b6268c authored Jul 08, 2025 by Yineng Zhang Committed by GitHub Jul 08, 2025
Hide whitespace changes
Inline Side-by-side

Showing with 1 addition and 1 deletion

python/sglang/srt/models/deepseek_v2.py python/sglang/srt/models/deepseek_v2.py +1 -1

No files found.
--- a/python/sglang/srt/models/deepseek_v2.py
+++ b/python/sglang/srt/models/deepseek_v2.py
@@ -1938,7 +1938,7 @@ class DeepseekV2Model(nn.Module):
        self.embed_tokens = VocabParallelEmbedding(
            config.vocab_size,
            config.hidden_size,
-            use_attn_tp_group=True,
+            enable_tp=not global_server_args_dict["enable_dp_attention"],
        )
        self.alt_stream = torch.cuda.Stream() if _is_cuda else None
        self.layers = nn.ModuleList(