Skip to content

GitLab

Unverified Commit ef8a29c4 authored by Gang Chen's avatar Gang Chen Committed by GitHub
Browse files

Embedding parallel by attn_tp (#7623)

parent 8e9fb43d
Hide whitespace changes
Inline Side-by-side
Showing with 1 addition and 1 deletion
python/sglang/srt/models/deepseek_v2.py
View file @ ef8a29c4
...@@ -1930,7 +1930,7 @@ class DeepseekV2Model(nn.Module): ...@@ -1930,7 +1930,7 @@ class DeepseekV2Model(nn.Module):
self.embed_tokens = VocabParallelEmbedding( self.embed_tokens = VocabParallelEmbedding(
config.vocab_size, config.vocab_size,
config.hidden_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.alt_stream = torch.cuda.Stream() if _is_cuda else None
self.layers = nn.ModuleList( self.layers = nn.ModuleList(
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment