Skip to content

GitLab

Unverified Commit b166ef20 authored by zzhxxx's avatar zzhxxx Committed by GitHub
Browse files

[refactor] Add prefix support to embed_tokens in DeepSeek MTP (#30788) 


Signed-off-by: default avatarzzhx1 <zzh_201018@outlook.com>
parent 5f2f3fba
Hide whitespace changes
Inline Side-by-side
Showing with 1 addition and 0 deletions
vllm/model_executor/models/deepseek_mtp.py
View file @ b166ef20
...@@ -141,6 +141,7 @@ class DeepSeekMultiTokenPredictor(nn.Module): ...@@ -141,6 +141,7 @@ class DeepSeekMultiTokenPredictor(nn.Module):
self.embed_tokens = VocabParallelEmbedding( self.embed_tokens = VocabParallelEmbedding(
config.vocab_size, config.vocab_size,
config.hidden_size, config.hidden_size,
prefix=maybe_prefix(prefix, "embed_tokens"),
) )
self.logits_processor = LogitsProcessor(config.vocab_size) self.logits_processor = LogitsProcessor(config.vocab_size)
......
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