xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-07-09 10:57:10 +08:00
Code Issues Packages Projects Releases Wiki Activity

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

Browse Source 
Signed-off-by: zzhx1 <zzh_201018@outlook.com>
This commit is contained in:
zzhxxx 2025-12-18 12:45:56 +08:00 committed by GitHub
parent 5f2f3fba1d
commit b166ef20e1
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 1 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
vllm/model_executor/models/deepseek_mtp.py
View File

@ -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)