karylab_agents/vllm
1
0
Fork 0
forked from Karylab-cklius/vllm
Code Pull Requests 2 Actions 1 Packages Activity

Move Roberta remaining nn.Embedding to VocabParallelEmbedding (#47452)

Browse Source 
Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
This commit is contained in:
Max de Bayser
2026-07-03 09:47:50 +00:00
committed by GitHub
parent 4875b4456b
commit bbdcbe4686
1 changed files with 3 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
vllm/model_executor/models/roberta.py
+3 -5
View File
@@ -53,13 +53,11 @@ class RobertaEmbedding(nn.Module):
config.vocab_size, config.hidden_size
)
self.padding_idx = config.pad_token_id
self.position_embeddings = nn.Embedding(
config.max_position_embeddings,
config.hidden_size,
padding_idx=self.padding_idx,
self.position_embeddings = VocabParallelEmbedding(
config.max_position_embeddings, config.hidden_size
)
self.token_type_embeddings = nn.Embedding(
self.token_type_embeddings = VocabParallelEmbedding(
config.type_vocab_size, config.hidden_size
)
self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)