Fix rollout_model init (#2881)

This commit is contained in:
Yuanle Liu
2025-07-17 13:36:21 +08:00
committed by GitHub
parent 1f15ca21e4
commit dbb9e2506b
9 changed files with 76 additions and 312 deletions

View File

@@ -58,7 +58,6 @@ class VocabParallelEmbedding(nn.Layer):
self.column_cut = False
self.world_size: int = hcg.get_model_parallel_world_size()
self.ring_id: int = hcg.get_model_parallel_group().id
self.use_rope: bool = fd_config.model_config.use_rope
self.use_ep: bool = fd_config.parallel_config.use_ep
self.hidden_dropout_prob: float = fd_config.model_config.hidden_dropout_prob
self.initializer_range: float = fd_config.model_config.initializer_range
@@ -92,14 +91,6 @@ class VocabParallelEmbedding(nn.Layer):
self.embeddings.weight.is_distributed = True
self.embeddings.weight.split_axis = 1
if not self.use_rope:
self.position_embeddings = nn.Embedding(
self.max_position_embeddings,
embedding_dim,
weight_attr=paddle.ParamAttr(initializer=nn.initializer.Normal(
mean=0.0, std=self.initializer_range), ),
)
self.prefix = prefix
self.dropout = nn.Dropout(self.hidden_dropout_prob)