Skip to content

Commit bfa9f7f

Browse files
authored
tweak ds3 model.py to reflect main branch for DS3 baseline can run (#1684)
1 parent 45647b3 commit bfa9f7f

File tree

1 file changed

+1
-3
lines changed
  • torchtitan/models/deepseek_v3/model

1 file changed

+1
-3
lines changed

torchtitan/models/deepseek_v3/model/model.py

Lines changed: 1 addition & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -270,9 +270,7 @@ def __init__(self, layer_id: int, model_args: DeepSeekV3ModelArgs):
270270
self.attention = Attention(model_args)
271271
self.attention_norm = nn.RMSNorm(model_args.dim, eps=model_args.norm_eps)
272272
self.ffn_norm = nn.RMSNorm(model_args.dim, eps=model_args.norm_eps)
273-
# self.moe_enabled = layer_id >= model_args.n_dense_layers
274-
# TODO: enable me when local_map works
275-
self.moe_enabled = False
273+
self.moe_enabled = layer_id >= model_args.n_dense_layers
276274

277275
if self.moe_enabled:
278276
self.moe = MoE(

0 commit comments

Comments
 (0)