xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-24 14:06:32 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix nemotron_nas intermediate_size computation (#30795)

Browse Source 
Signed-off-by: Grzegorz Karch <gkarch@nvidia.com>
This commit is contained in:
Grzegorz K. Karch 2025-12-17 02:06:28 +01:00 committed by GitHub
parent c0a88df7f7
commit f5db6385a1
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 7 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
vllm/model_executor/models/nemotron_nas.py
View File

@ -169,10 +169,13 @@ class DeciLMDecoderLayer(nn.Module):
self.input_layernorm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
if not self._is_no_op_ffn:
ffn_mult = block_config.ffn.ffn_mult
intermediate_size = _ffn_mult_to_intermediate_size(
ffn_mult, config.hidden_size
)
if hasattr(block_config.ffn, "ffn_mult"):
ffn_mult = block_config.ffn.ffn_mult
intermediate_size = _ffn_mult_to_intermediate_size(
ffn_mult, config.hidden_size
)
else:
intermediate_size = block_config.ffn.intermediate_size
self.mlp = LlamaMLP(
hidden_size=self.hidden_size,