Skip to content

GitLab

Unverified Commit f5db6385 authored by Grzegorz K. Karch's avatar Grzegorz K. Karch Committed by GitHub
Browse files

Fix nemotron_nas intermediate_size computation (#30795) 


Signed-off-by: default avatarGrzegorz Karch <gkarch@nvidia.com>
parent c0a88df7
Hide whitespace changes
Inline Side-by-side
Showing with 7 additions and 4 deletions
vllm/model_executor/models/nemotron_nas.py
View file @ f5db6385
...@@ -169,10 +169,13 @@ class DeciLMDecoderLayer(nn.Module): ...@@ -169,10 +169,13 @@ class DeciLMDecoderLayer(nn.Module):
self.input_layernorm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps) self.input_layernorm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
if not self._is_no_op_ffn: if not self._is_no_op_ffn:
ffn_mult = block_config.ffn.ffn_mult if hasattr(block_config.ffn, "ffn_mult"):
intermediate_size = _ffn_mult_to_intermediate_size( ffn_mult = block_config.ffn.ffn_mult
ffn_mult, config.hidden_size intermediate_size = _ffn_mult_to_intermediate_size(
) ffn_mult, config.hidden_size
)
else:
intermediate_size = block_config.ffn.intermediate_size
self.mlp = LlamaMLP( self.mlp = LlamaMLP(
hidden_size=self.hidden_size, hidden_size=self.hidden_size,
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment