diff --git a/vllm/model_executor/models/diffusion_gemma.py b/vllm/model_executor/models/diffusion_gemma.py index 59fe7e22ba..a700107ddb 100644 --- a/vllm/model_executor/models/diffusion_gemma.py +++ b/vllm/model_executor/models/diffusion_gemma.py @@ -250,6 +250,8 @@ class DiffusionGemmaForConditionalGeneration( self.lm_head = ParallelLMHead( num_embeddings=text_config.vocab_size, embedding_dim=text_config.hidden_size, + quant_config=vllm_config.quant_config, + prefix=maybe_prefix(prefix, "lm_head"), ) if text_config.tie_word_embeddings: