Efficient-Large-Model
/

Fast_dLLM_v2_1.5B

Model card Files Files and versions

Chengyue Wu commited on Oct 5

Commit

0f95288

·

1 Parent(s): da56081

update

Files changed (1) hide show

modeling.py +0 -1

modeling.py CHANGED Viewed

@@ -555,7 +555,6 @@ class Fast_dLLM_QwenForCausalLM(Fast_dLLM_QwenPreTrainedModel, GenerationMixin):
         top_p=0.95,
         temperature=0,
         use_block_cache=False,
-        block_cache_refresh_interval=16,
         **kwargs
     ):
         num_blocks = max_new_tokens // block_size

         top_p=0.95,
         temperature=0,
         use_block_cache=False,
         **kwargs
     ):
         num_blocks = max_new_tokens // block_size