diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py index 5c3b491904..364c0444bd 100644 --- a/unsloth/models/llama.py +++ b/unsloth/models/llama.py @@ -1169,7 +1169,8 @@ def CausalLM_fast_forward(fast_forward_inference): if not return_dict: output = (logits,) + outputs[1:] return (loss,) + output if loss is not None else output - print(outputs.past_key_values) + print(outputs.past_key_values, outputs.past_key_values[0][0].shape) + raise return CausalLMOutputWithPast( loss = loss, logits = logits,