Closed XinyuShe closed 8 months ago
I hope to use past_key_values for inferring longer texts, but even though I have set use_cache=True, I still cannot find it in the generated_tokens's attributes.
past_key_values
use_cache=True
generated_tokens
generated_tokens = model.generate( **model_input, max_new_tokens=1600, use_cache=True )
Hi @XinyuShe, I'm not sure what you're referring to since the repo doesn't feature a generate() function like the one you're describing.
generate()
I hope to use
past_key_values
for inferring longer texts, but even though I have setuse_cache=True
, I still cannot find it in thegenerated_tokens
's attributes.