Hi,
I noticed that the model_max_length argument in tokenizer_config.json is set to 2048. Shouldn't this be 8192 for an 8k context model?
model_max_length
tokenizer_config.json
· Sign up or log in to comment