Open kjappelbaum opened 4 weeks ago
perhaps not for batch inference
For the llama runs we do not use mattext tokenizers though.
Ah I see now. There was this issue of Llama tokenizer not including pad token. So we set tokenizer.pad_token = tokenizer.eos_token ref.
tokenizer.pad_token = tokenizer.eos_token
We also tried adding a token, this then resized the vocab and creates a set of problems
perhaps not for batch inference