Fix LlamaTokenizer eos_token (attempt)

This commit is contained in:
oobabooga 2023-04-09 21:19:39 -03:00
parent 120f5662cf
commit a3085dba07

View File

@ -174,6 +174,9 @@ def load_model(model_name):
tokenizer = AutoTokenizer.from_pretrained(Path(f"{shared.args.model_dir}/gpt-j-6B/"))
elif type(model) is transformers.LlamaForCausalLM:
tokenizer = LlamaTokenizer.from_pretrained(Path(f"{shared.args.model_dir}/{shared.model_name}/"), clean_up_tokenization_spaces=True)
tokenizer.eos_token_id = 2
tokenizer.bos_token_id = 1
tokenizer.pad_token_id = 0
else:
tokenizer = AutoTokenizer.from_pretrained(Path(f"{shared.args.model_dir}/{shared.model_name}/"))
tokenizer.truncation_side = 'left'