mmarch/text-generation-webui
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Add a tokenizer placeholder

Browse source
This commit is contained in:
oobabooga 2023-03-03 15:16:55 -03:00
parent 4cc36dc434
commit a345a2acd2
1 changed files with 4 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

5
modules/text_generation.py
View file

@ -21,8 +21,11 @@ def get_max_prompt_length(tokens):
return max_length return max_length
def encode(prompt, tokens_to_generate=0, add_special_tokens=True): def encode(prompt, tokens_to_generate=0, add_special_tokens=True):
# These models do not have explicit tokenizers for now, so
# we return an estimate on the number of tokens
if shared.is_RWKV or shared.is_LLaMA: if shared.is_RWKV or shared.is_LLaMA:
return prompt return np.zeros((1, len(prompt)//5))
input_ids = shared.tokenizer.encode(str(prompt), return_tensors='pt', truncation=True, max_length=get_max_prompt_length(tokens_to_generate), add_special_tokens=add_special_tokens) input_ids = shared.tokenizer.encode(str(prompt), return_tensors='pt', truncation=True, max_length=get_max_prompt_length(tokens_to_generate), add_special_tokens=add_special_tokens)
if shared.args.cpu: if shared.args.cpu: