Consider the softprompt in the maximum prompt length calculation

This commit is contained in:
oobabooga 2023-02-14 12:06:47 -03:00
parent 8b3bb512ef
commit d910d435cd

View file

@ -247,8 +247,15 @@ def fix_galactica(s):
s = s.replace(r'$$', r'$')
return s
def get_max_prompt_length(tokens):
global soft_prompt, soft_prompt_tensor
max_length = 2048-tokens
if soft_prompt:
max_length -= soft_prompt_tensor.shape[1]
return max_length
def encode(prompt, tokens_to_generate=0, add_special_tokens=True):
input_ids = tokenizer.encode(str(prompt), return_tensors='pt', truncation=True, max_length=2048-tokens_to_generate, add_special_tokens=add_special_tokens)
input_ids = 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 args.cpu:
return input_ids
elif args.deepspeed:
@ -497,7 +504,8 @@ def generate_chat_prompt(text, tokens, name1, name2, context, history_size, impe
rows = [f"{context.strip()}\n"]
i = len(history['internal'])-1
count = 0
while i >= 0 and len(encode(''.join(rows), tokens)[0]) < 2048-tokens:
max_length = get_max_prompt_length(tokens)
while i >= 0 and len(encode(''.join(rows), tokens)[0]) < max_length:
rows.insert(1, f"{name2}: {history['internal'][i][1].strip()}\n")
count += 1
if not (history['internal'][i][0] == '<|BEGIN-VISIBLE-CHAT|>'):
@ -515,7 +523,7 @@ def generate_chat_prompt(text, tokens, name1, name2, context, history_size, impe
rows.append(f"{name1}:")
limit = 2
while len(rows) > limit and len(encode(''.join(rows), tokens)[0]) >= 2048-tokens:
while len(rows) > limit and len(encode(''.join(rows), tokens)[0]) >= max_length:
rows.pop(1)
rows.pop(1)