Spaces:
Runtime error
Runtime error
File size: 1,671 Bytes
591004d |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 |
#!python
# -*- coding: utf-8 -*-
# @author: Kun
from models.vicuna_bin import max_token, temperature, top_p
from common import torch_gc
from global_config import lang_opt
def get_api_response(model, tokenizer, content: str, max_tokens=None):
if "en" == lang_opt:
system_role_content = 'You are a helpful and creative assistant for writing novel.'
elif "zh1" == lang_opt:
system_role_content = 'You are a helpful and creative assistant for writing novel.\
You are must always in Chinese.重要，你需要使用中文与我进行交流。'
elif "zh2" == lang_opt:
system_role_content = '你是写小说的好帮手，有创意的助手。'
else:
raise Exception(f"not supported language: {lang_opt}")
print("===> Question:")
print(content)
print("<==="+"="*100)
content = content.encode()
tokens = model.tokenize(content)
output = b""
count = 0
token_count = 10000
top_k = 40
repetition_penalty = 1.1
for token in model.generate(tokens,
top_k=top_k,
top_p=top_p,
temp=temperature,
repeat_penalty=repetition_penalty):
text = model.detokenize([token])
# print(text)
output += text
count += 1
if count >= token_count or (token == model.token_eos()):
break
response = output.decode()
# print("===> [vicuna][generate] response: {}".format(response))
torch_gc()
print("===> Generated Text: ")
print(response)
print("<==="+"="*100)
return response
|