Edit model card

Den4ikAI/FRED-T5-XL_instructor_chitchat

Инструкционная модель на FRED-T5-XL. Обратите внимание на промпты в примере чит-чата.

Пример использования [Instruct]

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, GenerationConfig
import torch
use_cuda = torch.cuda.is_available()
device = torch.device("cuda" if use_cuda else "cpu")
tokenizer = AutoTokenizer.from_pretrained("Den4ikAI/FRED-T5-XL_instructor_chitchat")
model = AutoModelForSeq2SeqLM.from_pretrained("Den4ikAI/FRED-T5-XL_instructor_chitchat", torch_dtype=torch.float16).to(device)
model.eval()
generation_config = GenerationConfig.from_pretrained("Den4ikAI/FRED-T5-XL_instructor_chitchat")
def generate(prompt):
  data = tokenizer(f"<SC6>Человек: {prompt}\nБот: <extra_id_0>", return_tensors="pt").to(model.device)
  output_ids = model.generate(
      **data,
      generation_config=generation_config
  )[0]
  print(tokenizer.decode(data["input_ids"][0].tolist()))
  out = tokenizer.decode(output_ids.tolist())
  return out
while 1:
  generate(input(":> "))

Пример использования [Chitchat]

import torch
import transformers

use_cuda = torch.cuda.is_available()
device = torch.device("cuda" if use_cuda else "cpu")

t5_tokenizer = transformers.GPT2Tokenizer.from_pretrained("Den4ikAI/FRED-T5-XL_instructor_chitchat")
t5_model = transformers.T5ForConditionalGeneration.from_pretrained("Den4ikAI/FRED-T5-XL_instructor_chitchat")
generation_config = transformers.GenerationConfig.from_pretrained("Den4ikAI/FRED-T5-XL_instructor_chitchat")

while True:
    print('-'*80)
    dialog = []
    while True:
        msg = input('H:> ').strip()
        if len(msg) == 0:
            break
        msg = msg[0].upper() + msg[1:]
        dialog.append('Собеседник сказал: ' + msg)
        # Данный пример промпта позволяет вести диалог и писать инструкции.
        # prompt = '<SC6>Тебя зовут Анфиса. Тебе интересно машинное обучение.' + '\n'.join(dialog) + '\nТы ответил: <extra_id_0>'
        # Второй пример - промпт просто для диалогов. В таком режиме не будет глюков, когда модель кидает кусок промпта в ответ.
        prompt = '<SC1>Тебя зовут Анфиса. Тебе интересно машинное обучение.' + '\n'.join(dialog) + '\nТы ответил: <extra_id_0>'

        input_ids = t5_tokenizer(prompt, return_tensors='pt').input_ids
        out_ids = t5_model.generate(input_ids=input_ids.to(device), generation_config=generation_config)
        t5_output = t5_tokenizer.decode(out_ids[0][1:])
        if '</s>' in t5_output:
            t5_output = t5_output[:t5_output.find('</s>')].strip()

        t5_output = t5_output.replace('<extra_id_0>', '').strip()
        t5_output = t5_output.split('Собеседник')[0].strip()
        print('B:> {}'.format(t5_output))
        dialog.append('Ты ответил: ' + t5_output)

Citation

@MISC{Den4ikAI/FRED-T5-XL_instructor_chitchat,
    author  = {Denis Petrov},
    title   = {Russian Instruct and Chitchat model},
    url     = {https://huggingface.co/Den4ikAI/FRED-T5-XL_instructor_chitchat/},
    year    = 2023
}
Downloads last month
17
Safetensors
Model size
1.74B params
Tensor type
BF16
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.