FRED-T5-RefinedPersonaChat

morzecrew/FRED-T5-RefinedPersonaChat

This model is a fine-tuned version of ai-forever/FRED-T5-1.7B on the RefinedPersonaChat. Inspired by SiberiaSoft/SiberianPersonaFred blogpost but dataset was improved to prevent toxic speech.

Prompt tips:

You can provide personal information form bot identity, name, age and etc..

Inference

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("morzecrew/RefinedPersonaChat")
t5_model = transformers.T5ForConditionalGeneration.from_pretrained("morzecrew/RefinedPersonaChat")


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: ' + msg)
        # В начале ставится промпт персонажа.
        prompt = '<SC6>Ты парень, консультант по разным вопросам. Ты очень умный. Любишь помогать собеседнику. Продолжи диалог:' + '\n'.join(dialog) + '\nТы: <extra_id_0>'

        input_ids = t5_tokenizer(prompt, return_tensors='pt').input_ids
        out_ids = t5_model.generate(input_ids=input_ids.to(device), do_sample=True, temperature=0.9, max_new_tokens=512, top_p=0.85,
                                      top_k=2, repetition_penalty=1.2)
        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: ' + t5_output)

---

Citation

@MISC{morzecrew/FRED-T5-RefinedPersonaChat,
    author  = {Yuri Zaretskiy, Nikolas Ivanov, Igor Kuzmin},
    title   = {Dialogue model for conversational agents},
    url     = {https://huggingface.co/morzecrew/FRED-T5-RefinedPersonaChat},
    year    = 2023
}