app.py

from transformers import AutoModelForCausalLM, AutoTokenizer, AutoModelForSeq2SeqLM, T5ForConditionalGeneration, T5Tokenizer
tokenizer = AutoTokenizer.from_pretrained("microsoft/DialoGPT-large")
model = AutoModelForCausalLM.from_pretrained("microsoft/DialoGPT-large")
grammar_tokenizer = T5Tokenizer.from_pretrained('deep-learning-analytics/GrammarCorrector')
grammar_model = T5ForConditionalGeneration.from_pretrained('deep-learning-analytics/GrammarCorrector')
import torch
import gradio as gr


# def chat(message, history):
#     history = history if history is not None else []
#     new_user_input_ids = tokenizer.encode(message+tokenizer.eos_token, return_tensors='pt')
#     bot_input_ids = torch.cat([torch.LongTensor(history), new_user_input_ids], dim=-1)
#     history = model.generate(bot_input_ids, max_length=500, pad_token_id=tokenizer.eos_token_id).tolist()
#     # response = tokenizer.decode(history[0]).replace("<|endoftext|>", "\n")
#     # pretty print last ouput tokens from bot
#     response =  tokenizer.decode(bot_input_ids.shape[-1][0], skip_special_tokens=True)
#     print("The response is ", [response])
#     # history.append((message, response, new_user_input_ids, chat_history_ids))
#     return response, history, feedback(message)


def chat(message, history=[]):
    new_user_input_ids = tokenizer.encode(message+tokenizer.eos_token, return_tensors='pt')
    if  len(history) > 0:
        last_set_of_ids = history[len(history)-1][2]
        bot_input_ids = torch.cat([last_set_of_ids, new_user_input_ids], dim=-1) 
    else:
        bot_input_ids = new_user_input_ids
    chat_history_ids = model.generate(bot_input_ids, max_length=5000, pad_token_id=tokenizer.eos_token_id)
    response_ids = chat_history_ids[:, bot_input_ids.shape[-1]:][0]
    response = tokenizer.decode(response_ids, skip_special_tokens=True)
    history.append((message, response, chat_history_ids))
    return history, history, feedback(message)


def feedback(text):
    num_return_sequences=1
    batch =  grammar_tokenizer([text],truncation=True,padding='max_length',max_length=64, return_tensors="pt")
    corrections= grammar_model.generate(**batch,max_length=64,num_beams=2, num_return_sequences=num_return_sequences, temperature=1.5)
    corrected_text = tokenizer.decode(corrections[0], clean_up_tokenization_spaces=True, skip_special_tokens=True)
    print("The corrections are: ", corrections)
    if corrected_text == text:
        feedback = f'Looks good! Keep up the good work'
    else:
        feedback = f'\'{" ".join(corrected_text)}\' might be a little better'
    return feedback

iface = gr.Interface(
    chat,
    ["text", "state"],
    ["chatbot", "state", "text"],
    allow_screenshot=False,
    allow_flagging="never",
)
iface.launch()