History and old tokens

[galaxyentity904]

I have tried to implement DialoGPT using this code

# encode the new user input, add the eos_token and return a tensor in Pytorch
    new_user_input_ids = tokenizer.encode(input(genpromptfinal) + tokenizer.eos_token, return_tensors='pt')

    # append the new user input tokens to the chat history
    bot_input_ids = torch.cat([chat_history_ids, new_user_input_ids], dim=-1) if step > 0 else new_user_input_ids

    # generated a response while limiting the total chat history to 1000 tokens, 
    chat_history_ids = model.generate(bot_input_ids, max_length=1000, pad_token_id=tokenizer.eos_token_id)

    # pretty print last ouput tokens from bot
    print("{}".format(tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)))

which is the code given in huggingface. The problem is when I try to generate something there is never a response. To fix this I have used this code I modified to work

# encode the new user input, add the eos_token and return a tensor in Pytorch
            new_user_input_ids = tokenizer.encode(genpromptfinal + tokenizer.eos_token, return_tensors='pt')
            # append the new user input tokens to the chat history
            step = 0
            bot_input_ids = torch.cat([new_user_input_ids],
                                      dim=-1) if step > 0 else new_user_input_ids
            chat_history_ids = model.generate(bot_input_ids, max_length=1000, pad_token_id=tokenizer.eos_token_id)
            bot_input_ids = torch.cat([chat_history_ids, new_user_input_ids],
                                      dim=-1) if step > 0 else new_user_input_ids
            # generated a response while limiting the total chat history to 1000 tokens,
            chat_history_ids = model.generate(bot_input_ids, max_length=1000, pad_token_id=tokenizer.eos_token_id)
            # pretty print last ouput tokens from bot
            "{}".format(tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True))

My concern is this may not retain past tokens when generating, and when I try to do that it gives me an error that chat_history_ids was never declared before bot_input_ids which needs chat_history_ids to include history.
How can I fix this?

[archmagos-dominus]

The code should look something like this instead

# Let's chat for 4 lines
for step in range(4):
    # encode the new user input, add the eos_token and return a tensor in Pytorch
    new_user_input_ids = tokenizer.encode(input(">> User:") + tokenizer.eos_token, return_tensors='pt')
    # print(new_user_input_ids)

    # append the new user input tokens to the chat history
    bot_input_ids = torch.cat([chat_history_ids, new_user_input_ids], dim=-1) if step > 0 else new_user_input_ids

    # generated a response while limiting the total chat history to 200 tokens,
    chat_history_ids = model.generate(
        bot_input_ids, max_length=200,
        pad_token_id=tokenizer.eos_token_id,
        no_repeat_ngram_size=1,
        do_sample=True,
        top_k=100,
        top_p=0.8,
        temperature=0.8
    )

    # pretty print last ouput tokens from bot
    print("{}".format(tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)))

Notice the fact that step is defined in the declaration of the for loop.