inference_fine_tune.py

import torch

from tokenizers import Tokenizer


from pathlib import Path
from config import get_config, get_weights_file_path
from train import get_model

def get_tokenizer(config)->Tokenizer:
    tokenizers_path = Path(config['tokenizer_file'])
    if Path.exists(tokenizers_path):
        print("Loading tokenizer from ", tokenizers_path)
        tokenizer = Tokenizer.from_file(str(tokenizers_path))
        return tokenizer
    else:
        raise FileNotFoundError("Cant find tokenizer file : ",tokenizers_path)


config = get_config("./openweb.config.json")
device = "cuda" if torch.cuda.is_available() else "cpu"
tokenizer = get_tokenizer(config)
pad_token_id = tokenizer.token_to_id("<pad>")
eos_token_id = tokenizer.token_to_id("</s>")
user_token_id = tokenizer.token_to_id("<user>") 
ai_token_id = tokenizer.token_to_id("<ai>")

model  = get_model(config, tokenizer.get_vocab_size()).to(device)
model_path = get_weights_file_path(config,config['preload'])
model.eval()        
state = torch.load(model_path,map_location=torch.device('cpu'))
model.load_state_dict(state['model_state_dict'])

def generate_response(prompt: str):
    input_tokens = tokenizer.encode(prompt).ids
    input_tokens = [user_token_id] + input_tokens + [ai_token_id]

    if len(input_tokens) > config['seq_len']:
        yield gr.Textbox.update(value="Prompt too long.")
        return

    input_tokens = torch.tensor(input_tokens).unsqueeze(0).to(device)
    temperature = 0.7
    top_k = 50
    i = 0
    generated_text = ""

    while input_tokens.shape[1] < 2000:
        out = model.decode(input_tokens)
        logits = model.project(out[:, -1])
        logits = logits / temperature
        top_k_logits, top_k_indices = torch.topk(logits, top_k)
        probs = torch.softmax(top_k_logits, dim=-1)
        next_token = torch.multinomial(probs, num_samples=1)
        next_token = top_k_indices.gather(-1, next_token)

        word = tokenizer.decode([next_token.item()])
        generated_text += word

        yield gr.Textbox.update(value=generated_text)

        input_tokens = torch.cat([input_tokens, next_token], dim=1)
        if input_tokens.shape[1] > config['seq_len']:
            input_tokens = input_tokens[:, -config['seq_len']:]

        if next_token.item() == eos_token_id or i >= 1024:
            break
        i += 1