inference_demo.py

# Copyright (c) Meta Platforms, Inc. and affiliates.
# This software may be used and distributed according to the terms of the Llama 2 Community License Agreement.

# from accelerate import init_empty_weights, load_checkpoint_and_dispatch
# Expects to be executed in folder: https://github.com/facebookresearch/llama-recipes/tree/main/src/llama_recipes/inference 

import fire
import torch
import os
import sys
import time
import json 
from typing import List

from transformers import LlamaTokenizer, LlamaForCausalLM
from safety_utils import get_safety_checker
from model_utils import load_model, load_peft_model

BASE_PROMPT = """Below is an interaction between a human and an AI fluent in English and Amharic, providing reliable and informative answers.
The AI is supposed to answer test questions from the human with short responses saying just the answer and nothing else.
Human: {}
Assistant [Amharic] : """

def main(
    model_name: str="",
    peft_model: str=None,
    quantization: bool=False,
    max_new_tokens =400, #The maximum numbers of tokens to generate
    prompt_file: str=None,
    seed: int=42, #seed value for reproducibility
    do_sample: bool=True, #Whether or not to use sampling ; use greedy decoding otherwise.
    min_length: int=None, #The minimum length of the sequence to be generated, input prompt + min_new_tokens
    use_cache: bool=True,  #[optional] Whether or not the model should use the past last key/values attentions Whether or not the model should use the past last key/values attentions (if applicable to the model) to speed up decoding.
    top_p: float=1.0, # [optional] If set to float < 1, only the smallest set of most probable tokens with probabilities that add up to top_p or higher are kept for generation.
    temperature: float=1.0, # [optional] The value used to modulate the next token probabilities.
    top_k: int=1, # [optional] The number of highest probability vocabulary tokens to keep for top-k-filtering.
    repetition_penalty: float=1.0, #The parameter for repetition penalty. 1.0 means no penalty.
    length_penalty: int=1, #[optional] Exponential penalty to the length that is used with beam-based generation. 
    enable_azure_content_safety: bool=False, # Enable safety check with Azure content safety api
    enable_sensitive_topics: bool=False, # Enable check for sensitive topics using AuditNLG APIs
    enable_saleforce_content_safety: bool=False, # Enable safety check woth Saleforce safety flan t5
    **kwargs
):    
    
    print("***Note: model is not set up for chat use case, history is reset after each response.")
    print("***Ensure that you have replaced the default LLAMA2 tokenizer with the Amharic tokenizer")
    
    # Set the seeds for reproducibility
    torch.cuda.manual_seed(seed)
    torch.manual_seed(seed)
    
    MAIN_PATH = '/path/to/llama2'
    peft_model = '/path/to/checkpoint'
    model_name = MAIN_PATH

    model = load_model(model_name, quantization)

    tokenizer = LlamaTokenizer.from_pretrained(model_name)
    embedding_size = model.get_input_embeddings().weight.shape[0]

    if len(tokenizer) != embedding_size:
        print("resize the embedding size by the size of the tokenizer")
        model.resize_token_embeddings(len(tokenizer))

    if peft_model:
        model = load_peft_model(model, peft_model)

    model.eval()

    while True:

        
        user_query = input('Type question in Amharic or English: ')
        user_prompt = BASE_PROMPT.format(user_query)
        batch = tokenizer(user_prompt, return_tensors="pt")
        batch = {k: v.to("cuda") for k, v in batch.items()}
        start = time.perf_counter()
        with torch.no_grad():
            outputs = model.generate(
                **batch,
                max_new_tokens=max_new_tokens,
                do_sample=do_sample,
                top_p=top_p,
                temperature=temperature,
                min_length=min_length,
                use_cache=use_cache,
                top_k=top_k,
                repetition_penalty=repetition_penalty,
                length_penalty=length_penalty,
                **kwargs 
            )
        e2e_inference_time = (time.perf_counter()-start)*1000
        print(f"the inference time is {e2e_inference_time} ms")
        
        output_text = tokenizer.decode(outputs[0], skip_special_tokens=True)

        print("MODEL_OUTPUT: {}".format(output_text))
        #user_prompt += output_text

if __name__ == "__main__":
    fire.Fire(main)