llava_util.py

# This file is modified from https://github.com/haotian-liu/LLaVA/

import argparse
import torch

from transformers import TextStreamer

from llava.constants import (
    IMAGE_TOKEN_INDEX,
    DEFAULT_IMAGE_TOKEN,
    DEFAULT_IM_START_TOKEN,
    DEFAULT_IM_END_TOKEN,
    IMAGE_PLACEHOLDER,
)
from llava.conversation import conv_templates, SeparatorStyle
from llava.model.builder import load_pretrained_model
from llava.utils import disable_torch_init
from llava.mm_utils import (
    process_images,
    tokenizer_image_token,
    get_model_name_from_path,
    KeywordsStoppingCriteria,
)

import requests
from PIL import Image
from io import BytesIO
import re


def run_llava(model_path, conv_mode, query, images, sep=",", temperature=0.2, top_p=None, num_beams=1, max_new_tokens=512, model_base=None, device="cuda:0"):
    args = argparse.Namespace(
        model_path=model_path,
        model_base=model_base,
        temperature=temperature,
        top_p=top_p,
        num_beams=num_beams,
        max_new_tokens=max_new_tokens,
        conv_mode=conv_mode,
        query=query,
        images=images,
        sep=sep,
        device=device
    )

    return eval_model(args)

def load_image(image_file):
    if image_file.startswith("http") or image_file.startswith("https"):
        response = requests.get(image_file)
        image = Image.open(BytesIO(response.content)).convert("RGB")
    else:
        image = Image.open(image_file).convert("RGB")
    return image

def load_images(image_files):
    out = []
    for image_file in image_files:
        image = load_image(image_file)
        out.append(image)
    return out

llavamodel = None

def eval_model(args):
    # Model
    disable_torch_init()
    global llavamodel

    model_name = get_model_name_from_path(args.model_path)
    if llavamodel is None:
        llavamodel = load_pretrained_model(
            args.model_path, args.model_base, model_name, device=args.device
        )
    tokenizer, model, image_processor, context_len = llavamodel 

    qs = args.query
    image_token_se = DEFAULT_IM_START_TOKEN + DEFAULT_IMAGE_TOKEN + DEFAULT_IM_END_TOKEN
    if IMAGE_PLACEHOLDER in qs:
        if model.config.mm_use_im_start_end:
            qs = re.sub(IMAGE_PLACEHOLDER, image_token_se, qs)
        else:
            qs = re.sub(IMAGE_PLACEHOLDER, DEFAULT_IMAGE_TOKEN, qs)
    else:
        if DEFAULT_IMAGE_TOKEN not in qs:
            if model.config.mm_use_im_start_end:
                qs = image_token_se + "\n" + qs
            else:
                qs = DEFAULT_IMAGE_TOKEN + "\n" + qs

    if "llama-2" in model_name.lower():
        conv_mode = "llava_llama_2"
    elif "v1" in model_name.lower():
        conv_mode = "llava_v1"
    elif "mpt" in model_name.lower():
        conv_mode = "mpt"
    else:
        conv_mode = "llava_v0"

    if args.conv_mode is not None and conv_mode != args.conv_mode:
        pass
        #print(
        #    "[WARNING] the auto inferred conversation mode is {}, while `--conv-mode` is {}, using {}".format(
        #        conv_mode, args.conv_mode, args.conv_mode
        #    )
        #)
    else:
        args.conv_mode = conv_mode

    conv = conv_templates[args.conv_mode].copy()
    conv.append_message(conv.roles[0], qs)
    conv.append_message(conv.roles[1], None)
    prompt = conv.get_prompt()

    images_tensor = process_images(
        args.images,
        image_processor,
        model.config
    ).to(args.device, dtype=torch.float16)

    input_ids = (
        tokenizer_image_token(prompt, tokenizer, IMAGE_TOKEN_INDEX, return_tensors="pt")
        .unsqueeze(0)
        .to(args.device)
    )

    stop_str = conv.sep if conv.sep_style != SeparatorStyle.TWO else conv.sep2
    keywords = [stop_str]
    stopping_criteria = KeywordsStoppingCriteria(keywords, tokenizer, input_ids)
    streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)

    with torch.inference_mode():
        output_ids = model.generate(
            input_ids,
            images=images_tensor,
            do_sample=True if args.temperature > 0 else False,
            temperature=args.temperature,
            #top_p=args.top_p,
            #num_beams=args.num_beams,
            max_new_tokens=args.max_new_tokens,
            use_cache=True,
            #stopping_criteria=[stopping_criteria],
            streamer=streamer,
            image_sizes=[image.size for image in args.images]
        )

    outputs = tokenizer.batch_decode(output_ids, skip_special_tokens=True)[0].strip()

    return outputs


if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    parser.add_argument("--model-path", type=str, default="facebook/opt-350m")
    parser.add_argument("--model-base", type=str, default=None)
    parser.add_argument("--image-file", type=str, required=True)
    parser.add_argument("--query", type=str, required=True)
    parser.add_argument("--conv-mode", type=str, default=None)
    parser.add_argument("--sep", type=str, default=",")
    parser.add_argument("--temperature", type=float, default=0.2)
    parser.add_argument("--top_p", type=float, default=None)
    parser.add_argument("--num_beams", type=int, default=1)
    parser.add_argument("--max_new_tokens", type=int, default=512)
    args = parser.parse_args()

    eval_model(args)