# pip install git+https://github.com/LLaVA-VL/LLaVA-NeXT.git from llava.model.builder import load_pretrained_model from llava.mm_utils import get_model_name_from_path, process_images, tokenizer_image_token from llava.constants import IMAGE_TOKEN_INDEX, DEFAULT_IMAGE_TOKEN, DEFAULT_IM_START_TOKEN, DEFAULT_IM_END_TOKEN, IGNORE_INDEX from llava.conversation import conv_templates, SeparatorStyle from PIL import Image import requests import copy import torch import sys import warnings import os from decord import VideoReader, cpu import numpy as np import json import argparse from vbench2.utils import load_dimension_info from tqdm import tqdm warnings.filterwarnings("ignore") def load_video(video_path, max_frames_num, fps=1, force_sample=False): if max_frames_num == 0: return np.zeros((1, 336, 336, 3)) vr = VideoReader(video_path, ctx=cpu(0),num_threads=1) total_frame_num = len(vr) video_time = total_frame_num / vr.get_avg_fps() fps = round(vr.get_avg_fps()/fps) frame_idx = [i for i in range(0, len(vr), fps)] frame_time = [i/fps for i in frame_idx] if len(frame_idx) > max_frames_num or force_sample: sample_fps = max_frames_num uniform_sampled_frames = np.linspace(0, total_frame_num - 1, sample_fps, dtype=int) frame_idx = uniform_sampled_frames.tolist() frame_time = [i/vr.get_avg_fps() for i in frame_idx] frame_time = ",".join([f"{i:.2f}s" for i in frame_time]) spare_frames = vr.get_batch(frame_idx).asnumpy() return spare_frames,frame_time,video_time def LLaVA_Video(prompt_dict_ls, model, tokenizer, image_processor, device): final_score = 0 valid_num = 0 processed_json=[] for prompt_dict in tqdm(prompt_dict_ls): base_question = prompt_dict['auxiliary_info']['question'] question_num = len(base_question) video_paths = prompt_dict['video_list'] num_judge0 = prompt_dict['auxiliary_info']['judge'][0] num_judge1 = prompt_dict['auxiliary_info']['judge'][1] for video_path in video_paths: max_frames_num = 64 video,frame_time,video_time = load_video(video_path, max_frames_num, 1, force_sample=True) video = image_processor.preprocess(video, return_tensors="pt")["pixel_values"].cuda().bfloat16() video = [video] conv_template = "qwen_1_5" # Make sure you use correct chat template for different models time_instruciton = f"The video lasts for {video_time:.2f} seconds, and {len(video[0])} frames are uniformly sampled from it. These frames are located at {frame_time}. " score=0 valid=True if num_judge0: for i in range(question_num+1): if i==0: prefix = "Judging whether the video occurs only one creature, answer yes or no only." question = DEFAULT_IMAGE_TOKEN + f"{time_instruciton}{prefix}" else: prefix = "For the following description, judging whether the description contains in the video, answer yes or no only." question = DEFAULT_IMAGE_TOKEN + f"{time_instruciton}{prefix}\n{base_question[i-1]}" conv = copy.deepcopy(conv_templates[conv_template]) conv.append_message(conv.roles[0], question) conv.append_message(conv.roles[1], None) prompt_question = conv.get_prompt() input_ids = tokenizer_image_token(prompt_question, tokenizer, IMAGE_TOKEN_INDEX, return_tensors="pt").unsqueeze(0).to(device) cont = model.generate( input_ids, images=video, modalities= ["video"], do_sample=False, temperature=0, max_new_tokens=4096, ) text_outputs = tokenizer.batch_decode(cont, skip_special_tokens=True)[0].strip() if i==0: if 'yes' not in text_outputs.lower(): valid=False break else: if "yes" in text_outputs.lower(): score+=1 else: for i in range(question_num): prefix = "For the following description, judging whether the description contains in the video, answer yes or no only." question = DEFAULT_IMAGE_TOKEN + f"{time_instruciton}{prefix}\n{base_question[i]}" conv = copy.deepcopy(conv_templates[conv_template]) conv.append_message(conv.roles[0], question) conv.append_message(conv.roles[1], None) prompt_question = conv.get_prompt() input_ids = tokenizer_image_token(prompt_question, tokenizer, IMAGE_TOKEN_INDEX, return_tensors="pt").unsqueeze(0).to(device) cont = model.generate( input_ids, images=video, modalities= ["video"], do_sample=False, temperature=0, max_new_tokens=4096, ) text_outputs = tokenizer.batch_decode(cont, skip_special_tokens=True)[0].strip() if "yes" in text_outputs.lower(): score+=1 if valid: if num_judge1: if score==question_num: final_score += 1 sco=1 else: sco=0 else: final_score += score/question_num sco=score/question_num else: final_score += 1/question_num sco=1/question_num valid_num+=1 processed_json.append({'video_path': video_path, 'video_results': sco}) return final_score/(valid_num), processed_json def compute_composition(json_dir, device, submodules_dict, **kwargs): _, prompt_dict_ls = load_dimension_info(json_dir, dimension='composition', lang='en') model_name = "llava_qwen" device_map = "auto" try: pretrained = submodules_dict['llava'] llava_tokenizer, llava_model, image_processor, max_length = load_pretrained_model(pretrained, None, model_name, torch_dtype="bfloat16", device_map=device_map) # Add any other thing you want to pass in llava_model_args except: pretrained = "lmms-lab/LLaVA-Video-7B-Qwen2" llava_tokenizer, llava_model, image_processor, max_length = load_pretrained_model(pretrained, None, model_name, torch_dtype="bfloat16", device_map=device_map) # Add any other thing you want to pass in llava_model_args llava_model.eval() all_results, video_results = LLaVA_Video(prompt_dict_ls, llava_model, llava_tokenizer, image_processor, device) all_results = sum([d['video_results'] for d in video_results]) / len(video_results) return all_results, video_results