import torch import os import json import gc import csv from PIL import Image from datasets import load_dataset,load_from_disk from collections import defaultdict from tqdm import tqdm from ..utils import save_json,judge_multi_choice from ..base_dataset import BaseDataset from ..question_formats import get_multiple_choice_prompt class MedFrameQA(BaseDataset): def __init__(self,model,dataset_path,output_path): self.model = model self.output_path = output_path self.dataset_path = dataset_path if dataset_path else "SuhaoYu1020/MedFrameQA" self.samples = [] self.chunk_idx = int(os.environ.get("chunk_idx",0)) self.num_chunks = int(os.environ.get("num_chunks",1)) def run(self,samples,model,batch_size = 500): out_samples = [] with torch.no_grad(): messages_list = [] current_messages = [] current_samples = [] for sample in tqdm(samples): messages = sample["messages"] current_messages.append(messages) current_samples.append(sample) if len(current_messages) >= batch_size: messages_list.append([current_messages,current_samples]) current_messages = [] current_samples = [] if current_messages: messages_list.append([current_messages,current_samples]) for current_messages,current_samples in tqdm(messages_list): outputs = model.generate_outputs(current_messages) try: for sample,response in zip(current_samples,outputs): del sample["messages"] sample["response"] = response out_samples.append(sample) except Exception as e: from pdb import set_trace;set_trace() print(e) gc.collect() return out_samples def load_data(self): dataset = load_dataset(self.dataset_path)["test"] for idx,sample in tqdm(enumerate(dataset)): if idx % self.num_chunks == self.chunk_idx: sample = self.construct_messages(sample) self.samples.append(sample) return self.samples def construct_messages(self,sample): question = sample["question"] choices = sample["options"] answer = sample["correct_answer"] image_1 = sample["image_1"] image_2 = sample["image_2"] image_3 = sample["image_3"] image_4 = sample["image_4"] image_5 = sample["image_5"] choices = [f"{chr(65+i)}.{choices[i]}" for i in range(len(choices))] images = [image_1,image_2,image_3,image_4,image_5] images = [image for image in images if image is not None] is_reasoning = True if os.environ.get("REASONING","False") == "True" else False prompt = get_multiple_choice_prompt(question,choices,is_reasoning) messages = {"prompt":prompt,"images":images} sample["messages"] = messages sample["choices"] = choices sample["answer"] = answer del sample["options"] del sample["correct_answer"] del sample["image_1"] del sample["image_2"] del sample["image_3"] del sample["image_4"] del sample["image_5"] return sample def cal_metrics(self,out_samples): total = 0 right = 0 total_modality = defaultdict(int) right_modality = defaultdict(int) for i,sample in enumerate(out_samples): response = sample["response"] choices = sample["choices"] answer = sample["answer"] correct = judge_multi_choice(choices,answer,response) out_samples[i]["correct"] = correct if correct: right += 1 total += 1 metrics = {"total metrics":{"total":total,"right":right,"acc":right/total}} return metrics,out_samples