File size: 5,236 Bytes
63fba53 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 |
import os
import json
from PIL import Image
from tqdm import tqdm
from pydash import at
from mathruler.grader import extract_boxed_content
from ..question_formats import get_judgement_prompt,get_multiple_choice_prompt
from ..utils import save_json,extract,judge_multi_choice,judge_close_end_vqa,judge_judgement
from ..base_dataset import BaseDataset
class Radrestruct(BaseDataset):
def __init__(self,model,dataset_path,output_path):
self.model = model
self.output_path = output_path
self.dataset_path = dataset_path
self.samples = []
self.chunk_idx = int(os.environ.get("chunk_idx",0))
self.num_chunks = int(os.environ.get("num_chunks",1))
def load_data(self):
dataset_path = self.dataset_path
id_to_img_mapping_path = os.path.join(dataset_path,"id_to_img_mapping_frontal_reports.json")
reports_path = os.listdir(os.path.join(dataset_path,"test_qa_pairs"))
answer_options_path = os.path.join(dataset_path,"answer_options.json")
dataset = []
with open(id_to_img_mapping_path, "r") as f:
id_to_img_mapping_frontal_reports = json.load(f)
for report_file in reports_path:
for elem in id_to_img_mapping_frontal_reports[report_file.split(".")[0]]:
dataset.append((report_file,elem))
with open(answer_options_path, "r") as f:
answer_options = json.load(f)
# ['index', 'Figure_path', 'Caption', 'Question', 'Choice A', 'Choice B', 'Choice C', 'Choice D', 'Answer', 'split']
for idx,sample in tqdm(enumerate(dataset)):
if idx % self.num_chunks == self.chunk_idx:
samples = self.construct_messages(sample,answer_options)
self.samples.extend(samples)
return self.samples
def construct_messages(self,sample,answer_options):
report,img = sample
report_id = report.split(".")[0]
img_path = os.path.join(self.dataset_path,"imgs",f"{img}.png")
img = Image.open(img_path).convert("RGB")
sample = {}
samples = []
with open(os.path.join(self.dataset_path,"test_qa_pairs",report), "r") as f:
qa_pairs = json.load(f)
for qa_pair_idx,qa_pair in enumerate(qa_pairs):
question,answer,history,info = qa_pair
options = info["options"]
answer_idxs = at(answer_options,*options)
answer_type = info["answer_type"]
is_reasoning = True if os.environ.get("REASONING","False") == "True" else False
if answer_type == "single_choice":
prompt = get_judgement_prompt(question,is_reasoning)
else:
prompt = get_multiple_choice_prompt(question,options,is_reasoning)
messages = {"prompt":prompt,"image":img}
sample["messages"] = messages
sample["answer_type"] = answer_type
sample["question"] = question
sample["choices"] = options
sample["answer"] = answer[0]
samples.append(sample)
return samples
def cal_metrics(self,out_samples):
metrics = {
"total metrics" : {
"total":0,
"right":0
},
"single_choice" : {
"total" : 0,
"right" : 0
},
"multiple_choice" : {
"total" : 0,
"right" : 0
}
}
for i,out_sample in tqdm(enumerate(out_samples)):
response = out_sample["response"]
if extract_boxed_content(response)!= "None":
response = extract_boxed_content(response)
elif "<answer>" in response:
response = extract(response,"answer")
answer = out_sample["answer"]
question = out_sample["question"]
answer_type = out_sample["answer_type"]
choices = out_sample["choices"]
answer = answer.lower().strip()
response = response.lower().strip()
metrics["total metrics"]["total"] += 1
if answer_type == "single_choice":
metrics["single_choice"]["total"] += 1
correct = judge_judgement(answer,response)
out_samples[i]["correct"] = correct
if correct:
metrics["single_choice"]["right"] += 1
metrics["total metrics"]["right"] += 1
else:
metrics["multiple_choice"]["total"] += 1
correct = judge_multi_choice(choices,answer,response)
out_samples[i]["correct"] = correct
if correct:
metrics["multiple_choice"]["right"] += 1
metrics["total metrics"]["right"] += 1
metrics["total metrics"]["acc"] = metrics["total metrics"]["right"]/metrics["total metrics"]["total"]
metrics["single_choice"]["acc"] = metrics["single_choice"]["right"]/metrics["single_choice"]["total"]
metrics["multiple_choice"]["acc"] = metrics["multiple_choice"]["right"]/metrics["multiple_choice"]["total"]
return metrics,out_samples
|