-
Notifications
You must be signed in to change notification settings - Fork 7
/
Copy pathmodel_vqa_mme_real_world.py
166 lines (133 loc) · 6.7 KB
/
model_vqa_mme_real_world.py
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
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
import argparse
import torch
import os
import json
from tqdm import tqdm
import shortuuid
from llava.constants import IMAGE_TOKEN_INDEX, DEFAULT_IMAGE_TOKEN, DEFAULT_IM_START_TOKEN, DEFAULT_IM_END_TOKEN
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 tokenizer_image_token, process_images, get_model_name_from_path
from torch.utils.data import Dataset, DataLoader
from PIL import Image
import math
def split_list(lst, n):
"""Split a list into n (roughly) equal-sized chunks"""
chunk_size = math.ceil(len(lst) / n) # integer division
return [lst[i:i+chunk_size] for i in range(0, len(lst), chunk_size)]
def get_chunk(lst, n, k):
chunks = split_list(lst, n)
return chunks[k]
# Custom dataset class
class CustomDataset(Dataset):
def __init__(self, questions, image_folder, tokenizer, image_processor, model_config, prompt):
self.questions = questions
self.image_folder = image_folder
self.tokenizer = tokenizer
self.image_processor = image_processor
self.model_config = model_config
self.prompt = prompt
def __getitem__(self, index):
line = self.questions[index]
choices = line['Answer choices']
image_file = line["Image"]
qs = line["Text"]
if self.model_config.mm_use_im_start_end:
qs = DEFAULT_IM_START_TOKEN + DEFAULT_IMAGE_TOKEN + DEFAULT_IM_END_TOKEN + '\n' + qs
else:
qs = DEFAULT_IMAGE_TOKEN + '\n' + qs
choice_prompt = ' The choices are listed below: \n'
for choice in choices:
choice_prompt += choice + "\n"
qs += choice_prompt + self.prompt + '\nThe best answer is:'
# print(qs)
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()
image = Image.open(os.path.join(self.image_folder, image_file)).convert('RGB')
# self.model_config.image_aspect_ratio = 'pad_and_divide'
# image_tensor = process_images([image], self.image_processor, self.model_config)
image_tensor = process_images([image], self.image_processor, self.model_config)[0]
input_ids = tokenizer_image_token(prompt, self.tokenizer, IMAGE_TOKEN_INDEX, return_tensors='pt')
return input_ids, image_tensor, image.size
def __len__(self):
return len(self.questions)
def collate_fn(batch):
input_ids, image_tensors, image_sizes = zip(*batch)
input_ids = torch.stack(input_ids, dim=0)
image_tensors = torch.stack(image_tensors, dim=0)
return input_ids, image_tensors, image_sizes
# DataLoader
def create_data_loader(questions, image_folder, tokenizer, image_processor, model_config, batch_size=1, num_workers=4, prompt=''):
assert batch_size == 1, "batch_size must be 1"
dataset = CustomDataset(questions, image_folder, tokenizer, image_processor, model_config, prompt)
data_loader = DataLoader(dataset, batch_size=batch_size, num_workers=num_workers, shuffle=False, collate_fn=collate_fn)
return data_loader
def eval_model(args):
# Model
disable_torch_init()
model_path = os.path.expanduser(args.model_path)
model_name = get_model_name_from_path(model_path)
tokenizer, model, image_processor, context_len = load_pretrained_model(model_path, args.model_base, model_name)
with open(args.question_file, 'r') as file:
questions = json.load(file)
questions = get_chunk(questions, args.num_chunks, args.chunk_idx)
answers_file = os.path.expanduser(args.answers_file)
os.makedirs(os.path.dirname(answers_file), exist_ok=True)
ans_file = open(answers_file, "w")
if 'plain' in model_name and 'finetune' not in model_name.lower() and 'mmtag' not in args.conv_mode:
args.conv_mode = args.conv_mode + '_mmtag'
print(f'It seems that this is a plain model, but it is not using a mmtag prompt, auto switching to {args.conv_mode}.')
data_loader = create_data_loader(questions, args.image_folder, tokenizer, image_processor, model.config, prompt=args.test_prompt)
index, cnt_images = 0, []
for (input_ids, image_tensor, image_sizes), line in tqdm(zip(data_loader, questions), total=len(questions)):
idx = line["Question_id"]
cur_prompt = line["Text"]
input_ids = input_ids.to(device='cuda', non_blocking=True)
cnt_images.append(image_tensor.shape[0])
with torch.inference_mode():
output_ids = model.generate(
input_ids,
images=image_tensor.to(dtype=torch.float16, device='cuda', non_blocking=True),
image_sizes=image_sizes,
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)
outputs = tokenizer.batch_decode(output_ids, skip_special_tokens=True)[0].strip()
index += 1
if index % 100 == 0:
print(f'Prompt: {cur_prompt}\n\n Output: {outputs}')
ans_id = shortuuid.uuid()
line['output'] = outputs
ans_file.write(json.dumps(line) + "\n")
ans_file.flush()
ans_file.close()
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--model-path", type=str, default="liuhaotian/llava-v1.5-13b")
parser.add_argument("--model-base", type=str, default=None)
parser.add_argument("--image-folder", type=str, default="MME-RealWorld/")
parser.add_argument("--question-file", type=str, default="MME-RealWorld.json")
parser.add_argument("--answers-file", type=str, default="./llava_15_13b_answer.jsonl")
parser.add_argument("--conv-mode", type=str, default="vicuna_v1")
parser.add_argument("--num-chunks", type=int, default=1)
parser.add_argument("--chunk-idx", type=int, default=0)
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=64)
parser.add_argument("--use-qlora", type=bool, default=False)
parser.add_argument("--qlora-path", type=str, default="")
parser.add_argument(
"--test-prompt",
type=str,
default="Select the best answer to the above multiple-choice question based on the image. Respond with only the letter (A, B, C, D, or E) of the correct option.",
)
args = parser.parse_args()
print(args)
eval_model(args)