-
Notifications
You must be signed in to change notification settings - Fork 0
/
amber_llava.py
184 lines (152 loc) · 6.92 KB
/
amber_llava.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
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
import argparse
import torch
import os
import json
from tqdm import tqdm
import shortuuid
import sys
import os
sys.path.append(os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__)))))
sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
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, KeywordsStoppingCriteria
from PIL import Image
import base64
import requests
from io import BytesIO
import re
import math
import torch.distributed as dist
from utils import dist_util
from utils.logger import create_logger
from glob import glob
from transformers import set_seed
def image_parser(args):
out = args.image_file.split(args.sep)
return out
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
def eval_model(args):
dist_util.setup_dist(args)
device = dist_util.device()
if dist.get_rank() == 0:
os.makedirs(
args.log_path, exist_ok=True
) # Make results folder (holds all experiment subfolders)
model_string_name = args.model_path.split("/")[-1]
experiment_index = len(glob(f"{args.log_path}/{model_string_name}/*"))
experiment_dir = f"{args.log_path}" # Create an experiment folder
os.makedirs(experiment_dir, exist_ok=True)
logger = create_logger(experiment_dir)
logger.info(f"Experiment directory created at {experiment_dir}")
else:
logger = create_logger(None)
# logger.info(f"use_cd: {args.use_cd}, method: {args.use_avisc}, layer_gamma: {args.layer_gamma}, masking_scheme: {args.masking_scheme}, lamb: {args.lamb}")
logger.info(f"question_file : {args.question_file}")
# 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,device="cuda:0")
with open(args.question_file, "r", encoding="utf-8") as f:
query_list_data = json.load(f)
answers_file = os.path.expanduser(args.answers_file)
# os.makedirs(os.path.dirname(answers_file), exist_ok=True)
ans_file = open(answers_file, "a")
for line in tqdm(query_list_data):
idx = line["id"]
image_file = line["image"]
qs = line["query"]
cur_prompt = qs
if 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
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()
input_ids = tokenizer_image_token(prompt, tokenizer, IMAGE_TOKEN_INDEX, return_tensors='pt').unsqueeze(0).cuda(0)
image = Image.open(image_file)
image_tensor = image_processor.preprocess(image, return_tensors='pt')['pixel_values'][0]
stop_str = conv.sep if conv.sep_style != SeparatorStyle.TWO else conv.sep2
keywords = [stop_str]
stopping_criteria = KeywordsStoppingCriteria(keywords, tokenizer, input_ids)
with torch.inference_mode():
with torch.no_grad():
output_dict = model.generate(
input_ids,
images=image_tensor.unsqueeze(0).half().cuda(0),
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,
output_attentions=True,
return_dict_in_generate=True,
output_hidden_states=True,
output_scores = True,
stopping_criteria=[stopping_criteria],
use_deco = True,
alpha = args.alpha,
threshold_top_p=args.threshold_top_p,
threshold_top_k=args.threshold_top_k,
early_exit_layers=[i for i in range(args.start_layer, args.end_layer)],
return_dict=True
)
output_ids = output_dict.sequences
input_token_len = input_ids.shape[1]
outputs = tokenizer.batch_decode(
output_ids[:, input_token_len:], skip_special_tokens=True
)[0]
outputs = outputs.strip()
logger.info(f"[{image_file}]")
logger.info(f"prompt: {cur_prompt}")
logger.info(f"text: {outputs}")
res_dict = {"id": idx,"response": outputs}
ans_file.write(json.dumps(res_dict, ensure_ascii=False) + "\n")
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--model-path", type=str, default="./llava-v1.5-7b")
parser.add_argument("--model-base", type=str, default=None)
parser.add_argument("--question-file", type=str, default="./AMBER/data/query/query_generative.json")
parser.add_argument("--answers-file", type=str, default="")
parser.add_argument("--conv-mode", type=str, default="llava_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=1.0) # 可以考虑调大温度试试
parser.add_argument("--top_p", type=float, default=0.9)
parser.add_argument("--top_k", type=int, default=None)
parser.add_argument("--num_beams", type=int, default=1)
parser.add_argument("--max_new_tokens", type=int, default=1024)
parser.add_argument("--log_path", type=str, default="./Deco")
parser.add_argument("--batch_size", type=int, default=1)
parser.add_argument("--alpha", type=float, default=0.6)
parser.add_argument("--threshold_top_p", type=float, default=0.9)
parser.add_argument("--threshold_top_k", type=int, default=20)
parser.add_argument("--start_layer", type=int, default=20)
parser.add_argument("--end_layer", type=int, default=29)
parser.add_argument("--seed", type=int, default=42)
args = parser.parse_args()
set_seed(args.seed)
eval_model(args)