File size: 4,405 Bytes
7cdf421 |
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 |
import json
import os.path
from torch.utils.data import Dataset
from tqdm import tqdm
import pandas as pd
import re
import random
import numpy as np
import torch
def load_alpaca(data_path, sample_data=False, sample_numer=1000, save_dir=''):
"""
sample and process the alpaca dataset in to the following format:
[
{
"image_name": "00000000000",
"output_modality": "text",
"conversation": [
{
"from": "human",
"value": "Give three tips for staying healthy.",
"input_modality": "text"
},
{
"from": "gpt",
"value": "1. Eat a balanced and nutritious diet: ...",
"caption": "",
"output_modality": "text"
}
]
},
...
]
"""
with open(data_path, 'r') as f:
data = json.load(f)
print('the total instance is {}'.format(len(data)))
if sample_data and sample_numer > 0:
data = random.sample(data, sample_numer)
res = []
for d in data:
_temp = dict()
_temp['image_name'] = '00000000000'
_temp['output_modality'] = 'text'
conversation = []
conversation.append(
{'from': 'human',
'value': d['instruction'] + d['input'],
'input_modality': 'text'}
)
conversation.append(
{'from': 'gpt',
'value': d['output'],
'caption': '',
'output_modality': 'text'}
)
_temp['conversation'] = conversation
res.append(_temp)
if not os.path.exists(save_dir):
os.makedirs(save_dir)
save_path = os.path.join(save_dir, os.path.basename(data_path))
with open(save_path, 'w', encoding='utf-8') as f:
json.dump(res, f, indent=4)
return res
def load_llava(data_path, sample_data=False, sample_numer=1000, save_dir=''):
"""
sample and process the llava instruction dataset into the following format:
[
{
"image_name": "00000000000.jpg",
"output_modality": "text",
"conversation": [
{
"from": "human",
"value": "Give three tips for staying healthy.",
"input_modality": "image"
},
{
"from": "gpt",
"value": "1. Eat a balanced and nutritious diet: ...",
"caption": "",
"output_modality": "text"
}
]
},
...
]
"""
with open(data_path, 'r') as f:
data = json.load(f)
print('the total instance is {}'.format(len(data)))
if sample_data and sample_numer > 0:
res = random.sample(data, sample_numer)
else:
res = data
# res = data
save_path = os.path.join(save_dir, os.path.basename(data_path))
for x in res:
i = 0
x['output_modality'] = 'text'
for j in x['conversation']:
if j['from'] == 'gpt':
j['caption'] = ''
j['output_modality'] = 'text'
elif j['from'] == 'human':
if i == 0:
j['input_modality'] = 'image'
i += 1
else:
j['input_modality'] = 'text'
with open(save_path, 'w', encoding='utf-8') as f:
json.dump(res, f, indent=4)
return res
def load_t2x(data_path):
with open(data_path, 'r', encoding='utf-8') as f:
data = json.load(f)
return data
if __name__ == '__main__':
save_dir = '../../data/IT_data/T+X-T_data'
res = []
# audios = load_t2x(os.path.join(save_dir, 'audio_t2x.json'))
# videos = load_t2x(os.path.join(save_dir, 'video_t2x.json'))
# images = load_t2x(os.path.join(save_dir, 'image_t2x.json'))
# sample_number = max(len(audios), len(videos), len(images))
#
# print(sample_number)
sample_number = 1000
print('Load aplaca dataset ...')
text = load_alpaca('../../data/IT_data/T+X-T_data/alpaca/alpaca.json', False, sample_number, save_dir)
res.extend(text)
print('Load llava dataset ...')
data = load_llava('../../data/IT_data/T+X-T_data/llava/llava.json', False, sample_number, save_dir)
|