Openproblem比赛总结
kuro 11/14/2022 kaggle
# hyperopt调参
hyperopt调参会出现一个问题，就是best参数的返回值对于np.choice对单独返回一个索引，issue见此处 (opens new window)。
from hyperopt import fmin, tpe, hp, STATUS_OK, Trials, space_eval
print space_eval(tuner_space, best)
1
2
3
2
3
调参的时候，如果觉得参数传递和dict的写法很别扭，可以试一下globals()全局变量，例如：
globals().update({'HP_LEARNING_RATE': False})
加上字符串前缀过滤的时候也会比较容易找到
# hyperopt
更自定义化的训练过程 https://github.com/hyperopt/hyperopt/issues/694
# 数据集创建
upd = f'cnn1d-bestparam-sub-cite-predstd-mlp12-seed{my_seed}'
!mkdir -p {upd}
!kaggle datasets init -p {upd}
assert (len(upd) < 50 and re.findall(r'[^a-zA-Z0-9-]',upd) == []),f'upd name {upd} is not valid'
with open(f'{upd}/dataset-metadata.json','w') as f:
f.write('''
{
"title": "%s",
"id": "galegale05/%s",
"licenses": [
{
"name": "CC0-1.0"
}
]
}
''' % (upd,upd))
!cat {upd}/dataset-metadata.json
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
# kaggle token
!pip install kaggle
!mkdir .kaggle
!mkdir ~/.kaggle/
import json
token = {"username":"galegale05","key":""}
with open('kaggle.json', 'w') as file:
json.dump(token, file)
!cp kaggle.json ~/.kaggle/kaggle.json
!chmod 600 /root/.kaggle/kaggle.json
# !kaggle config set -n path -v ./input
!kaggle datasets list
1
2
3
4
5
6
7
8
9
10
11
2
3
4
5
6
7
8
9
10
11