Wttch's train helper
Project description
wttch 的 AI 训练工具包
一、消息通知
1.1 钉钉 webhook 通知
from wttch.train.notification import DingtalkNotification
# 钉钉机器人 webhook 的链接
webhook_url = ''
# 消息签名的密钥
secret = ''
notification = DingtalkNotification(webhook_url, secret)
# 发送文本通知
notification.send_text("")
# 发送markdown
notification.send_markdown('')
1.2 企业微信 webhook 通知
from wttch.train.notification import WechatNotification
# 企业微信机器人 webhook 的链接
webhook_url = ''
notification = WechatNotification(webhook_url)
# 发送文本通知
notification.send_text("")
# 发送markdown
notification.send_markdown('')
二、训练工具包
2.1 缓存工具
不需要修改太多代码就可以帮助你缓存数据到指定的缓存文件去。
(1). 添加 cache_wrapper; (2). 正常调用你的函数。
from wttch.train.utils import cache_wrapper
# 缓存的文件名字前缀,函数的参数会被添加到该名字后面
prefix = 'dataset'
# 缓存的文件夹位置
save_path = './dataset_cache'
@cache_wrapper(prefix=prefix, save_path=save_path)
def you_load_dataset_function():
return {'a': 1, 'b': 2}
you_load_dataset_function()
2.2 计时器
from wttch.train.utils import StopWatch
stopwatch = StopWatch()
stopwatch.start("job 1")
# 费时操作
stopwatch.stop()
stopwatch.start("job 2")
# 费时操作
stopwatch.stop()
# 格式化打印
stopwatch.display()
2.3 进度条
简单包装了
tqdm
工具。
from wttch.train.utils.progress import Progress
with Progress(total=1000) as progress:
for i in range(1000):
# 在这里训练
# 进度条末尾显示训练结果
progress.train_result(loss=0.01, acc=0.02)
三、torch 工具包
3.1 方便设备获取
(1). 将使用的设备写入 thread local; (2). 从 thread local 中获取设备数据; (3). 训练。
from wttch.train.torch.utils import try_gpu, get_device_local, set_device_local
# 尝试获取 gpu 并写入 thread local
set_device_local(try_gpu(device_no=0))
# 从 thread local 读取设备
device = get_device_local()
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
wttch-train-helper-0.0.14.tar.gz
(10.8 kB
view details)
Built Distribution
File details
Details for the file wttch-train-helper-0.0.14.tar.gz
.
File metadata
- Download URL: wttch-train-helper-0.0.14.tar.gz
- Upload date:
- Size: 10.8 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.2 CPython/3.9.6
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 04d38cf883b31f5914eeda922f2ac0ca85a5b211952bacc8a2e06f562e957d8b |
|
MD5 | 27954aab5b5d007b7a71858b4f890226 |
|
BLAKE2b-256 | 1b7c7ed773042eb769adf7e7b74f167b5be5a4f8da3fce6dc7f2806b412ed471 |
File details
Details for the file wttch_train_helper-0.0.14-py3-none-any.whl
.
File metadata
- Download URL: wttch_train_helper-0.0.14-py3-none-any.whl
- Upload date:
- Size: 11.6 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.2 CPython/3.9.6
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 999f75e0a624c2f27b8f11eafe901dadfffb5ef96241a32de03ed204cb5fdeb4 |
|
MD5 | 2eb5cee1d10a830a1011a9ef570dc309 |
|
BLAKE2b-256 | b88779afc078cf24dbfd2864f3db7720d71a73faf7fc7402bb6545a6855e4f1f |