Skip to main content

llm model process for pytorch

Project description

kerastorch 是一个通用的pytorch模型训练模版工具,你可以像使用keras一样,使用pytorch,它具有如下优点:

  • 好看 (代码优雅,日志美丽,自带可视化)

  • 好用 (使用方便,支持 进度条、评估指标、early-stopping等常用功能,支持tensorboard,wandb回调函数等扩展功能)

  • 好改 (修改简单,核心代码模块化,仅约200行,并提供丰富的修改使用案例)

使用方法 🍊🍊

安装 kerastorch

pip install kerastorch

通过使用kerastorch,你不需要写自己的pytorch模型训练循环。你只要做这样两步就可以了。

(1) 创建你的模型结构net,然后把它和损失函数传入kerastorch.KerasModel构建一个model。

(2) 使用model的fit方法在你的训练数据和验证数据上进行训练,训练数据和验证数据需要封装成两个DataLoader.

核心使用代码就像下面这样:

import torch 
import kerastorch
import torchmetrics
model = kerastorch.KerasModel(net,
                              loss_fn = nn.BCEWithLogitsLoss(),
                              optimizer= torch.optim.Adam(net.parameters(),lr = 1e-4),
                              metrics_dict = {"acc":torchmetrics.Accuracy(task='binary')}
                             )
dfhistory=model.fit(train_data=dl_train, 
                    val_data=dl_val, 
                    epochs=20, 
                    patience=3, 
                    ckpt_path='checkpoint.pt',
                    monitor="val_acc",
                    mode="max",
                    plot=True,
                    
                   )

在jupyter notebook中执行训练代码,你将看到类似下面的训练可视化图像和训练日志进度条。

主要特性 🍉🍉

kerastorch 支持以下这些功能特性,稳定支持这些功能的起始版本以及这些功能借鉴或者依赖的库的来源见下表。

功能 稳定支持起始版本 依赖或借鉴库
✅ 训练进度条 3.0.0 依赖tqdm,借鉴keras
✅ 训练评估指标 3.0.0 借鉴pytorch_lightning
✅ notebook中训练自带可视化 3.8.0 借鉴fastai
✅ early stopping 3.0.0 借鉴keras
✅ gpu training 3.0.0 依赖accelerate
✅ multi-gpus training(ddp) 3.6.0 依赖accelerate
✅ fp16/bf16 training 3.6.0 依赖accelerate
✅ tensorboard callback 3.7.0 依赖tensorboard
✅ wandb callback 3.7.0 依赖wandb

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

kerastorch-1.0.3.tar.gz (43.1 kB view details)

Uploaded Source

Built Distribution

kerastorch-1.0.3-py3-none-any.whl (91.1 kB view details)

Uploaded Python 3

File details

Details for the file kerastorch-1.0.3.tar.gz.

File metadata

  • Download URL: kerastorch-1.0.3.tar.gz
  • Upload date:
  • Size: 43.1 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.2 CPython/3.8.16

File hashes

Hashes for kerastorch-1.0.3.tar.gz
Algorithm Hash digest
SHA256 005779bb620a595c0f3607738e77ca11485b128d1fe6dde08e0a1ceb88933cef
MD5 2f6c1db4c676334954cf9e82eb08eeea
BLAKE2b-256 e33f0cc24395ba6334111ed373267209eabe27bcc139d639b139c48cb5cc99d8

See more details on using hashes here.

File details

Details for the file kerastorch-1.0.3-py3-none-any.whl.

File metadata

  • Download URL: kerastorch-1.0.3-py3-none-any.whl
  • Upload date:
  • Size: 91.1 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.2 CPython/3.8.16

File hashes

Hashes for kerastorch-1.0.3-py3-none-any.whl
Algorithm Hash digest
SHA256 0c31766f88f77cc4084f372bc1e001d40cf66cf91a240a692ed6353eedb1ed8d
MD5 fef9f1b9f8fd40a7505a64f931cf089e
BLAKE2b-256 0f05b099d62a05132ef06cf30e73beec291d9425b34cbf0db9c5ca24d07766ba

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page