fastai

fastai is a high-level deep learning library built on top of PyTorch, designed to make modern deep learning applications easier and more efficient. It provides a simple API that allows users to quickly build, train, and evaluate complex models without needing to delve into the underlying details.

You can use fastai to quickly train models while using SwanLab for experiment tracking and visualization.

1. Import SwanLabCallback

from swanlab.integration.fastai import SwanLabCallback

SwanLabCallback is a logging class adapted for fastai.

SwanLabCallback can define parameters such as:

• project, experiment_name, description, and other parameters consistent with swanlab.init, used for initializing the SwanLab project.
• You can also create the project externally via swanlab.init, and the integration will log the experiment to the project you created externally.

2. Pass to Trainer

from fastai.vision.all import *
from swanlab.integration.fastai import SwanLabCallback

...

# Define the model
learn = vision_learner(...)

# Add SwanLabCallback
learn.fit_one_cycle(5, cbs=SwanLabCallback)

3. Example - Pet Classification

from fastai.vision.all import *
from swanlab.integration.fastai import SwanLabCallback

# Load data
path = untar_data(URLs.PETS)
dls = ImageDataLoaders.from_name_re(
    path, get_image_files(path / "images"), pat=r"([^/]+)_\d+.jpg$", item_tfms=Resize(224)
)

# Define the model
learn = vision_learner(dls, resnet34, metrics=error_rate)

# Add SwanLabCallback
learn.fit_one_cycle(
    5,
    cbs=SwanLabCallback(
        project="fastai-swanlab-integration-test",
        experiment_name="super-test",
        description="Test fastai integration with swanlab",
        logdir="./logs",
    ),
)