sefixlines 0.3.0

A tool for quick ML task solving with model training optimization and a wide range of capabilities

🚀 Sefixlines Model Pipelines

 

 

🆕 UPDATE: Regression tasks
🆕 UPDATE: Text classification

✨ Features

• ⚡ Quick start without tons of code
• 🖼️ Image and text classification
• 🎯 Image semantic segmentation
• 💾 Automatic weights saving/loading
• 🔧 Easy customization (loss_fn, optimizer, scheduler, augmentation)

⚙️ Installation

pip install sefixlines

🎯 Get Started with Basic Example

For a quick start, use ready-made templates with configured pipelines:

from sefixlines import baseline

# Creates a ready-to-use notebook with an example for your task
baseline.create('raw')                         # Universal
baseline.create('image_classification')        # Image classification
baseline.create('text_classification')         # Text classification
baseline.create('image_regression')            # Image regression
baseline.create('text_regression')             # Text regression
baseline.create('image_semantic_segmentation') # Semantic segmentation

This command will create a sefixline.ipynb file in the current directory with a fully working example, including:

• 📊 Data loading and preparation
• 🤖 Model setup
• 🏋️ Training with visualization
• 📈 Results evaluation

⚡ This is the fastest way to get started! Just open the created notebook and adapt it to your data.

🚦 Minimal Manual Run

1. Prepare your data

from sefixlines import datasets

datasets.ImageClassificationDataset(paths, labels)                  # Image classification
datasets.TextClassificationDataset(texts, labels)                   # Text classification
datasets.ImageRegressionDataset(paths, labels)                      # Image regression
datasets.TextRegressionDataset(texts, labels)                       # Text regression
datasets.ImageSemanticSegmentationDataset(image_paths, mask_paths)  # Semantic segmentation

2. Choose a model (any model that returns logits).
3. Train

from sefixlines import models

# For classification
model_wrapper = models.Classifier(model, "MyModel")
model_wrapper.fit(train_set, valid_set, num_epochs=3)

# For regression
regressor = models.Regressor(model, "MyRegressor")
regressor.fit(train_set, valid_set, num_epochs=3)

# For semantic segmentation
segmenter = models.SemanticSegmenter(model, "MySemanticSegmenter")
segmenter.fit(train_set, valid_set, num_epochs=3)

License

MIT. See LICENSE file.