ml-experiments 0.1.2

Monitor you model training anywhere.

ML Experiments

This package is under development.

Installation

pip install ml-experiments

# or (if you are not getting the latest version)
pip install git+https://github.com/abdalrhmanu/ml-experiments.git --upgrade

Notification

from ml_experiments.notify import notify_desktop, notify_email

@notify_desktop(title='Testing Completed!')
  def train():
    // .. some training code ..
    return {"loss": 5}

@notify_email(recipient_emails=['[e-mail]'], sender_email=['[e-mail]','[e-mail]'])
  def train():
    // .. some training code ..
    return {"loss": 5}

Dataset Splitting

from ml_experiments.data_prep import Dataset

dataset = Dataset(
  dataset_path=r'..\\dev\\dataset', 
  csv_path=r'..\\dev\\test.csv',
  output_directory=r'..\\dev\\out',
  output_format="dir")

# split (copy) to the defined output directory to every label
dataset.split_to_directory()

dataset2 = Dataset(
  dataset_path=r'..\\dev\\dataset', 
  csv_path=r'..\\dev\\test.csv',
  output_directory=None,
  output_format="df")

# splits to a dataframe without creating any new directory 
df, x_col, y_col = dataset2.split_to_df()

Running Tests

cd tests
python3 test_filename.py