Colab vs Kaggle

Key Commands

# Mount Google Drive
from google.colab import drive
drive.mount('/content/drive')

# Check GPU
!nvidia-smi

# Install packages
!pip install -q transformers datasets

# Save to Drive
import shutil
shutil.copy('model.pkl', '/content/drive/MyDrive/')

Interactive Notebook

Quiz

Test your understanding of Google Colab -- 10 questions, 70% to pass.

PyTorch Core Concepts

Training Loop

for epoch in range(epochs):
    optimizer.zero_grad()       # 1. clear gradients
    output = model(X_train)     # 2. forward pass
    loss = criterion(output, y) # 3. compute loss
    loss.backward()             # 4. backpropagation
    optimizer.step()            # 5. update weights

Interactive Notebook

Quiz

Test your understanding of PyTorch Foundations -- 10 questions, 70% to pass.

Keras Sequential Model

import tensorflow as tf

model = tf.keras.Sequential([
    tf.keras.layers.Dense(64, activation='relu', input_shape=(30,)),
    tf.keras.layers.Dropout(0.3),
    tf.keras.layers.Dense(1, activation='sigmoid')
])
model.compile(optimizer='adam', loss='binary_crossentropy',
              metrics=['accuracy'])
model.fit(X_train, y_train, epochs=50, validation_split=0.2,
          callbacks=[tf.keras.callbacks.EarlyStopping(patience=5)])

PyTorch vs Keras

Interactive Notebook

Quiz

Test your understanding of TensorFlow / Keras -- 10 questions, 70% to pass.

ML Project .gitignore

# Python
.venv/
__pycache__/
*.py[cod]

# Data (use DVC/S3)
data/raw/
*.csv
*.parquet

# Models (use HuggingFace Hub/S3)
*.pkl
*.pt
*.h5

# Secrets
.env
*.pem

# ML outputs
mlruns/
logs/

DVC for Data Versioning

pip install dvc[s3]
dvc init
dvc add data/train.csv    # creates train.csv.dvc
git add data/train.csv.dvc .gitignore
git commit -m "track data with DVC"
dvc push                  # push to S3

Interactive Notebook

Quiz

Test your understanding of Git & GitHub for ML -- 10 questions, 70% to pass.

GPU Instance Types

Cost Management

• Use Spot instances for training (70-90% cheaper)
• Save checkpoints to S3 every N epochs
• Stop instances immediately after training
• Use SageMaker for auto-termination on completion

Interactive Notebook

Quiz

Test your understanding of AWS EC2 with GPU -- 10 questions, 70% to pass.

Detailed Comparison

Recommendation

Interactive Notebook

Quiz

Test your understanding of PyTorch vs TensorFlow -- 10 questions, 70% to pass.

Jupyter Best Practices

• Keep cells small and focused on one thing
• Run cells top-to-bottom — avoid hidden state bugs
• Restart and Run All before sharing
• Use nbconvert to export to HTML/PDF for reports
• Use %timeit and %memit for profiling

VS Code ML Extensions

Interactive Notebook

Quiz

Test your understanding of Jupyter & VS Code for ML -- 10 questions, 70% to pass.