Skip to content

Keras Training History - Example

This notebook demonstrates how to use FlowyML's enhanced Keras callback to automatically track and visualize training history.

Setup

import numpy as np
from tensorflow import keras
from tensorflow.keras import layers
from flowyml.integrations.keras import FlowymlKerasCallback

# Generate synthetic data
def generate_data(samples=1000):
    X = np.random.randn(samples, 20)
    y = (X[:, 0] + X[:, 1] > 0).astype(int)
    return X, y

X_train, y_train = generate_data(1000)
X_val, y_val = generate_data(200)

Create a Simple Model

model = keras.Sequential([
    layers.Dense(64, activation='relu', input_shape=(20,)),
    layers.Dropout(0.3),
    layers.Dense(32, activation='relu'),
    layers.Dropout(0.3),
    layers.Dense(1, activation='sigmoid')
])

model.compile(
    optimizer='adam',
    loss='binary_crossentropy',
    metrics=['accuracy']
)

Train with FlowyML Callback

# Create the callback - that's all you need!
flowyml_callback = FlowymlKerasCallback(
    experiment_name="keras-demo",
    project="my-project",
    run_name="binary-classifier-v1",
    auto_log_history=True  # Automatically tracks training history
)

# Train the model
history = model.fit(
    X_train, y_train,
    validation_data=(X_val, y_val),
    epochs=50,
    batch_size=32,
    callbacks=[flowyml_callback],  # Add the callback
    verbose=1
)

What Happens Automatically

The callback automatically: 1. ✅ Tracks all metrics per epoch (loss, accuracy, val_loss, val_accuracy) 2. ✅ Saves model with complete training history attached 3. ✅ Logs to experiment for comparison 4. ✅ Captures model architecture as artifact

View in UI

  1. Go to http://localhost:3000/projects/my-project
  2. Navigate to Structure tab
  3. Find your run and expand the tree
  4. Click on the model-binary-classifier-v1 artifact
  5. See beautiful training history graphs!

Advanced: Custom Metrics

# Add custom metrics
import tensorflow.keras.backend as K

def f1_score(y_true, y_pred):
    true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1)))
    possible_positives = K.sum(K.round(K.clip(y_true, 0, 1)))
    predicted_positives = K.sum(K.round(K.clip(y_pred, 0, 1)))
    precision = true_positives / (predicted_positives + K.epsilon())
    recall = true_positives / (possible_positives + K.epsilon())
    return 2*((precision*recall)/(precision+recall+K.epsilon()))

model.compile(
    optimizer='adam',
    loss='binary_crossentropy',
    metrics=['accuracy', f1_score]  # Custom metric
)

flowyml_callback = FlowymlKerasCallback(
    experiment_name="keras-demo",
    project="my-project",
    auto_log_history=True
)

history = model.fit(
    X_train, y_train,
    validation_data=(X_val, y_val),
    epochs=50,
    callbacks=[flowyml_callback]
)

# The callback automatically captures your custom 'f1_score' metric too!

Disable Auto-Logging (Optional)

If you need manual control:

callback = FlowymlKerasCallback(
    experiment_name="keras-demo",
    project="my-project",
    auto_log_history=False  # Disable automatic history tracking
)

What Gets Stored

The training_history field in the artifact metadata looks like:

{
  "epochs": [1, 2, 3, ..., 50],
  "train_loss": [0.693, 0.612, 0.580, ...],
  "train_accuracy": [0.501, 0.625, 0.701, ...],
  "val_loss": [0.701, 0.625, 0.595, ...],
  "val_accuracy": [0.490, 0.615, 0.690, ...],
  "f1_score": [0.48, 0.61, 0.68, ...]  // Custom metrics included!
}

This format is automatically recognized by the FlowyML UI and rendered as interactive charts.

Next Steps

  • Try with your own datasets and models
  • Compare multiple runs in the experiments view
  • Use different optimizers and see the impact on training curves
  • Share your training history visualizations with your team!