Keras

knowledge skill

Train, evaluate, or predict with a Keras neural network model. Produces a trained model file (.keras or .h5) or inference outputs given input data and model architecture.

Worked examples

• {"input": {"model_architecture": "Sequential([Dense(64, activation='relu'), Dense(1, activation='sigmoid')])", "training_data": [[0.1, 0.2], [0.3, 0.4], [0.5, 0.6]], "epochs": 10, "batch_size": 32}, "output": {"model_file": "model.keras", "training_metrics": {"loss": 0.12, "accuracy": 0.95}, "predictions": [0.87, 0.11, 0.92]}}
• {"input": {"model_architecture": "Sequential([Dense(128, activation='relu'), Dense(10, activation='softmax')])", "training_data": [[0.1, 0.2, 0.3], [0.4, 0.5, 0.6]], "epochs": 5, "batch_size": 2}, "output": {"model_file": "classifier.keras", "training_metrics": {"loss": 0.34, "accuracy": 0.88}, "predictions": [[0.01, 0.95, 0.02, 0.01, 0.0, 0.0, 0.0, 0.0, 0.0, 0.01]]}}

Input

• model_architecture: str
• training_data: list[list[float]]
• epochs: int
• batch_size: int

Output

• model_file: str
• training_metrics: dict[str, float]
• predictions: list[float]

Details

• Skill type: knowledge skill
• Safety level: safe_public_research
• Version: 1.0.0