instmodel 0.5.0

A package that provides advanced coding tools for training and deploying AI

instmodel

instmodel is a Python package designed to simplify the creation of instruction-based neural network models using a Keras backend. With instmodel, you can quickly build, train, and export your models into a compact “instruction” format suitable for lightweight inference, serialization, or deployment.

---

Features

• High-Level Model Building: Easily construct neural networks using familiar Keras-like layers such as Dense, Attention, and more.
• Instruction Model Export: Convert your trained models into a JSON-based “instruction model” that precisely reflects the network architecture, weights, and activations.
• Validation & Debugging: Verify that the instruction model produces the same outputs as the original Keras model with built-in validation functions.
• Easy Deployment: Use the exported instruction model for lightweight or custom inference scenarios (e.g., embedded systems).

---

Installation

Install the latest release from PyPI:

pip install instmodel

---

Quick Example

Below is a simplified example (adapted from the test suite) showing how to:

1. Define a feed-forward network using Keras-like syntax.
2. Train it on dummy data.
3. Export it as an instruction model.
4. Validate it against the original Keras model outputs.

import numpy as np
from instmodel.model import (
    Dense,
    InputBuffer,
    ModelGraph,
    ff_model,
    validate_keras_model
)
from instmodel.instruction_model import validate_instruction_model

# 1. Define a simple feed-forward model (three Dense layers).
input_buffer = InputBuffer(4, name="simple_input")
hidden = Dense(8, activation="relu", name="hidden_relu_1")(input_buffer)
hidden = Dense(6, activation="relu", name="hidden_relu_2")(hidden)
output = Dense(1, activation="sigmoid", name="output_sigmoid")(hidden)

model_graph = ModelGraph(input_buffer, output)
model_graph.compile(optimizer="adam", loss="binary_crossentropy")

# 2. Train the model on dummy data.
x_data = np.random.random((10, 4))
y_data = np.random.randint(0, 2, size=(10, 1))
model_graph.fit(x_data, y_data, epochs=1, verbose=0)

# 3. Export the trained Keras model to an instruction model.
instruction_model = model_graph.create_instruction_model()

# 4. Validate the exported instruction model against the original Keras outputs.
keras_pred = model_graph.predict(x_data, verbose=0)
instruction_model["validation_data"] = {
    "inputs": x_data.tolist(),
    "expected_outputs": keras_pred.tolist(),
}

validate_instruction_model(instruction_model)  # Check instruction-model output
validate_keras_model(model_graph.get_keras(), instruction_model["validation_data"])  # Compare with Keras model

---

License

This project is licensed under the MIT License.