Skip to main content

A modular, scalable AI framework for PyPI with dynamic architecture plugins and benchmarking suite.

Project description

PYMBBO ⚡

PyPI version License: MIT Python Version

PYMBBO es un framework modular, intuitivo y de alto rendimiento en Python diseñado para facilitar la experimentación, entrenamiento, evaluación y despliegue de redes neuronales, ofreciendo un sistema Plug-and-Play de Arquitecturas y herramientas avanzadas de Benchmarking y Comparación de Modelos.


🔥 Características Principales

  1. Gestión de Hiperparámetros (Config / Hyperparameters): Objeto central con acceso tipo atributo/diccionario y guardado/carga en formato JSON.
  2. Ingesta y Preparación de Datos (load_dataset, Dataset, Batcher): Carga desde CSV, matrices NumPy, tensores y carpetas, con división de conjuntos (split) y flujo de transformaciones (transform).
  3. Construcción y Ensamblado de Redes (build_model): Soporta modo secuencial (add_layer), resúmenes visuales (summary()), congelamiento de capas (freeze_layers) y fine-tuning.
  4. Sistema Plug-and-Play de Arquitecturas (pymbbo/architectures/): Agrega nuevas arquitecturas creando una simple subcarpeta. El framework la descubre y registra automáticamente.
  5. Entrenamiento Avanzado (fit): Motor de entrenamiento con soporte para Callbacks (EarlyStopping, ModelCheckpoint, LRScheduler).
  6. Métricas Especializadas y Benchmarking (token_scaling_benchmark, compare_models): Pruebas automáticas de velocidad de tokens, rendimiento (tokens/seg), latencia, costo estimado y comparación simultánea lado a lado de modelos.
  7. Persistencia y Exportación (save, load_model, export): Exportación directa a .mbbo, ONNX y TorchScript.

⚡ Instalación Rápida

pip install pymbbo

O desde el código fuente:

git clone https://github.com/pymbbo/pymbbo.git
cd pymbbo
pip install -e .

💡 Ejemplo Rápido de Uso

import numpy as np
from pymbbo import Config, load_dataset, build_model, EarlyStopping, ModelCheckpoint

# 1. Configurar Hiperparámetros
config = Config(learning_rate=0.001, batch_size=32, epochs=10)

# 2. Cargar y Preparar Datos
X = np.random.randn(500, 10).astype(np.float32)
Y = np.random.randint(0, 2, (500, 1)).astype(np.float32)
dataset = load_dataset((X, Y))
train_ds, val_ds, test_ds = dataset.split(train=0.8, val=0.1, test=0.1)

# 3. Construir el Modelo (Usando arquitectura incorporada 'mlp')
model = build_model("mlp", input_dim=10, hidden_units=[64, 32], output_dim=1)
model.summary()

# 4. Compilar y Entrenar
model.compile(optimizer="adam", loss_function="bce", metrics=["accuracy"])

callbacks = [
    EarlyStopping(patience=3),
    ModelCheckpoint("best_model.mbbo")
]

model.fit(train_ds, validation_data=val_ds, epochs=10, callbacks=callbacks)

# 5. Inferencia y Evaluación
predicciones = model.predict(test_ds)
reporte = model.evaluate(test_ds)
print("Resultado Evaluación:", reporte)

# 6. Guardar y Exportar
model.save("modelo_final.mbbo")
model.export("modelo.onnx", format="onnx")

📖 Documentación Completa

Para acceder al manual completo de desarrollador, explicación detallada de cada módulo y la guía paso a paso para crear plugins de arquitectura, consulta DOCS.md.


📄 Licencia

Este proyecto está bajo la Licencia MIT. Consulta el archivo LICENSE para más detalles.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pymbbo-0.1.7.tar.gz (80.2 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

pymbbo-0.1.7-py3-none-any.whl (64.1 kB view details)

Uploaded Python 3

File details

Details for the file pymbbo-0.1.7.tar.gz.

File metadata

  • Download URL: pymbbo-0.1.7.tar.gz
  • Upload date:
  • Size: 80.2 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.14

File hashes

Hashes for pymbbo-0.1.7.tar.gz
Algorithm Hash digest
SHA256 a1822726efa3be2c727e3b9b3624d91c1c3e8e176e4828d89e743721dde9c8fd
MD5 382adcf614997c86ca32a5ef56c67a98
BLAKE2b-256 ed3e63bb2bc4fde553b5388d4f54fe2b9438ac404b51a58e009c2e7ee04e8214

See more details on using hashes here.

Provenance

The following attestation bundles were made for pymbbo-0.1.7.tar.gz:

Publisher: publish.yml on bueormnew/pymbbo

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file pymbbo-0.1.7-py3-none-any.whl.

File metadata

  • Download URL: pymbbo-0.1.7-py3-none-any.whl
  • Upload date:
  • Size: 64.1 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.14

File hashes

Hashes for pymbbo-0.1.7-py3-none-any.whl
Algorithm Hash digest
SHA256 dc3a50279c871fe214c9caf7dc891889ccb6b33a67036b56a4ff55afe3d59cf3
MD5 149e4bfd7ff025698580d5f51ba8056c
BLAKE2b-256 c31a3ae6d294c5244d84f993221532ea8734aa966b56b686db1b067341155155

See more details on using hashes here.

Provenance

The following attestation bundles were made for pymbbo-0.1.7-py3-none-any.whl:

Publisher: publish.yml on bueormnew/pymbbo

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page