ML Deep Learning

Purpose

Design and train deep neural networks with PyTorch and TensorFlow. Build CNN, RNN, and transformer architectures. Optimize training with distributed strategies, mixed precision, and proper GPU utilization.

Agent Protocol

Trigger

Exact user phrases: "PyTorch", "TensorFlow", "Keras", "neural network", "CNN", "RNN", "LSTM", "transformer", "GAN", "autoencoder", "training loop", "backpropagation", "gradient descent", "CUDA", "distributed training", "mixed precision", "DDP", "FSDP", "AMP", "CUDA", "ResNet", "EfficientNet".

Input Context

Before activating, verify:

• Framework (PyTorch, TensorFlow, JAX)
• Problem type (image classification, NLP, time-series, generative)
• Dataset size (images, sequences, total samples)
• Hardware (single GPU, multi-GPU, TPU, CPU)
• GPU type and memory (A100, V100, RTX, etc.)
• Training constraints (time budget, memory budget)
• Prior experiments or baseline results