USAGE.md

Model Usage Instructions

Loading the Model

import torch
from pathlib import Path

# Load the main LAM model (14GB)
model_path = "model_weights/lam_model.pt"
model = torch.load(model_path, map_location='cuda' if torch.cuda.is_available() else 'cpu')

# Or load the smaller checkpoint
checkpoint = torch.load("checkpoints/best.pt")

Model Files

• model_weights/lam_model.pt - Main unified model (14GB)
• checkpoints/best.pt - Best training checkpoint (6MB)
• model_weights/sample_vision_weights.pt - Vision module weights
• model_weights/config.json - Model configuration
• model_weights/capabilities.json - Model capabilities definition

Requirements

pip install torch torchvision transformers pillow

Inference Example

# Coming soon - full inference pipeline