llama-aviation-pte

Quantized Llama 3.2 1B Aviation model for iOS deployment

Model Details

• Base Model: Llama 3.2 1B Instruct
• Format: PTE (PyTorch ExecuTorch)
• Backend: XNNPACK (optimized for iOS)
• Quantization: 4-bit GPTQ
• Size: 1.07 GB
• Platform: iOS 16+

Files

• llama3_2_1b_aviation_spinquant.pte (1083.3 MB)
• tokenizer.json (8.7 MB)
• tokenizer_config.json (0.1 MB)

Usage

Download

wget https://huggingface.co/bluelarkaviation/llama-aviation-pte/resolve/main/model.pte

iOS Integration

1. Add model.pte to your Xcode project
2. Install ExecuTorch framework
3. Load the model:

import ExecuTorch

let modelPath = Bundle.main.path(forResource: "model", ofType: "pte")!
let module = try Module(filePath: modelPath)

Performance

• Optimized for: iPhone 12 and newer
• RAM Required: 2-3 GB
• Backend: XNNPACK (5-10x faster on ARM CPUs)

License

Same as base Llama 3.2 model