Alpamayo Recipes

June 1, 2026 ยท View on GitHub

A collection of end-to-end Alpamayo recipes for multiple versions (v1, v1.5, and beyond), designed to help developers quickly build, adapt, and deploy Alpamayo-based applications. This repo brings together battle-tested workflows across the Alpamayo ecosystem, including post-training recipes (supervised fine-tuning and reinforcement learning), quantization recipes, etc. Whether you are experimenting locally or building a full production stack, this repository is intended as the primary starting point for developers to learn, customize, and extend Alpamayo for their own use cases.

Contributing

See CONTRIBUTING.md for the repository layout, recipe packaging conventions, and guidance on adding new recipes for released Alpamayo models.

Recipes

Each recipe folder contains its own README with installation and training instructions.

RecipeDescription
recipes/alpamayo1_sft/Alpamayo 1 supervised fine-tuning (HuggingFace Trainer + DeepSpeed)
recipes/alpamayo1_5_sft/Alpamayo 1.5 SFT (HuggingFace Trainer + DeepSpeed)
recipes/alpamayo1_x_rl/Alpamayo 1 and 1.5 RL post-training (Cosmos-RL / GRPO)

Utility Scripts

ScriptPurpose
scripts/curate_pai_samples.pyCurate a subset of PAI samples
scripts/convert_checkpoint.pyConvert between Alpamayo 1 and 1.5 checkpoints
scripts/convert_release_config_to_training.pyConvert a release checkpoint to training format
scripts/convert_cosmos_rl_checkpoint.pyConvert a Cosmos-RL checkpoint to HuggingFace format
scripts/download_pai.pyDownload the Physical AI AV dataset from HuggingFace