Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
NVIDIA Alpamayo 1.5 Nano is an open 10B reasoning VLA model for autonomous vehicles with reinforcement-learning enhanced reasoning, navigation guidance, and visual question answering.
| Date | Stars |
|---|---|
| 2026-07-24 | 341 |
| 2026-07-25 | 342 |
| 2026-07-28 | 345 |
| 2026-07-30 | 345 |
| 2026-07-31 | 346 |
| 2026-08-03 | 346 |
| 2026-08-05 | 351 |
| 2026-08-06 | 352 |
Today
+1 stars today
This week
+7 stars this week
This month
— stars this month
Momentum
11.0
growth rate 2.03%/day
<div align="center"> # 🏔️ Alpamayo 1.5 ### Supercharging Autonomous Driving with Interactive, Steerable Reasoning [](https://huggingface.co/nvidia/Alpamayo-1.5-10B) [](./LICENSE) </div> ## Updates - [May 2026] SFT and RL post-training scripts are available in [Alpamayo Recipes](https://github.com/NVlabs/alpamayo-recipes): [Alpamayo 1.5 SFT](https://github.com/NVlabs/alpamayo-recipes/tree/main/recipes/alpamayo1_5_sft) and [Alpamayo 1.x RL post-training](https://github.com/NVlabs/alpamayo-recipes/tree/main/recipes/alpamayo1_x_rl). **📖 Please read the [HuggingFace Model Card](https://huggingface.co/nvidia/Alpamayo-1.5-10B) first!** The model card contains comprehensive details on model architecture, inputs/outputs, licensing, and tested hardware configurations. This GitHub README focuses on setup, usage, and frequently asked questions. ## Support 📣 **Usage questions and discussion about Alpamayo 1.5**: please join us on the [Alpamayo NV Developer Forum](https://forums.developer.nvidia.com/c/autonomous-vehicles/alpamayo/766). 🐛 **Code-level bugs, documentation issues, and feature requests**: file a [GitHub issue](../../issues/new/choose) using the appropriate template (Bug report, Documentation request, or Feature request). The relevant NVIDIA responder is auto-assigned via the `assignees:` field on the template. 🚨 **Security vulnerabilities**: please use [NVIDIA's Vulnerability Disclosure Program](https://app.intigriti.com/programs/nvidia/nvidiavdp/detail). Do not file security issues publicly here. ## Prerequisites - **NVIDIA GPU** with CUDA support - **CUDA Toolkit 12.x** with `nvcc` (required to compile `flash-attn` from source). If you don't have it, see [Troubleshooting](#flash-attention-issues) for a fallback using PyTorch's built-in SDPA. - **Python 3.12** ### Hardware requirements | Configuration | VRAM | | ------------------------------------------------------- | ------ | | Single-sample inference (`num_traj_samples=1`) | ~24 GB | | Multi-sample inference (`num_traj_samples=16`) | ~40 GB | | Multi-sample inference with CFG (`num_traj_samples=16`) | ~60 GB | Measured on an NVIDIA H100 80GB GPU. ## Getting Started ### 1. Install uv (if not already installed) ```bash curl -LsSf https://astral.sh/uv/install.sh | sh export PATH="$HOME/.local/bin:$PATH" ``` ### 2. Set up the environment ```bash uv venv a1_5_venv source a1_5_venv/bin/activate uv sync --active ``` > **Note:** If `uv sync` fails on `flash-attn`, see [Troubleshooting](#flash-attention-issues) below. ### 3. Authenticate with HuggingFace The model and dataset require access to gated resources. Request access here: - 🤗 [PhysicalAI-Autonomous-Vehicles Dataset](https://huggingface.co/datasets/nvidia/PhysicalAI-Autonomous-Vehicles) - 🤗 [Alpamayo-1.5-10B Model](https://huggingface.co/nvidia/Alpamayo-1.5-10B) Then authenticate: ```bash hf auth login ``` Get your token at: https://huggingface.co/settings/tokens > **Note:** The `physical_ai_av` package (auto-installed via dependencies) streams data from the HuggingFace dataset. You must have accepted the dataset access request above before running inference. ## Running Inference ### Test script NOTE: This script will download both some example data (relatively small) and the model weights (22 GB). The latter can be particularly slow depending on network bandwidth. For reference, it takes around 2.5 minutes on a 100 MB/s wired connection. ```bash python src/alpamayo1_5/test_inference.py ``` In case you would like to obtain more trajectories and reasoning traces, please feel free to increase the `num_traj_samples` argument in the script. ### Interactive notebooks We provide notebooks that demonstrate the different capabilities of Alpamayo 1.5 under `notebooks/`, including
Excerpt of 13,129 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:a746208ec6687f31, topic:robotics, topic:autonomous-driving, topic:self-driving-car