Update README.md
Browse files
README.md
CHANGED
|
@@ -23,13 +23,13 @@ Alpamayo 2 Edge is a vision-language-action model for navigation-conditioned dri
|
|
| 23 |
|
| 24 |
### License/Terms of Use:
|
| 25 |
|
| 26 |
-
**Model weights:** The
|
| 27 |
|
| 28 |
**Source code:** See the [code repository license](https://github.com/NVlabs/alpamayo2-edge/blob/ga-release/LICENSE). Source-code licensing does not establish the license for these model weights.
|
| 29 |
|
| 30 |
### Deployment Geography:
|
| 31 |
|
| 32 |
-
|
| 33 |
|
| 34 |
### Use Case:
|
| 35 |
|
|
@@ -39,7 +39,7 @@ Navigation-conditioned driving reasoning and trajectory prediction for autonomou
|
|
| 39 |
|
| 40 |
### Release Date:
|
| 41 |
|
| 42 |
-
**Hugging Face:**
|
| 43 |
|
| 44 |
## Model Architecture:
|
| 45 |
|
|
@@ -146,9 +146,8 @@ This checkpoint uses the Cosmos3-Edge vision-language backbone and the course-he
|
|
| 146 |
|
| 147 |
**Acceleration Engine:** PyTorch and Hugging Face Transformers.
|
| 148 |
|
| 149 |
-
**Hardware Requirements (GPU Architecture, Model):**
|
| 150 |
|
| 151 |
-
**Latency, throughput, and peak memory:** TBD.
|
| 152 |
|
| 153 |
**Sampling:** Reasoning generation is followed by ten Euler diffusion steps by default. `--num-traj-sets` controls independently generated reasoning continuations; `--num-traj-samples` controls trajectories per continuation.
|
| 154 |
|
|
|
|
| 23 |
|
| 24 |
### License/Terms of Use:
|
| 25 |
|
| 26 |
+
**Model weights:** The model weights are released under the [OpenMDW-1.1](https://openmdw.ai/license/1-1/) license.
|
| 27 |
|
| 28 |
**Source code:** See the [code repository license](https://github.com/NVlabs/alpamayo2-edge/blob/ga-release/LICENSE). Source-code licensing does not establish the license for these model weights.
|
| 29 |
|
| 30 |
### Deployment Geography:
|
| 31 |
|
| 32 |
+
Global
|
| 33 |
|
| 34 |
### Use Case:
|
| 35 |
|
|
|
|
| 39 |
|
| 40 |
### Release Date:
|
| 41 |
|
| 42 |
+
**Hugging Face:** 11/10/2026, at [nvidia/Alpamayo2-Edge](https://huggingface.co/nvidia/Alpamayo2-Edge).
|
| 43 |
|
| 44 |
## Model Architecture:
|
| 45 |
|
|
|
|
| 146 |
|
| 147 |
**Acceleration Engine:** PyTorch and Hugging Face Transformers.
|
| 148 |
|
| 149 |
+
**Hardware Requirements (GPU Architecture, Model):** 16GB minimum GPU memory, validated devices (Thor-U), and measured runtime configuration (6 trajectories).
|
| 150 |
|
|
|
|
| 151 |
|
| 152 |
**Sampling:** Reasoning generation is followed by ten Euler diffusion steps by default. `--num-traj-sets` controls independently generated reasoning continuations; `--num-traj-samples` controls trajectories per continuation.
|
| 153 |
|