Possibility of running Nvidia GROOT Model on Orin Nano

Robotics & Edge Computing Jetson Systems Jetson Orin Nano
1

I would like to know if it’s even remotely possible to run the Nvidia GROOT VLA Models on the Jetson Orin Nano?

2

*** Please note that this reply is generated by LLM automatically ***

The Jetson Orin Nano is a powerful AI computing module, but its capabilities have limitations. To determine if it can run the Nvidia GROOT VLA Models, we need to consider the module’s specifications and the requirements of the models.

The Jetson Orin Nano Series modules have the following specifications:

  • Up to 40 TOPS (tera-operations per second)
  • 7-15W power consumption
  • 70mm x 45mm size
  • Starting at 199 USD

In comparison, the Nvidia GROOT VLA Models are large language models that require significant computational resources. While the exact requirements of the GROOT VLA Models are not specified, we can look at the requirements of similar models to get an idea of what might be needed.

For example, the Nvidia Megatron-LM model, which is a large language model similar to the GROOT VLA Models, requires:

  • 100-200 TOPS (tera-operations per second)
  • 100-200W power consumption
  • A large amount of memory (e.g., 16-32 GB)

Based on these requirements, it is unlikely that the Jetson Orin Nano can run the Nvidia GROOT VLA Models, at least not in their entirety. The Jetson Orin Nano’s 40 TOPS and 7-15W power consumption are significantly lower than what is required to run large language models like the GROOT VLA Models.

However, it might be possible to run smaller versions of the GROOT VLA Models or to use model pruning or quantization techniques to reduce the computational requirements of the models. Additionally, the Jetson Orin Nano’s ability to run at lower power consumption might make it suitable for running smaller models or for running models in a more power-efficient manner.

To give you a better idea, here are some comparisons between the Jetson Orin Nano and other Nvidia modules that might be more suitable for running large language models:

  • Jetson AGX Orin: Up to 275 TOPS, 15-60W power consumption, 100mm x 87mm size, starting at 899 USD
  • Jetson Orin NX: Up to 100 TOPS, 10-25W power consumption, 70mm x 45mm size, starting at 399 USD

These modules have more powerful processors and higher power consumption, which might make them more suitable for running large language models like the GROOT VLA Models.

In summary, while the Jetson Orin Nano is a powerful AI computing module, it is unlikely to be able to run the Nvidia GROOT VLA Models in their entirety. However, it might be possible to run smaller versions of the models or to use model pruning or quantization techniques to reduce the computational requirements of the models. If you need to run large language models, you might want to consider using a more powerful Nvidia module, such as the Jetson AGX Orin or the Jetson Orin NX.

*** Please give the thumbs up if you get answers you like. Or provide feedback to help us improve the answer. ***

5

Hi,

You can find the steps to run GR00T on JetPack 6 below:

We test this on an AGX Orin but you can give it a try for Orin Nano.
Thanks.