Originally published at: Maximize Robotics Performance by Post-Training NVIDIA Cosmos Reason | NVIDIA Technical Blog
First unveiled at NVIDIA GTC 2025, NVIDIA Cosmos Reason is an open and fully customizable reasoning vision language model (VLM) for physical AI and robotics. The VLM enables robots and vision AI agents to reason using prior knowledge, physics understanding, and common sense to understand and act in the real world. Given a video and…
if i want to run in docker environment, which base image should i use to run inference on the cosmos reasoning 1 model? I am thinking of development and deployment to my jetson agx orin. It has jetpack 6.2.1 installed, cuda 12.6, L4T 36.4.4
I’m new to development with nvidia models and on nvidia hardware, so I’m not sure how I should go about inferencing the cosmos reasoning 1 model. Are the inference scripts in the github repo are meant to be run on x86 architecture? I got the following error when i tried running the inference_sample.py in my nvidia jetson agx orin.
➜ cosmos-reason1 git:(main) ./scripts/inference_sample.py
error: Distribution `torchcodec==0.6.0 @ registry+https://pypi.org/simple` can't be installed because it doesn't have a source distribution or wheel for the current platform
How can I run inference on the cosmos reason 1 model? The article mentioned
Cosmos Reason is optimized to perform best on NVIDIA GPUs. To run the models, developers can set up a Docker environment or run it in their environment.
But how do I set up a docker environment to run the model for inference (what image should i use)? I would really appreciate more elaboration or a guide on this.
Hello @zhiwei.thean,
I suggest posting your issue to the Jetson forums, as this category is not monitored by the Jetson engineers.