8-2 Environment

1. All the following steps are performed inside the Docker container.

First, create a Docker container to serve as the execution environment.

(1) Use -v $(pwd):/workspace to mount the current directory. Since AI large models require a significant amount of storage space, it is recommended to use a hard drive with ample capacity.

sudo docker run -it --rm --runtime nvidia \

--network host \

--name AImodel \

-v $(pwd):/workspace \

dustynv/l4t-pytorch:r36.4.0

Install the required packages inside the Docker container.

cd /workspace/ pip install -r requirements.txt --index-url https://pypi.jetson-ai-lab.io/jp6/cu126

Create a new requirements.txt with the following content

torch==2.8.0

torchvision==0.23.0

transformers==4.48.2

accelerate==1.6.0

flash_attn==2.8.1

triton==3.1.0

# Florence-2-base

einops

timm

# Phi-4-multimodal-instruct

scipy

peft==0.13.2

backoff==2.2.1

# Janus

sentencepiece

attrdict