NVIDIA GPU Configuration

FileFlows Forge can utilize NVIDIA GPUs for hardware-accelerated AI and machine learning tasks (such as Whisper audio transcription) using CUDA.

Prerequisites

Before enabling NVIDIA GPU support, ensure your host system meets the following requirements:

1. NVIDIA Drivers: Installed on the host machine (nvidia-smi command should execute successfully).
2. NVIDIA Container Toolkit: Installed and configured on the host so Docker can pass GPU access to the container.

---

Docker Compose Setup

To enable NVIDIA GPU access, add the ENABLE_CUDA=1 environment variable and specify the deploy.resources.reservations.devices section in your docker-compose.yml:

services:
  fileflows-forge:
    image: revenz/fileflows-forge
    restart: unless-stopped
    ports:
      - 19201:19201
    environment:
      - TZ=Pacific/Auckland
      - ENABLE_CUDA=1
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    volumes:
      - ./data:/app/data

---

Environment Variables

Variable	Default	Description
ENABLE_CUDA	0	Set to 1 to enable CUDA GPU hardware acceleration inside Forge.

---

Verifying GPU Access

Once the container is running with NVIDIA GPU support enabled, you can verify that Forge detects your GPU:

1. Check the FileFlows Forge Logs during container startup for CUDA device initialization messages.
2. Run nvidia-smi inside the running container to test GPU visibility:

docker exec -it <container_name_or_id> nvidia-smi

If configured correctly, this command will output your host GPU details inside the container environment.