Skip to content

stakeordie/emp-mgpu

BranchesTags

Repository files navigation

Docker Setup Guide

This guide explains how to set up and run the ComfyUI environment both locally and on a remote server, with support for both GPU and mock (CPU) modes.

Prerequisites

  • Docker and Docker Compose installed
  • AWS credentials (for S3 sync features)
  • HuggingFace token
  • OpenAI API key (optional)
  • SSH key for NGINX repository access

Local Development Setup

1. Environment Configuration

  1. Copy the sample environment file:

    cp .env.sample .env.local

  2. Configure your environment variables in .env.local:

    • HF_TOKEN: Your HuggingFace token
    • OPENAI_API_KEY: Your OpenAI API key (if needed)
    • AWS_ACCESS_KEY_ID: Your AWS access key
    • AWS_SECRET_ACCESS_KEY_ENCODED: encoded with SLASH replacing /: hd/d/df -> hd_SLASH_d_SLASH_df
    • AWS_DEFAULT_REGION: "us-east-1"
    • SERVER_CREDS: HTTP Basic Auth for NGINX. Provide only the password.
    • TEST_GPUS: Number of GPUs to mock in test mode
    • AWS_TEST_MODE: Set to true for test mode

  3. Set up SSH key parts (for repository access):

    # Generate the CUT_PK parts from your SSH key
openssl base64 -in ~/.ssh/id_ed25519 | tr -d '\n' | cut -c-250 > cut1
openssl base64 -in ~/.ssh/id_ed25519 | tr -d '\n' | cut -c251-500 > cut2
openssl base64 -in ~/.ssh/id_ed25519 | tr -d '\n' | cut -c501- > cut3

    Add these values to your .env.local as CUT_PK_1, CUT_PK_2, and CUT_PK_3.

2. Building and Running

2.1. Build Types

Test and Mock GPU setups are defined in the .env.local file:

  • AWS_TEST_MODE: Set to true for test mode, false or unset for production
  • TEST_GPUS: Number of GPUs to mock in test mode (0 or unset for script to determine GPUs

AWS_TEST_MODE

Bucket Sync: When AWS_TEST_MODE is set to true will use the AWS bucket emprops-share-test for model and config sync. This is useful for testing and debugging in a controlled environment. The production bucket is 100 + GB so it will take a while. The test bucket only has the defualt ComryUI model SD1.5.

Config_nodes_test will also be used which limits the to EmProps Nodes.

TEST_GPUS

Mock GPU: When TEST_GPUS is set to a number it will mock that many GPUs in test mode, this results in MOCK_GPU being set to 1 or true and the number of TEST_GPUS being set will run all with the --cpu flag in the comfyui service start command. This is useful for testing and debugging in a controlled environment. If you don't set TEST_GPUS the system will still check for GPU support and if it can't find any it set the number of GPUs to 1 but will also set MOCK_GPU=1.

2.2. Docker Compose

Run cd docker to enter the Docker Directory.

Run docker compose build comfyui to build the Docker images. Run docker compose up comfyui to start the container. or Run docker compose up --build comfyui to build and start the container.

Run docker compose down to stop and remove the container.

Run docker exec -it comfy /bin/bash to enter the container.

Run docker push emprops/docker:latest to push the latest image to Docker Hub.

mgpu CLI usage

Commands:

  • start [gpu_id] Start ComfyUI service for GPU
  • stop [gpu_id] Stop ComfyUI service for GPU
  • restart [gpu_id] Restart ComfyUI service for GPU
  • status [gpu_id] Show status of ComfyUI service for GPU
  • logs [gpu_id] Show logs for GPU
  • setup [gpu_id] Setup ComfyUI for GPU
  • count Show number of available GPUs

Logs

mgpu uses multitail for consolidated logging.

  • "b" to scroll | up down pageup pagedown
  • "/" to search
  • "q" to quite

logs accepts multiple gpu_ids. so mgpu logs 1 3 4 will work.\5

MODELS FOLDER

#!/bin/bash

cd /workspace/shared/models/checkpoints && wget https://huggingface.co/Lightricks/LTX-Video/resolve/main/ltxv-13b-0.9.7-dev-fp8.safetensors

cd /workspace/shared/models/clip && wget https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/text_encoders/clip_g_hidream.safetensors https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/text_encoders/clip_l_hidream.safetensors https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/text_encoders/llama_3.1_8b_instruct_fp8_scaled.safetensors https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn_scaled.safetensors

cd /workspace/shared/models/diffusion_models && wget https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors

cd /workspace/shared/models/upscale_models && wget https://huggingface.co/Lightricks/LTX-Video/resolve/main/ltxv-spatial-upscaler-0.9.7.safetensors

cd /workspace/shared/models/vae && wget https://huggingface.co/lovis93/testllm/resolve/ed9cf1af7465cebca4649157f118e331cf2a084f/ae.safetensors

mkdir -p /workspace/shared/models/text_encoder/PixArt-XL-2-1024-MS/text_encoder && cd /workspace/shared/models/text_encoder/PixArt-XL-2-1024-MS/text_encoder && wget https://huggingface.co/PixArt-alpha/PixArt-XL-2-1024-MS/resolve/main/text_encoder/config.json https://huggingface.co/PixArt-alpha/PixArt-XL-2-1024-MS/resolve/main/text_encoder/model-00001-of-00002.safetensors https://huggingface.co/PixArt-alpha/PixArt-XL-2-1024-MS/resolve/main/text_encoder/model-00002-of-00002.safetensors https://huggingface.co/PixArt-alpha/PixArt-XL-2-1024-MS/resolve/main/text_encoder/model.safetensors.index.json

checkpoints wget https://huggingface.co/Lightricks/LTX-Video/resolve/main/ltxv-13b-0.9.7-dev-fp8.safetensors

clip wget https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/text_encoders/clip_g_hidream.safetensors https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/text_encoders/clip_l_hidream.safetensors https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/text_encoders/llama_3.1_8b_instruct_fp8_scaled.safetensors https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn_scaled.safetensors

diffusion_models wget https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors

text_encoder wget https://huggingface.co/PixArt-alpha/PixArt-XL-2-1024-MS/resolve/main/text_encoder/config.json https://huggingface.co/PixArt-alpha/PixArt-XL-2-1024-MS/resolve/main/text_encoder/model-00001-of-00002.safetensors https://huggingface.co/PixArt-alpha/PixArt-XL-2-1024-MS/resolve/main/text_encoder/model-00002-of-00002.safetensors https://huggingface.co/PixArt-alpha/PixArt-XL-2-1024-MS/resolve/main/text_encoder/model.safetensors.index.json

upscale_models wget https://huggingface.co/Lightricks/LTX-Video/resolve/main/ltxv-spatial-upscaler-0.9.7.safetensors

vae wget https://huggingface.co/lovis93/testllm/resolve/ed9cf1af7465cebca4649157f118e331cf2a084f/ae.safetensors

── animatediff_models
├── checkpoints
│   ├── epiCPhotoGasm.safetensors -> /workspace/shared/sd_models/Stable-diffusion/epiCPhotoGasm.safetensors
│   ├── ltxv-13b-0.9.7-dev-fp8.safetensors @ =
│   ├── sd_xl_base_1.0_0.9vae.safetensors -> /workspace/shared/sd_models/Stable-diffusion/sd_xl_base_1.0_0.9vae.safetensors
│   ├── sd_xl_refiner_1.0_0.9vae.safetensors -> /workspace/shared/sd_models/Stable-diffusion/sd_xl_refiner_1.0_0.9vae.safetensors
│   ├── v1-5-pruned.safetensors -> /workspace/shared/sd_models/Stable-diffusion/v1-5-pruned.safetensors
│   └── v2-1_768-ema-pruned.safetensors -> /workspace/shared/sd_models/Stable-diffusion/v2-1_768-ema-pruned.safetensors
├── clip
│   ├── clip_g_hidream.safetensors @ https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/text_encoders/clip_g_hidream.safetensors
│   ├── clip_l_hidream.safetensors @ https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/text_encoders/clip_l_hidream.safetensors
│   ├── llama_3.1_8b_instruct_fp8_scaled.safetensors @ https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/text_encoders/llama_3.1_8b_instruct_fp8_scaled.safetensors
│   ├── t5xxl_fp16.safetensors @ https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors
│   └── t5xxl_fp8_e4m3fn_scaled.safetensors @ https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn_scaled.safetensors
├── clip_vision
├── configs
├── controlnet
├── diffusers
├── diffusion_models
│   └── hidream_i1_dev_bf16.safetensors @ https://huggingface.co/Comfy-Org/HiDream-I1_ComfyUI/resolve/main/split_files/diffusion_models/hidream_i1_dev_bf16.safetensors
├── embeddings
├── gligen
├── hypernetworks
├── ipadapter
├── loras
├── photomaker
├── sams
├── style_models
├── text_encoder
│   └── PixArt-XL-2-1024-MS
│       └── text_encoder
│           ├── config.json @ https://huggingface.co/PixArt-alpha/PixArt-XL-2-1024-MS/resolve/main/text_encoder/config.json
│           ├── model-00001-of-00002.safetensors @ https://huggingface.co/PixArt-alpha/PixArt-XL-2-1024-MS/resolve/main/text_encoder/model-00001-of-00002.safetensors
│           ├── model-00002-of-00002.safetensors @ https://huggingface.co/PixArt-alpha/PixArt-XL-2-1024-MS/resolve/main/text_encoder/model-00002-of-00002.safetensors
│           └── model.safetensors.index.json @ https://huggingface.co/PixArt-alpha/PixArt-XL-2-1024-MS/resolve/main/text_encoder/model.safetensors.index.json
├── ultralytics
├── unet
├── upscale_models
│   └── ltxv-spatial-upscaler-0.9.7.safetensors @ https://huggingface.co/Lightricks/LTX-Video/resolve/main/ltxv-spatial-upscaler-0.9.7.safetensors
├── vae
│   └── ae.safetensors @ https://huggingface.co/lovis93/testllm/resolve/ed9cf1af7465cebca4649157f118e331cf2a084f/ae.safetensors
└── vae_approx

About

No description, website, or topics provided.

Resources

Readme
Activity
Custom properties

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •  

Languages