Categories
AI ComfyUI

ComfyUI on AMD GPU with DirectML

Just in case anyone is trying to get this to work, these are the steps I followed. These steps assume you have a rough idea of what you are trying to accomplish and how to get there; this is not a detailed step-by-step guide. I’m using Windows 11 with a 6700XT and the 23.11.1 drivers.

This was handy:
https://learn.microsoft.com/en-us/windows/ai/directml/gpu-pytorch-windows

Install miniconda.
https://docs.conda.io/en/latest/miniconda.html#windows-installers

Edit 11/28: Start an Anaconda Powershell Prompt.

Clone the ComfyUI code somewhere (you’ll need git installed).

mkdir d:\ai
cd d:\ai

git clone https://github.com/comfyanonymous/ComfyUI.git
cd comfyui

Start a conda environment for python 3.10 and activate it:

conda create --name dml python=3.10
conda activate dml

Verify the python version is 3.10 within the new environment:

python --version

Start installing other bits and pieces (the requirements.txt is in the comfyui root folder):

conda install numpy pandas tensorboard matplotlib tqdm pyyaml -y

pip install opencv-python
pip install wget
pip install torchvision
conda install pytorch cpuonly -c pytorch
pip install -r requirements.txt

Start comfyUI with:

python main.py --directml

You should then get the URL to be able to view ComfyUI.

Categories
AI ChatGPT OpenAI

Some AI resources

Some good youtube channels:

Free courses:

Categories
AI ChatGPT OpenAI

FrameFit

I’ve created a ChatGPT ‘GPT’ to assist in adding borders to images. The general purpose was to add a border to emulate a matte so the end result could be framed in cheap off-the-shelf frames of standard sizes.

You can ask the tool to ‘add a white border of 20 pixels to the image’ or ‘Please add borders to this image to allow for framing in a landscape 11×14 inch frame’ as examples.

Sometimes it does take a couple attempts to get where you need. The more detail you can specify up front about the required output (dimensions for example, or state the output is simply for web, or if the end result should be portrait or landscape) will assist.

The tool can be found here ( FrameFit GPT, and you should be able to provide feedback from within it.

You will need to be a ‘ChatGPT Plus’ user.

Categories
AI ChatGPT OpenAI Photography

ChatGPT GPT: PhotoFocus

ChatGPT introduced ‘GPTs’. I made one to answer photography questions.

PhotoFocus GPT

You do have to be a ‘ChatGPT Plus’ user.