Skip to content

Latest commit

 

History

History
71 lines (55 loc) · 2.04 KB

local.md

File metadata and controls

71 lines (55 loc) · 2.04 KB

Local Testing (not required if you don't want to test locally)

Clone the repo, create a venv and install the requirements

git clone https://github.com/ashleykleynhans/runpod-worker-instantid.git
cd runpod-worker-instantid
python3 -m venv venv
source venv/bin/activate
cd src
pip3 install --no-cache-dir torch==2.0.1 torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
pip3 install --no-cache-dir xformers==0.0.22 runpod
pip3 install -r requirements.txt

Install the Checkpoints

python3 download_checkpoints.py

Start the local RunPod Handler API

Use --rp_serve_api command line argument to serve the API locally.

python3 -u rp_handler.py --rp_serve_api

NOTE: You need to keep the RunPod Handler API running in order to run the tests, so open a new terminal window to run the tests.

Set your test data files

You can either overwrite the images in the data directory with your own source and target files, or alternatively, you can edit the scripts in the tests directory to reference the source and target images somewhere else on your system.

Remove credentials from .env

If you have added your RUNPOD_API_KEY and RUNPOD_ENDPOINT_ID to the .env file within this directory, you should first comment them out before attempting to test locally. If the .env file exists and the values are provided, the tests will attempt to send the requests to your RunPod endpoint instead of running locally.

Run test scripts

  1. Ensure that the RunPod Handler API is still running.
  2. Change directory to the tests directory and run one of the scripts, for example:
cd tests
python3 generate.py
  1. This will display the HTTP status code and the filename of the output image, for example:
Status code: 200
Saving image: 792a7e9f-9c36-4d35-b408-0d45d8e2bbcb.jpg

You can then open the output image (in this case 792a7e9f-9c36-4d35-b408-0d45d8e2bbcb.jpg) to view the results.

You obviously need to edit the payload within the script to achieve the desired results.