model_server
Serving Vision to Living Things.
Summary
model_server is a service for on-demand computer vision, adapted specifically to image-based feedback in microscopy and other life sciences applications. It abstracts image data access, persists machine learning models, and exposes an extensible API to facilitate low-latency analysis.
Install Git and miniforge
- Install Miniforge for environment management:
https://github.com/conda-forge/miniforge/releases - Under the Start menu, open
Miniforge3 > Miniforge Prompt
Option 1: install model_server as a package:
- Download the most recent version of the built package from:
https://git.embl.de/rhodes/model_server/-/packages/1280 - (optional) activate the target conda environment:
mamba activate <target_environment>
- From the package repository https://git.embl.de/rhodes/model_server/-/packages/ download:
- The most recent requirements.yml
- The most recent .tar.bz2 file containing the built conda package
- In a text editor, open requirements.yml and remove all but the "channels" and "dependencies" blocks, then save.
- Change directories to the location of 'requirements.yml' and install dependencies:
mamba env update -f requirements.yml
- Download the most recent .tar.bz2 file containing the built conda package from:
https://git.embl.de/rhodes/model_server/-/packages/1283 - Change directories to the downloaded file and install model_server package:
mamba install model_server-<version>-py_0.tar.bz2
Option 2: install model_server from source:
- Install Git:
https://git-scm.com/download/win - In the new terminal, clone the model_server repository:
cd %userprofile%
git clone https://almf-staff:KJmFvyPRbpzoVZDqfMzV@git.embl.de/rhodes/model_server.git
- Create the target environment:
mamba env create --file requirements.yml --name model_server_env
- Activate the target environment:
mamba activate model_server_env
- Add the project source as a Python package:
pip install --no-deps -e .
To start the server:
- From the Miniforge prompt, run
mamba activate <target_environment>
- Then run
python -m scripts.run_server --port 6221
- A browser window should appear, with basic status information.