vessl 0.1.129

A library and CLI for VESSL

vessl-python-sdk

Basic usage

import vessl

vessl.init(organization_name="my-organization")
vessl.create_experiment(...)

Keras

• Use ExperimentCallback

import vessl
from vessl.integration.keras import ExperimentCallback

vessl.init()

# Keras training code
model = Model()
model.compile(...)

# Add integration
model.fit(x, y, epochs=5, callbacks=[ExperimentCallback()])

• Run experiment on Vessl using Web UI or SDK

For M1

docker context create remote --docker "host=ssh://ec2-user@10.110.3.24"
docker context use remote
docker build . -t vessl-python-sdk
docker run vessl-python-sdk

Development Setting

Poetry

Install poetry

Documentation

Add & Install new package

Write Package and it's version to pyproject.toml

poetry lock
poetry install

Enter the virtual environment

poetry shell

Set python version environment

1. Install each python distribution package to your local environment (you can use pyenv)
2. Follow instructions

For faster development

VESSL_ENV constant

VESSL_ENV envvar is default set to prod.

You can override this with dev by setting envvar VESSL_ENV=dev

If you set VESSL_ENV to dev, it automatically sets WEB_HOST and API_HOST to dev server.

• WEB_HOST: https://dev.vssl.ai
• API_HOST: https://api.dev.vssl.ai

It also switches off the sentry configuration on vessl/cli/_main.py:L55.

But also, you can override these automatically set variables(WEB_HOST, API_HOST, etc..) with .env file.

.env file

You can use .env file to set environment variables for your development.

You should put .env file on the root of this project.

See python-dotenv.

e.g.

ENABLE_SENTRY=false             # disables sentry configuration, added in order to disable while VESSL_ENV=dev
VESSL_ENV=dev                   # this automatically sets dev api
VESSL_LOG=DEBUG                 # you don't have to pass it through shell command
# API_HOST=http://localhost     # Set freely when attaching to local devenv
# WEB_HOST=http://localhost

Set alias for local CLI run

You can run your local code as cli with python -m vessl.cli._main.

But for convienience, you can set alias for that and put in your .zshrc or .bashrc

e.g.

alias vessldev="python -m vessl.cli._main"

Disable sentry for local development

Set envvar ENABLE_SENTRY=false on your shell or on .env.

(TEMPORARY GUIDE) How to update dependencies?

will be resolved in PR#210

1. pyproject.toml 업데이트를 한 후,

• poetry export --with dev --without-hashes > requirements.txt로 requirements.txt 업데이트

• poetry export --only test --without-hashes > requirements_test.txt로 업데이트

2. setup.py에도 업데이트

• 이쪽 의존성은 pyproject.toml쪽 내용이랑 다르기 때문에(ㅠㅠㅠ) 적당히 눈치 보면서 업데이트