Skip to content

Instantly share code, notes, and snippets.

@VolkerH
VolkerH / travis_to_conda.md
Created March 28, 2019 05:25 — forked from zshaheen/travis_to_conda.md
How to Setup Automatic Uploads to Anaconda from Travis CI in 15 minutes

How to Setup Automatic Uploads to Anaconda from Travis CI in 15 minutes

TL;DR: Edit .travis.yaml to install Anaconda and to run conda_upload.sh after testing. Edit meta.yaml to take in the environmental variables $VERSION and $CONDA_BLD_PATH. Create conda_upload.sh which sets the needed environmental variables, builds the tar archive, and uploads it to Anaconda. Finally edit some stuff on your Anaconda and Travis CI account so they can talk.

Intro

The following steps will detail how to automatically trigger Anaconda builds and uploads from Travis CI. This will only upload successful builds in the master branch and if there are multiple commits in a single day, it'll only keep the latest one. Both of these settings can easily be changed.

Edit .travis.yaml

First, edit .travis.yml so that it installs Anaconda.

install:
@VolkerH
VolkerH / keras_weighted_binary_crossentropy.py
Created February 9, 2019 14:27 — forked from ksugar/keras_weighted_binary_crossentropy.py
Custom loss function for weighted binary crossentropy in Keras with Tensorflow
# code picked up from https://github.com/keras-team/keras/blob/master/keras/backend/tensorflow_backend.py
# Just used tf.nn.weighted_cross_entropy_with_logits instead of tf.nn.sigmoid_cross_entropy_with_logits with input pos_weight in calculation
import tensorflow as tf
from keras import backend as K
""" Weighted binary crossentropy between an output tensor and a target tensor.
# Arguments
pos_weight: A coefficient to use on the positive examples.
# Returns