-
Notifications
You must be signed in to change notification settings - Fork 32
/
.travis.yml
39 lines (34 loc) · 1.46 KB
/
.travis.yml
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
dist: trusty
sudo: false # to use container-based infra, see: http://docs.travis-ci.com/user/migrating-from-legacy/
language:
- python
python:
- "2.7"
cache:
directories:
- download
- $HOME/.cache/pip
- $HOME/miniconda/envs/test # to avoid repetitively setting up Ana/Miniconda environment
before_install:
- deactivate # leaving Travis' virtualenv first since otherwise Jupyter/IPython gets confused with conda inside a virtualenv (See: https://github.com/ipython/ipython/issues/8898)
- mkdir -p download
- cd download
- travis_retry wget -c https://repo.continuum.io/miniconda/Miniconda2-latest-Linux-x86_64.sh -O miniconda.sh
- chmod +x miniconda.sh
- bash miniconda.sh -b -f -p ~/miniconda
- cd ..
- export PATH=~/miniconda/bin:$PATH
- conda update --yes conda
- conda install --yes pytorch torchvision cuda80 -c soumith
- conda install --yes -c anaconda pandas
- conda install --yes -c anaconda scipy
script:
# memory based checkpointing test (default)
- unzip numerical_test/yf_data.dat.zip -d ./numerical_test/
- python numerical_test/test-ckpt-memory.py --nhidden=50 --debug --use_cuda --use_lstm
# disk based checkpointing (default)
- python numerical_test/test-ckpt-disk.py --nhidden=50 --debug --use_cuda --use_lstm
# simulate cases with gradient spike
- python numerical_test/Simulate_Extreme_Cases_Jump.py
# simulate cases with continous 0 gradients
- python numerical_test/Simulate_Extreme_Cases_Zero_Epoch.py