stable/hyperoptNote.md

41 lines
1.5 KiB
Markdown
Raw Normal View History

2018-03-23 22:55:15 +00:00
Basically, to sum things up:
2018-03-23 04:05:37 +00:00
2018-03-23 22:55:15 +00:00
Make these changes per this PR in these files:
https://github.com/hyperopt/hyperopt/pull/287/files
2018-03-23 04:05:37 +00:00
2018-03-23 22:55:15 +00:00
/usr/local/lib/python3.6/dist-packages/hyperopt/fmin.py
removetrial['state'] == base.JOB_STATE_RUNNING
addtrial['state'] = base.JOB_STATE_RUNNING
Make these changes per this PR into the hyperopt files:
https://github.com/hyperopt/hyperopt/pull/288/files
remove:import six.moves.cPickle as pickle
add: import dill as pickle in the following files:
2018-03-23 04:05:37 +00:00
/usr/local/lib/python3.6/dist-packages/hyperopt/fmin.py
2018-03-23 23:10:40 +00:00
2018-03-23 22:55:15 +00:00
/usr/local/lib/python3.6/dist-packages/hyperopt/main.py
2018-03-23 23:10:40 +00:00
2018-03-23 22:55:15 +00:00
/usr/local/lib/python3.6/dist-packages/hyperopt/mongoexp.py
2018-03-23 23:10:40 +00:00
2018-03-23 22:55:15 +00:00
/usr/local/lib/python3.6/dist-packages/hyperopt/utils.py
Install dill:
pip3.6 install dill
Get this raw file and replace your hyperopy.py with this one in freqtrade/optmize/:
https://raw.githubusercontent.com/MoonGem/freqtrade/develop/freqtrade/optimize/hyperopt.py
then run the mongodb scripts, wait until the worker shows that it does not appear to be working then ctrl+c the worker, however, this is as far as I could get. I think this is the correct process.
However, the queue does eventually empty after awhile and you shouldn't need to ctrlc the worker.
Kinda buggy I guess still.
You may also need to install the latest hyperopt from their github to avoid subscription error:
pip3.6 install git+https://github.com/hyperopt/hyperopt.git
2018-03-23 04:05:37 +00:00
2018-03-24 01:31:53 +00:00
Lastly, copy hyperopt-mongodb.py over freqtrade/optimise/hyperopt.py then run scripts. Wait until the database fills and get result!