adelorenzo/stable
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Dev test notes.

Browse Source
This commit is contained in:
MoonGem 2018-03-23 20:59:26 -05:00 committed by GitHub
parent 07a90f51c0
commit ad3541b376
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 13 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
hyperoptNote.md → hyperopt-note.md
View File

@ -1,5 +1,7 @@
Basically, to sum things up:
The locations of your hyperopt directories may vary. Some reside in .local.
Make these changes per this PR in these files:
https://github.com/hyperopt/hyperopt/pull/287/files
@ -30,21 +32,18 @@ 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
Lastly, copy hyperopt-mongodb.py over freqtrade/optimise/hyperopt.py then run scripts. Wait until the database fills and get result!
replace freqtrade/optimize/hyperopt.py with the hyperopt-mongodb.py in the freqtrade root.
Management tool for mongodb:
https://robomongo.org/download
Then, run the scipts. 5000 epochs can take a few hours even on a high-end machine. You can use the mongodb took above to check the progress on the local machine.
Lastly, read the test_strategy.py file located in user_data/strategy/test_strategy.py
Once you've done that, run scripts. You can restart the worker and the freqtrader hyperopt process but I do not advise it. Monitor the output with mongodb.