❓ Exception: inputs are all NaN
rodrigogs opened this issue · comments
I'm having the following error using mgm-hurry on Ubuntu 21.
I just executed mgm-hurry up
and left everything as default. Any idea?
INFO - Using optimizer random state: 46429
INFO - Using indicator startup period: 400 ...
INFO - NumExpr defaulting to 8 threads.
INFO - Loading data from 2021-04-29 14:40:00 up to 2021-10-26 00:00:00 (179 days).
INFO - Dataload complete. Calculating indicators
WARNING - No history data for pair: "BTC/USDT", timeframe: 4h. Use `freqtrade download-data` to download the data
ERROR - Fatal exception!
Traceback (most recent call last):
File "/home/rodrigosilva/Workspace/mgm-dev/freqtrade-mgm/freqtrade/main.py", line 37, in main
return_code = args['func'](args)
File "/home/rodrigosilva/Workspace/mgm-dev/freqtrade-mgm/freqtrade/commands/optimize_commands.py", line 87, in start_hyperopt
hyperopt.start()
File "/home/rodrigosilva/Workspace/mgm-dev/freqtrade-mgm/freqtrade/optimize/hyperopt.py", line 398, in start
self.prepare_hyperopt_data()
File "/home/rodrigosilva/Workspace/mgm-dev/freqtrade-mgm/freqtrade/optimize/hyperopt.py", line 378, in prepare_hyperopt_data
preprocessed = self.backtesting.strategy.ohlcvdata_to_dataframe(data)
File "/home/rodrigosilva/Workspace/mgm-dev/freqtrade-mgm/freqtrade/strategy/interface.py", line 716, in ohlcvdata_to_dataframe
return {pair: self.advise_indicators(pair_data.copy(), {'pair': pair})
File "/home/rodrigosilva/Workspace/mgm-dev/freqtrade-mgm/freqtrade/strategy/interface.py", line 716, in <dictcomp>
return {pair: self.advise_indicators(pair_data.copy(), {'pair': pair})
File "/home/rodrigosilva/Workspace/mgm-dev/freqtrade-mgm/freqtrade/strategy/interface.py", line 733, in advise_indicators
return self.populate_indicators(dataframe, metadata)
File "/home/rodrigosilva/Workspace/mgm-dev/freqtrade-mgm/monigomani/user_data/strategies/MoniGoManiHyperStrategy.py", line 159, in populate_indicators
return self._populate_indicators(dataframe=dataframe, metadata=metadata)
File "/home/rodrigosilva/Workspace/mgm-dev/freqtrade-mgm/monigomani/user_data/strategies/MasterMoniGoManiHyperStrategy.py", line 514, in _populate_indicators
core_trend = self._populate_core_trend(core_trend, metadata)
File "/home/rodrigosilva/Workspace/mgm-dev/freqtrade-mgm/monigomani/user_data/strategies/MasterMoniGoManiHyperStrategy.py", line 431, in _populate_core_trend
dataframe['ht_trendmode'] = ta.HT_TRENDMODE(dataframe)
File "talib/_abstract.pxi", line 399, in talib._ta_lib.Function.__call__
File "talib/_abstract.pxi", line 441, in talib._ta_lib.Function.__call_function
File "/home/rodrigosilva/Workspace/mgm-dev/freqtrade-mgm/.env/lib/python3.9/site-packages/talib/__init__.py", line 27, in wrapper
return func(*args, **kwargs)
File "talib/_func.pxi", line 3169, in talib._ta_lib.HT_TRENDMODE
File "talib/_func.pxi", line 68, in talib._ta_lib.check_begidx1
Exception: inputs are all NaN
ERROR - 🤷 No "best" results found during the HyperOpt, aborting further tests since there are no HyperOpt results to test upon.
You still need to mgm-hurry download_candle_data
for 4h
candles
Sorry for commenting in this closed thread @Rikj000 , but I'm facing this issue as well even though my configuration isn't set to use 4h timeframe. It's true that the moment I include in my downloaded data the 4h candles then the error disappears and I'm able to run the hyperopt. However, in this situation the hyperopt doesn't seem to find any best result.
In short, despite of setting in the configuration to use 5m/30m candles the hyperopt seems to require 4h, causing (I'm assuming) not to find any best result.
Please take this as a very humble input from somebody that wants to participate and that values the hard work you've invested in this project.
Sorry for commenting in this closed thread @Rikj000 , but I'm facing this issue as well even though my configuration isn't set to use 4h timeframe. It's true that the moment I include in my downloaded data the 4h candles then the error disappears and I'm able to run the hyperopt. However, in this situation the hyperopt doesn't seem to find any best result.
In short, despite of setting in the configuration to use 5m/30m candles the hyperopt seems to require 4h, causing (I'm assuming) not to find any best result.
Please take this as a very humble input from somebody that wants to participate and that values the hard work you've invested in this project.
Hi @Bloodsucker, no problem glad you took the time to write down your concerns! 🦾
My guess would be that it's just not that clear that MGM was using 4h
candles.
In the default mgm-config
we use a "timeframe": "30m"
and a "core_trend_timeframe_multiplier": 8
,
which resulted into a "core_trend_timeframe": "4h"
(30m x 8 = 4h).
Don't worry though, the core_trend_timeframe_multiplier
wasn't documented yet and I actually just switched it back to a more simple core_trend_timeframe
again + I added a new TimeFrames
Documentation section 🙂
FYI: When you update with mgm-hurry install_mgm
you currently still will manually have to update your mgm-config
This was such a fast and informative answer. I've managed to get that new commit and I've also let the script to replace to the new one. Very much appreciated.
Seriously, keep this hard work and attitude for this dream project.