stable/freqtrade/freqai/freqai_interface.py

174 lines
6.7 KiB
Python
Raw Normal View History

import gc
2022-05-04 15:53:40 +00:00
import logging
import shutil
2022-05-04 15:53:40 +00:00
from abc import ABC, abstractmethod
from pathlib import Path
from typing import Any, Dict, Tuple
import numpy as np
import pandas as pd
from pandas import DataFrame
from freqtrade.freqai.data_handler import DataHandler
pd.options.mode.chained_assignment = None
2022-05-04 15:53:40 +00:00
logger = logging.getLogger(__name__)
class IFreqaiModel(ABC):
"""
Class containing all tools for training and prediction in the strategy.
User models should inherit from this class as shown in
templates/ExamplePredictionModel.py where the user overrides
train(), predict(), fit(), and make_labels().
2022-05-03 08:28:13 +00:00
Author: Robert Caulk, rob.caulk@gmail.com
"""
def __init__(self, config: Dict[str, Any]) -> None:
self.config = config
self.freqai_info = config["freqai"]
self.data_split_parameters = config["freqai"]["data_split_parameters"]
self.model_training_parameters = config["freqai"]["model_training_parameters"]
self.feature_parameters = config["freqai"]["feature_parameters"]
self.full_path = Path(
config["user_data_dir"]
/ "models"
/ str(self.freqai_info["full_timerange"] + self.freqai_info["identifier"])
)
self.time_last_trained = None
self.current_time = None
self.model = None
self.predictions = None
if not self.full_path.is_dir():
self.full_path.mkdir(parents=True, exist_ok=True)
shutil.copy(
self.config["config_files"][0],
Path(self.full_path / self.config["config_files"][0]),
)
def start(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
"""
Entry point to the FreqaiModel, it will train a new model if
necesssary before making the prediction.
The backtesting and training paradigm is a sliding training window
with a following backtest window. Both windows slide according to the
length of the backtest window. This function is not intended to be
overridden by children of IFreqaiModel, but technically, it can be
if the user wishes to make deeper changes to the sliding window
logic.
:params:
:dataframe: Full dataframe coming from strategy - it contains entire
backtesting timerange + additional historical data necessary to train
the model.
:metadata: pair metadataa coming from strategy.
"""
self.pair = metadata["pair"]
self.dh = DataHandler(self.config, dataframe)
2022-05-04 15:53:40 +00:00
logger.info(
"going to train",
len(self.dh.training_timeranges),
"timeranges:",
self.dh.training_timeranges,
)
# Loop enforcing the sliding window training/backtesting paragigm
# tr_train is the training time range e.g. 1 historical month
# tr_backtest is the backtesting time range e.g. the week directly
# following tr_train. Both of these windows slide through the
# entire backtest
for tr_train, tr_backtest in zip(
self.dh.training_timeranges, self.dh.backtesting_timeranges
):
gc.collect()
# self.config['timerange'] = tr_train
self.dh.data = {} # clean the pair specific data between models
self.freqai_info["training_timerange"] = tr_train
dataframe_train = self.dh.slice_dataframe(tr_train, dataframe)
dataframe_backtest = self.dh.slice_dataframe(tr_backtest, dataframe)
2022-05-04 15:53:40 +00:00
logger.info("training", self.pair, "for", tr_train)
# self.dh.model_path = self.full_path + "/" + "sub-train" + "-" + str(tr_train) + "/"
self.dh.model_path = Path(self.full_path / str("sub-train" + "-" + str(tr_train)))
if not self.model_exists(self.pair, training_timerange=tr_train):
self.model = self.train(dataframe_train, metadata)
self.dh.save_data(self.model)
else:
self.model = self.dh.load_data(self.dh.model_path)
preds, do_preds = self.predict(dataframe_backtest)
self.dh.append_predictions(preds, do_preds, len(dataframe_backtest))
self.dh.fill_predictions(len(dataframe))
return self.dh.predictions, self.dh.do_predict, self.dh.target_mean, self.dh.target_std
def make_labels(self, dataframe: DataFrame) -> DataFrame:
"""
User defines the labels here (target values).
:params:
:dataframe: the full dataframe for the present training period
"""
return dataframe
2022-05-04 15:53:40 +00:00
@abstractmethod
2022-05-03 08:28:13 +00:00
def train(self, unfiltered_dataframe: DataFrame, metadata: dict) -> Any:
"""
Filter the training data and train a model to it. Train makes heavy use of the datahandler
2022-05-03 08:28:13 +00:00
for storing, saving, loading, and analyzing the data.
:params:
:unfiltered_dataframe: Full dataframe for the current training period
:metadata: pair metadata from strategy.
:returns:
:model: Trained model which can be used to inference (self.predict)
"""
2022-05-03 08:28:13 +00:00
return Any
2022-05-04 15:53:40 +00:00
@abstractmethod
def fit(self) -> Any:
"""
Most regressors use the same function names and arguments e.g. user
can drop in LGBMRegressor in place of CatBoostRegressor and all data
management will be properly handled by Freqai.
:params:
:data_dictionary: the dictionary constructed by DataHandler to hold
all the training and test data/labels.
"""
return Any
2022-05-04 15:53:40 +00:00
@abstractmethod
def predict(self, dataframe: DataFrame) -> Tuple[np.array, np.array]:
"""
Filter the prediction features data and predict with it.
:param: unfiltered_dataframe: Full dataframe for the current backtest period.
:return:
:predictions: np.array of predictions
:do_predict: np.array of 1s and 0s to indicate places where freqai needed to remove
data (NaNs) or felt uncertain about data (PCA and DI index)
"""
return np.array([]), np.array([])
def model_exists(self, pair: str, training_timerange: str) -> bool:
"""
Given a pair and path, check if a model already exists
:param pair: pair e.g. BTC/USD
:param path: path to model
"""
coin, _ = pair.split("/")
self.dh.model_filename = "cb_" + coin.lower() + "_" + training_timerange
path_to_modelfile = Path(self.dh.model_path / str(self.dh.model_filename + "_model.joblib"))
file_exists = path_to_modelfile.is_file()
if file_exists:
2022-05-04 15:53:40 +00:00
logger.info("Found model at", self.dh.model_path / self.dh.model_filename)
else:
2022-05-04 15:53:40 +00:00
logger.info("Could not find model at", self.dh.model_path / self.dh.model_filename)
return file_exists