Merge branch 'develop' into backtest_live_models

This commit is contained in:
Wagner Costa Santos
2022-10-20 11:59:37 -03:00
43 changed files with 650 additions and 236 deletions

View File

@@ -15,9 +15,9 @@ from freqtrade.commands.db_commands import start_convert_db
from freqtrade.commands.deploy_commands import (start_create_userdir, start_install_ui,
start_new_strategy)
from freqtrade.commands.hyperopt_commands import start_hyperopt_list, start_hyperopt_show
from freqtrade.commands.list_commands import (start_list_exchanges, start_list_markets,
start_list_strategies, start_list_timeframes,
start_show_trades)
from freqtrade.commands.list_commands import (start_list_exchanges, start_list_freqAI_models,
start_list_markets, start_list_strategies,
start_list_timeframes, start_show_trades)
from freqtrade.commands.optimize_commands import (start_backtesting, start_backtesting_show,
start_edge, start_hyperopt)
from freqtrade.commands.pairlist_commands import start_test_pairlist

View File

@@ -42,6 +42,8 @@ ARGS_EDGE = ARGS_COMMON_OPTIMIZE + ["stoploss_range"]
ARGS_LIST_STRATEGIES = ["strategy_path", "print_one_column", "print_colorized",
"recursive_strategy_search"]
ARGS_LIST_FREQAIMODELS = ["freqaimodel_path", "print_one_column", "print_colorized"]
ARGS_LIST_HYPEROPTS = ["hyperopt_path", "print_one_column", "print_colorized"]
ARGS_BACKTEST_SHOW = ["exportfilename", "backtest_show_pair_list"]
@@ -107,8 +109,8 @@ ARGS_ANALYZE_ENTRIES_EXITS = ["exportfilename", "analysis_groups", "enter_reason
"exit_reason_list", "indicator_list"]
NO_CONF_REQURIED = ["convert-data", "convert-trade-data", "download-data", "list-timeframes",
"list-markets", "list-pairs", "list-strategies", "list-data",
"hyperopt-list", "hyperopt-show", "backtest-filter",
"list-markets", "list-pairs", "list-strategies", "list-freqaimodels",
"list-data", "hyperopt-list", "hyperopt-show", "backtest-filter",
"plot-dataframe", "plot-profit", "show-trades", "trades-to-ohlcv"]
NO_CONF_ALLOWED = ["create-userdir", "list-exchanges", "new-strategy"]
@@ -193,10 +195,11 @@ class Arguments:
start_create_userdir, start_download_data, start_edge,
start_hyperopt, start_hyperopt_list, start_hyperopt_show,
start_install_ui, start_list_data, start_list_exchanges,
start_list_markets, start_list_strategies,
start_list_timeframes, start_new_config, start_new_strategy,
start_plot_dataframe, start_plot_profit, start_show_trades,
start_test_pairlist, start_trading, start_webserver)
start_list_freqAI_models, start_list_markets,
start_list_strategies, start_list_timeframes,
start_new_config, start_new_strategy, start_plot_dataframe,
start_plot_profit, start_show_trades, start_test_pairlist,
start_trading, start_webserver)
subparsers = self.parser.add_subparsers(dest='command',
# Use custom message when no subhandler is added
@@ -363,6 +366,15 @@ class Arguments:
list_strategies_cmd.set_defaults(func=start_list_strategies)
self._build_args(optionlist=ARGS_LIST_STRATEGIES, parser=list_strategies_cmd)
# Add list-freqAI Models subcommand
list_freqaimodels_cmd = subparsers.add_parser(
'list-freqaimodels',
help='Print available freqAI models.',
parents=[_common_parser],
)
list_freqaimodels_cmd.set_defaults(func=start_list_freqAI_models)
self._build_args(optionlist=ARGS_LIST_FREQAIMODELS, parser=list_freqaimodels_cmd)
# Add list-timeframes subcommand
list_timeframes_cmd = subparsers.add_parser(
'list-timeframes',

View File

@@ -1,7 +1,6 @@
import csv
import logging
import sys
from pathlib import Path
from typing import Any, Dict, List
import rapidjson
@@ -10,7 +9,6 @@ from colorama import init as colorama_init
from tabulate import tabulate
from freqtrade.configuration import setup_utils_configuration
from freqtrade.constants import USERPATH_STRATEGIES
from freqtrade.enums import RunMode
from freqtrade.exceptions import OperationalException
from freqtrade.exchange import market_is_active, validate_exchanges
@@ -41,7 +39,7 @@ def start_list_exchanges(args: Dict[str, Any]) -> None:
print(tabulate(exchanges, headers=['Exchange name', 'Valid', 'reason']))
def _print_objs_tabular(objs: List, print_colorized: bool, base_dir: Path) -> None:
def _print_objs_tabular(objs: List, print_colorized: bool) -> None:
if print_colorized:
colorama_init(autoreset=True)
red = Fore.RED
@@ -55,7 +53,7 @@ def _print_objs_tabular(objs: List, print_colorized: bool, base_dir: Path) -> No
names = [s['name'] for s in objs]
objs_to_print = [{
'name': s['name'] if s['name'] else "--",
'location': s['location'].relative_to(base_dir),
'location': s['location_rel'],
'status': (red + "LOAD FAILED" + reset if s['class'] is None
else "OK" if names.count(s['name']) == 1
else yellow + "DUPLICATE NAME" + reset)
@@ -76,9 +74,8 @@ def start_list_strategies(args: Dict[str, Any]) -> None:
"""
config = setup_utils_configuration(args, RunMode.UTIL_NO_EXCHANGE)
directory = Path(config.get('strategy_path', config['user_data_dir'] / USERPATH_STRATEGIES))
strategy_objs = StrategyResolver.search_all_objects(
directory, not args['print_one_column'], config.get('recursive_strategy_search', False))
config, not args['print_one_column'], config.get('recursive_strategy_search', False))
# Sort alphabetically
strategy_objs = sorted(strategy_objs, key=lambda x: x['name'])
for obj in strategy_objs:
@@ -90,7 +87,22 @@ def start_list_strategies(args: Dict[str, Any]) -> None:
if args['print_one_column']:
print('\n'.join([s['name'] for s in strategy_objs]))
else:
_print_objs_tabular(strategy_objs, config.get('print_colorized', False), directory)
_print_objs_tabular(strategy_objs, config.get('print_colorized', False))
def start_list_freqAI_models(args: Dict[str, Any]) -> None:
"""
Print files with FreqAI models custom classes available in the directory
"""
config = setup_utils_configuration(args, RunMode.UTIL_NO_EXCHANGE)
from freqtrade.resolvers.freqaimodel_resolver import FreqaiModelResolver
model_objs = FreqaiModelResolver.search_all_objects(config, not args['print_one_column'])
# Sort alphabetically
model_objs = sorted(model_objs, key=lambda x: x['name'])
if args['print_one_column']:
print('\n'.join([s['name'] for s in model_objs]))
else:
_print_objs_tabular(model_objs, config.get('print_colorized', False))
def start_list_timeframes(args: Dict[str, Any]) -> None:

View File

@@ -540,6 +540,8 @@ CONF_SCHEMA = {
"properties": {
"enabled": {"type": "boolean", "default": False},
"keras": {"type": "boolean", "default": False},
"write_metrics_to_disk": {"type": "boolean", "default": False},
"purge_old_models": {"type": "boolean", "default": True},
"conv_width": {"type": "integer", "default": 2},
"train_period_days": {"type": "integer", "default": 0},
"backtest_period_days": {"type": "number", "default": 7},

View File

@@ -410,11 +410,13 @@ class Exchange:
else:
return DataFrame()
def get_contract_size(self, pair: str) -> float:
def get_contract_size(self, pair: str) -> Optional[float]:
if self.trading_mode == TradingMode.FUTURES:
market = self.markets[pair]
market = self.markets.get(pair, {})
contract_size: float = 1.0
if market['contractSize'] is not None:
if not market:
return None
if market.get('contractSize') is not None:
# ccxt has contractSize in markets as string
contract_size = float(market['contractSize'])
return contract_size
@@ -1934,6 +1936,7 @@ class Exchange:
candle_limit = self.ohlcv_candle_limit(timeframe, self._config['candle_type_def'])
# Age out old candles
ohlcv_df = ohlcv_df.tail(candle_limit + self._startup_candle_count)
ohlcv_df = ohlcv_df.reset_index(drop=True)
self._klines[(pair, timeframe, c_type)] = ohlcv_df
else:
self._klines[(pair, timeframe, c_type)] = ohlcv_df

View File

@@ -1,14 +1,15 @@
import collections
import json
import logging
import re
import shutil
import threading
from datetime import datetime, timezone
from pathlib import Path
from typing import Any, Dict, Tuple, TypedDict
import numpy as np
import pandas as pd
import psutil
import rapidjson
from joblib import dump, load
from joblib.externals import cloudpickle
@@ -65,6 +66,8 @@ class FreqaiDataDrawer:
self.pair_dict: Dict[str, pair_info] = {}
# dictionary holding all actively inferenced models in memory given a model filename
self.model_dictionary: Dict[str, Any] = {}
# all additional metadata that we want to keep in ram
self.meta_data_dictionary: Dict[str, Dict[str, Any]] = {}
self.model_return_values: Dict[str, DataFrame] = {}
self.historic_data: Dict[str, Dict[str, DataFrame]] = {}
self.historic_predictions: Dict[str, DataFrame] = {}
@@ -78,30 +81,60 @@ class FreqaiDataDrawer:
self.historic_predictions_bkp_path = Path(
self.full_path / "historic_predictions.backup.pkl")
self.pair_dictionary_path = Path(self.full_path / "pair_dictionary.json")
self.metric_tracker_path = Path(self.full_path / "metric_tracker.json")
self.follow_mode = follow_mode
if follow_mode:
self.create_follower_dict()
self.load_drawer_from_disk()
self.load_historic_predictions_from_disk()
self.load_metric_tracker_from_disk()
self.training_queue: Dict[str, int] = {}
self.history_lock = threading.Lock()
self.save_lock = threading.Lock()
self.pair_dict_lock = threading.Lock()
self.metric_tracker_lock = threading.Lock()
self.old_DBSCAN_eps: Dict[str, float] = {}
self.empty_pair_dict: pair_info = {
"model_filename": "", "trained_timestamp": 0,
"data_path": "", "extras": {}}
self.metric_tracker: Dict[str, Dict[str, Dict[str, list]]] = {}
def update_metric_tracker(self, metric: str, value: float, pair: str) -> None:
"""
General utility for adding and updating custom metrics. Typically used
for adding training performance, train timings, inferenc timings, cpu loads etc.
"""
with self.metric_tracker_lock:
if pair not in self.metric_tracker:
self.metric_tracker[pair] = {}
if metric not in self.metric_tracker[pair]:
self.metric_tracker[pair][metric] = {'timestamp': [], 'value': []}
timestamp = int(datetime.now(timezone.utc).timestamp())
self.metric_tracker[pair][metric]['value'].append(value)
self.metric_tracker[pair][metric]['timestamp'].append(timestamp)
def collect_metrics(self, time_spent: float, pair: str):
"""
Add metrics to the metric tracker dictionary
"""
load1, load5, load15 = psutil.getloadavg()
cpus = psutil.cpu_count()
self.update_metric_tracker('train_time', time_spent, pair)
self.update_metric_tracker('cpu_load1min', load1 / cpus, pair)
self.update_metric_tracker('cpu_load5min', load5 / cpus, pair)
self.update_metric_tracker('cpu_load15min', load15 / cpus, pair)
def load_drawer_from_disk(self):
"""
Locate and load a previously saved data drawer full of all pair model metadata in
present model folder.
:return: bool - whether or not the drawer was located
Load any existing metric tracker that may be present.
"""
exists = self.pair_dictionary_path.is_file()
if exists:
with open(self.pair_dictionary_path, "r") as fp:
self.pair_dict = json.load(fp)
self.pair_dict = rapidjson.load(fp, number_mode=rapidjson.NM_NATIVE)
elif not self.follow_mode:
logger.info("Could not find existing datadrawer, starting from scratch")
else:
@@ -110,7 +143,18 @@ class FreqaiDataDrawer:
"sending null values back to strategy"
)
return exists
def load_metric_tracker_from_disk(self):
"""
Tries to load an existing metrics dictionary if the user
wants to collect metrics.
"""
if self.freqai_info.get('write_metrics_to_disk', False):
exists = self.metric_tracker_path.is_file()
if exists:
with open(self.metric_tracker_path, "r") as fp:
self.metric_tracker = rapidjson.load(fp, number_mode=rapidjson.NM_NATIVE)
else:
logger.info("Could not find existing metric tracker, starting from scratch")
def load_historic_predictions_from_disk(self):
"""
@@ -146,7 +190,7 @@ class FreqaiDataDrawer:
def save_historic_predictions_to_disk(self):
"""
Save data drawer full of all pair model metadata in present model folder.
Save historic predictions pickle to disk
"""
with open(self.historic_predictions_path, "wb") as fp:
cloudpickle.dump(self.historic_predictions, fp, protocol=cloudpickle.DEFAULT_PROTOCOL)
@@ -154,6 +198,15 @@ class FreqaiDataDrawer:
# create a backup
shutil.copy(self.historic_predictions_path, self.historic_predictions_bkp_path)
def save_metric_tracker_to_disk(self):
"""
Save metric tracker of all pair metrics collected.
"""
with self.save_lock:
with open(self.metric_tracker_path, 'w') as fp:
rapidjson.dump(self.metric_tracker, fp, default=self.np_encoder,
number_mode=rapidjson.NM_NATIVE)
def save_drawer_to_disk(self):
"""
Save data drawer full of all pair model metadata in present model folder.
@@ -453,9 +506,14 @@ class FreqaiDataDrawer:
)
# if self.live:
# store as much in ram as possible to increase performance
self.model_dictionary[coin] = model
self.pair_dict[coin]["model_filename"] = dk.model_filename
self.pair_dict[coin]["data_path"] = str(dk.data_path)
if coin not in self.meta_data_dictionary:
self.meta_data_dictionary[coin] = {}
self.meta_data_dictionary[coin]["train_df"] = dk.data_dictionary["train_features"]
self.meta_data_dictionary[coin]["meta_data"] = dk.data
self.save_drawer_to_disk()
return
@@ -466,7 +524,7 @@ class FreqaiDataDrawer:
presaved backtesting (prediction file loading).
"""
with open(dk.data_path / f"{dk.model_filename}_metadata.json", "r") as fp:
dk.data = json.load(fp)
dk.data = rapidjson.load(fp, number_mode=rapidjson.NM_NATIVE)
dk.training_features_list = dk.data["training_features_list"]
dk.label_list = dk.data["label_list"]
@@ -492,14 +550,19 @@ class FreqaiDataDrawer:
/ dk.data_path.parts[-1]
)
with open(dk.data_path / f"{dk.model_filename}_metadata.json", "r") as fp:
dk.data = json.load(fp)
dk.training_features_list = dk.data["training_features_list"]
dk.label_list = dk.data["label_list"]
if coin in self.meta_data_dictionary:
dk.data = self.meta_data_dictionary[coin]["meta_data"]
dk.data_dictionary["train_features"] = self.meta_data_dictionary[coin]["train_df"]
else:
with open(dk.data_path / f"{dk.model_filename}_metadata.json", "r") as fp:
dk.data = rapidjson.load(fp, number_mode=rapidjson.NM_NATIVE)
dk.data_dictionary["train_features"] = pd.read_pickle(
dk.data_path / f"{dk.model_filename}_trained_df.pkl"
)
dk.data_dictionary["train_features"] = pd.read_pickle(
dk.data_path / f"{dk.model_filename}_trained_df.pkl"
)
dk.training_features_list = dk.data["training_features_list"]
dk.label_list = dk.data["label_list"]
# try to access model in memory instead of loading object from disk to save time
if dk.live and coin in self.model_dictionary:
@@ -627,22 +690,3 @@ class FreqaiDataDrawer:
).reset_index(drop=True)
return corr_dataframes, base_dataframes
# to be used if we want to send predictions directly to the follower instead of forcing
# follower to load models and inference
# def save_model_return_values_to_disk(self) -> None:
# with open(self.full_path / str('model_return_values.json'), "w") as fp:
# json.dump(self.model_return_values, fp, default=self.np_encoder)
# def load_model_return_values_from_disk(self, dk: FreqaiDataKitchen) -> FreqaiDataKitchen:
# exists = Path(self.full_path / str('model_return_values.json')).resolve().exists()
# if exists:
# with open(self.full_path / str('model_return_values.json'), "r") as fp:
# self.model_return_values = json.load(fp)
# elif not self.follow_mode:
# logger.info("Could not find existing datadrawer, starting from scratch")
# else:
# logger.warning(f'Follower could not find pair_dictionary at {self.full_path} '
# 'sending null values back to strategy')
# return exists, dk

View File

@@ -7,7 +7,7 @@ from collections import deque
from datetime import datetime, timezone
from pathlib import Path
from threading import Lock
from typing import Any, Dict, List, Tuple
from typing import Any, Dict, List, Literal, Tuple
import numpy as np
import pandas as pd
@@ -148,7 +148,7 @@ class IFreqaiModel(ABC):
dataframe = dk.remove_features_from_df(dk.return_dataframe)
self.clean_up()
if self.live:
self.inference_timer('stop')
self.inference_timer('stop', metadata["pair"])
return dataframe
def clean_up(self):
@@ -217,12 +217,14 @@ class IFreqaiModel(ABC):
logger.warning(f"Training {pair} raised exception {msg.__class__.__name__}. "
f"Message: {msg}, skipping.")
self.train_timer('stop')
self.train_timer('stop', pair)
# only rotate the queue after the first has been trained.
self.train_queue.rotate(-1)
self.dd.save_historic_predictions_to_disk()
if self.freqai_info.get('write_metrics_to_disk', False):
self.dd.save_metric_tracker_to_disk()
def start_backtesting(
self, dataframe: DataFrame, metadata: dict, dk: FreqaiDataKitchen
@@ -677,7 +679,7 @@ class IFreqaiModel(ABC):
return
def inference_timer(self, do='start'):
def inference_timer(self, do: Literal['start', 'stop'] = 'start', pair: str = ''):
"""
Timer designed to track the cumulative time spent in FreqAI for one pass through
the whitelist. This will check if the time spent is more than 1/4 the time
@@ -688,7 +690,10 @@ class IFreqaiModel(ABC):
self.begin_time = time.time()
elif do == 'stop':
end = time.time()
self.inference_time += (end - self.begin_time)
time_spent = (end - self.begin_time)
if self.freqai_info.get('write_metrics_to_disk', False):
self.dd.update_metric_tracker('inference_time', time_spent, pair)
self.inference_time += time_spent
if self.pair_it == self.total_pairs:
logger.info(
f'Total time spent inferencing pairlist {self.inference_time:.2f} seconds')
@@ -699,7 +704,7 @@ class IFreqaiModel(ABC):
self.inference_time = 0
return
def train_timer(self, do='start'):
def train_timer(self, do: Literal['start', 'stop'] = 'start', pair: str = ''):
"""
Timer designed to track the cumulative time spent training the full pairlist in
FreqAI.
@@ -709,7 +714,11 @@ class IFreqaiModel(ABC):
self.begin_time_train = time.time()
elif do == 'stop':
end = time.time()
self.train_time += (end - self.begin_time_train)
time_spent = (end - self.begin_time_train)
if self.freqai_info.get('write_metrics_to_disk', False):
self.dd.collect_metrics(time_spent, pair)
self.train_time += time_spent
if self.pair_it_train == self.total_pairs:
logger.info(
f'Total time spent training pairlist {self.train_time:.2f} seconds')

View File

@@ -1,4 +1,5 @@
import logging
import sys
from pathlib import Path
from typing import Any, Dict
@@ -30,6 +31,14 @@ class CatboostClassifier(BaseClassifierModel):
label=data_dictionary["train_labels"],
weight=data_dictionary["train_weights"],
)
if self.freqai_info.get("data_split_parameters", {}).get("test_size", 0.1) == 0:
test_data = None
else:
test_data = Pool(
data=data_dictionary["test_features"],
label=data_dictionary["test_labels"],
weight=data_dictionary["test_weights"],
)
cbr = CatBoostClassifier(
allow_writing_files=True,
@@ -40,6 +49,7 @@ class CatboostClassifier(BaseClassifierModel):
init_model = self.get_init_model(dk.pair)
cbr.fit(train_data, init_model=init_model)
cbr.fit(X=train_data, eval_set=test_data, init_model=init_model,
log_cout=sys.stdout, log_cerr=sys.stderr)
return cbr

View File

@@ -1,4 +1,5 @@
import logging
import sys
from pathlib import Path
from typing import Any, Dict
@@ -47,6 +48,7 @@ class CatboostRegressor(BaseRegressionModel):
**self.model_training_parameters,
)
model.fit(X=train_data, eval_set=test_data, init_model=init_model)
model.fit(X=train_data, eval_set=test_data, init_model=init_model,
log_cout=sys.stdout, log_cerr=sys.stderr)
return model

View File

@@ -1,4 +1,5 @@
import logging
import sys
from pathlib import Path
from typing import Any, Dict
@@ -58,8 +59,10 @@ class CatboostRegressorMultiTarget(BaseRegressionModel):
fit_params = []
for i in range(len(eval_sets)):
fit_params.append(
{'eval_set': eval_sets[i], 'init_model': init_models[i]})
fit_params.append({
'eval_set': eval_sets[i], 'init_model': init_models[i],
'log_cout': sys.stdout, 'log_cerr': sys.stderr,
})
model = FreqaiMultiOutputRegressor(estimator=cbr)
thread_training = self.freqai_info.get('multitarget_parallel_training', False)

View File

@@ -0,0 +1,85 @@
import logging
from typing import Any, Dict, Tuple
import numpy as np
import numpy.typing as npt
import pandas as pd
from pandas import DataFrame
from pandas.api.types import is_integer_dtype
from sklearn.preprocessing import LabelEncoder
from xgboost import XGBRFClassifier
from freqtrade.freqai.base_models.BaseClassifierModel import BaseClassifierModel
from freqtrade.freqai.data_kitchen import FreqaiDataKitchen
logger = logging.getLogger(__name__)
class XGBoostRFClassifier(BaseClassifierModel):
"""
User created prediction model. The class needs to override three necessary
functions, predict(), train(), fit(). The class inherits ModelHandler which
has its own DataHandler where data is held, saved, loaded, and managed.
"""
def fit(self, data_dictionary: Dict, dk: FreqaiDataKitchen, **kwargs) -> Any:
"""
User sets up the training and test data to fit their desired model here
:params:
:data_dictionary: the dictionary constructed by DataHandler to hold
all the training and test data/labels.
"""
X = data_dictionary["train_features"].to_numpy()
y = data_dictionary["train_labels"].to_numpy()[:, 0]
le = LabelEncoder()
if not is_integer_dtype(y):
y = pd.Series(le.fit_transform(y), dtype="int64")
if self.freqai_info.get('data_split_parameters', {}).get('test_size', 0.1) == 0:
eval_set = None
else:
test_features = data_dictionary["test_features"].to_numpy()
test_labels = data_dictionary["test_labels"].to_numpy()[:, 0]
if not is_integer_dtype(test_labels):
test_labels = pd.Series(le.transform(test_labels), dtype="int64")
eval_set = [(test_features, test_labels)]
train_weights = data_dictionary["train_weights"]
init_model = self.get_init_model(dk.pair)
model = XGBRFClassifier(**self.model_training_parameters)
model.fit(X=X, y=y, eval_set=eval_set, sample_weight=train_weights,
xgb_model=init_model)
return model
def predict(
self, unfiltered_df: DataFrame, dk: FreqaiDataKitchen, **kwargs
) -> Tuple[DataFrame, npt.NDArray[np.int_]]:
"""
Filter the prediction features data and predict with it.
:param: unfiltered_df: Full dataframe for the current backtest period.
:return:
:pred_df: dataframe containing the 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)
"""
(pred_df, dk.do_predict) = super().predict(unfiltered_df, dk, **kwargs)
le = LabelEncoder()
label = dk.label_list[0]
labels_before = list(dk.data['labels_std'].keys())
labels_after = le.fit_transform(labels_before).tolist()
pred_df[label] = le.inverse_transform(pred_df[label])
pred_df = pred_df.rename(
columns={labels_after[i]: labels_before[i] for i in range(len(labels_before))})
return (pred_df, dk.do_predict)

View File

@@ -0,0 +1,45 @@
import logging
from typing import Any, Dict
from xgboost import XGBRFRegressor
from freqtrade.freqai.base_models.BaseRegressionModel import BaseRegressionModel
from freqtrade.freqai.data_kitchen import FreqaiDataKitchen
logger = logging.getLogger(__name__)
class XGBoostRFRegressor(BaseRegressionModel):
"""
User created prediction model. The class needs to override three necessary
functions, predict(), train(), fit(). The class inherits ModelHandler which
has its own DataHandler where data is held, saved, loaded, and managed.
"""
def fit(self, data_dictionary: Dict, dk: FreqaiDataKitchen, **kwargs) -> Any:
"""
User sets up the training and test data to fit their desired model here
:param data_dictionary: the dictionary constructed by DataHandler to hold
all the training and test data/labels.
"""
X = data_dictionary["train_features"]
y = data_dictionary["train_labels"]
if self.freqai_info.get("data_split_parameters", {}).get("test_size", 0.1) == 0:
eval_set = None
else:
eval_set = [(data_dictionary["test_features"], data_dictionary["test_labels"])]
eval_weights = [data_dictionary['test_weights']]
sample_weight = data_dictionary["train_weights"]
xgb_model = self.get_init_model(dk.pair)
model = XGBRFRegressor(**self.model_training_parameters)
model.fit(X=X, y=y, sample_weight=sample_weight, eval_set=eval_set,
sample_weight_eval_set=eval_weights, xgb_model=xgb_model)
return model

View File

@@ -155,6 +155,8 @@ class Backtesting:
self.trading_mode: TradingMode = config.get('trading_mode', TradingMode.SPOT)
# strategies which define "can_short=True" will fail to load in Spot mode.
self._can_short = self.trading_mode != TradingMode.SPOT
self._position_stacking: bool = self.config.get('position_stacking', False)
self.enable_protections: bool = self.config.get('enable_protections', False)
self.init_backtest()
@@ -923,30 +925,23 @@ class Backtesting:
return trade
def handle_left_open(self, open_trades: Dict[str, List[LocalTrade]],
data: Dict[str, List[Tuple]]) -> List[LocalTrade]:
data: Dict[str, List[Tuple]]) -> None:
"""
Handling of left open trades at the end of backtesting
"""
trades = []
for pair in open_trades.keys():
if len(open_trades[pair]) > 0:
for trade in open_trades[pair]:
if trade.open_order_id and trade.nr_of_successful_entries == 0:
# Ignore trade if entry-order did not fill yet
continue
exit_row = data[pair][-1]
self._exit_trade(trade, exit_row, exit_row[OPEN_IDX], trade.amount)
trade.orders[-1].close_bt_order(exit_row[DATE_IDX].to_pydatetime(), trade)
for trade in list(open_trades[pair]):
if trade.open_order_id and trade.nr_of_successful_entries == 0:
# Ignore trade if entry-order did not fill yet
continue
exit_row = data[pair][-1]
self._exit_trade(trade, exit_row, exit_row[OPEN_IDX], trade.amount)
trade.orders[-1].close_bt_order(exit_row[DATE_IDX].to_pydatetime(), trade)
trade.close_date = exit_row[DATE_IDX].to_pydatetime()
trade.exit_reason = ExitType.FORCE_EXIT.value
trade.close(exit_row[OPEN_IDX], show_msg=False)
LocalTrade.close_bt_trade(trade)
# Deepcopy object to have wallets update correctly
trade1 = deepcopy(trade)
trade1.is_open = True
trades.append(trade1)
return trades
trade.close_date = exit_row[DATE_IDX].to_pydatetime()
trade.exit_reason = ExitType.FORCE_EXIT.value
trade.close(exit_row[OPEN_IDX], show_msg=False)
LocalTrade.close_bt_trade(trade)
def trade_slot_available(self, max_open_trades: int, open_trade_count: int) -> bool:
# Always allow trades when max_open_trades is enabled.
@@ -970,9 +965,8 @@ class Backtesting:
return 'short'
return None
def run_protections(
self, enable_protections, pair: str, current_time: datetime, side: LongShort):
if enable_protections:
def run_protections(self, pair: str, current_time: datetime, side: LongShort):
if self.enable_protections:
self.protections.stop_per_pair(pair, current_time, side)
self.protections.global_stop(current_time, side)
@@ -1078,10 +1072,78 @@ class Backtesting:
return None
return row
def backtest(self, processed: Dict, # noqa: max-complexity: 13
def backtest_loop(
self, row: Tuple, pair: str, current_time: datetime, end_date: datetime,
max_open_trades: int, open_trade_count_start: int) -> int:
"""
NOTE: This method is used by Hyperopt at each iteration. Please keep it optimized.
Backtesting processing for one candle/pair.
"""
for t in list(LocalTrade.bt_trades_open_pp[pair]):
# 1. Manage currently open orders of active trades
if self.manage_open_orders(t, current_time, row):
# Close trade
open_trade_count_start -= 1
LocalTrade.remove_bt_trade(t)
self.wallets.update()
# 2. Process entries.
# without positionstacking, we can only have one open trade per pair.
# max_open_trades must be respected
# don't open on the last row
trade_dir = self.check_for_trade_entry(row)
if (
(self._position_stacking or len(LocalTrade.bt_trades_open_pp[pair]) == 0)
and self.trade_slot_available(max_open_trades, open_trade_count_start)
and current_time != end_date
and trade_dir is not None
and not PairLocks.is_pair_locked(pair, row[DATE_IDX], trade_dir)
):
trade = self._enter_trade(pair, row, trade_dir)
if trade:
# TODO: hacky workaround to avoid opening > max_open_trades
# This emulates previous behavior - not sure if this is correct
# Prevents entering if the trade-slot was freed in this candle
open_trade_count_start += 1
# logger.debug(f"{pair} - Emulate creation of new trade: {trade}.")
LocalTrade.add_bt_trade(trade)
self.wallets.update()
for trade in list(LocalTrade.bt_trades_open_pp[pair]):
# 3. Process entry orders.
order = trade.select_order(trade.entry_side, is_open=True)
if order and self._get_order_filled(order.price, row):
order.close_bt_order(current_time, trade)
trade.open_order_id = None
self.wallets.update()
# 4. Create exit orders (if any)
if not trade.open_order_id:
self._get_exit_trade_entry(trade, row) # Place exit order if necessary
# 5. Process exit orders.
order = trade.select_order(trade.exit_side, is_open=True)
if order and self._get_order_filled(order.price, row):
order.close_bt_order(current_time, trade)
trade.open_order_id = None
sub_trade = order.safe_amount_after_fee != trade.amount
if sub_trade:
order.close_bt_order(current_time, trade)
trade.recalc_trade_from_orders()
else:
trade.close_date = current_time
trade.close(order.price, show_msg=False)
# logger.debug(f"{pair} - Backtesting exit {trade}")
LocalTrade.close_bt_trade(trade)
self.wallets.update()
self.run_protections(pair, current_time, trade.trade_direction)
return open_trade_count_start
def backtest(self, processed: Dict,
start_date: datetime, end_date: datetime,
max_open_trades: int = 0, position_stacking: bool = False,
enable_protections: bool = False) -> Dict[str, Any]:
max_open_trades: int = 0) -> Dict[str, Any]:
"""
Implement backtesting functionality
@@ -1094,12 +1156,9 @@ class Backtesting:
:param start_date: backtesting timerange start datetime
:param end_date: backtesting timerange end datetime
:param max_open_trades: maximum number of concurrent trades, <= 0 means unlimited
:param position_stacking: do we allow position stacking?
:param enable_protections: Should protections be enabled?
:return: DataFrame with trades (results of backtesting)
"""
trades: List[LocalTrade] = []
self.prepare_backtest(enable_protections)
self.prepare_backtest(self.enable_protections)
# Ensure wallets are uptodate (important for --strategy-list)
self.wallets.update()
# Use dict of lists with data for performance
@@ -1110,15 +1169,12 @@ class Backtesting:
indexes: Dict = defaultdict(int)
current_time = start_date + timedelta(minutes=self.timeframe_min)
open_trades: Dict[str, List[LocalTrade]] = defaultdict(list)
open_trade_count = 0
self.progress.init_step(BacktestState.BACKTEST, int(
(end_date - start_date) / timedelta(minutes=self.timeframe_min)))
# Loop timerange and get candle for each pair at that point in time
while current_time <= end_date:
open_trade_count_start = open_trade_count
open_trade_count_start = LocalTrade.bt_open_open_trade_count
self.check_abort()
for i, pair in enumerate(data):
row_index = indexes[pair]
@@ -1130,81 +1186,17 @@ class Backtesting:
indexes[pair] = row_index
self.dataprovider._set_dataframe_max_index(row_index)
for t in list(open_trades[pair]):
# 1. Manage currently open orders of active trades
if self.manage_open_orders(t, current_time, row):
# Close trade
open_trade_count -= 1
open_trades[pair].remove(t)
LocalTrade.trades_open.remove(t)
self.wallets.update()
# 2. Process entries.
# without positionstacking, we can only have one open trade per pair.
# max_open_trades must be respected
# don't open on the last row
trade_dir = self.check_for_trade_entry(row)
if (
(position_stacking or len(open_trades[pair]) == 0)
and self.trade_slot_available(max_open_trades, open_trade_count_start)
and current_time != end_date
and trade_dir is not None
and not PairLocks.is_pair_locked(pair, row[DATE_IDX], trade_dir)
):
trade = self._enter_trade(pair, row, trade_dir)
if trade:
# TODO: hacky workaround to avoid opening > max_open_trades
# This emulates previous behavior - not sure if this is correct
# Prevents entering if the trade-slot was freed in this candle
open_trade_count_start += 1
open_trade_count += 1
# logger.debug(f"{pair} - Emulate creation of new trade: {trade}.")
open_trades[pair].append(trade)
LocalTrade.add_bt_trade(trade)
self.wallets.update()
for trade in list(open_trades[pair]):
# 3. Process entry orders.
order = trade.select_order(trade.entry_side, is_open=True)
if order and self._get_order_filled(order.price, row):
order.close_bt_order(current_time, trade)
trade.open_order_id = None
self.wallets.update()
# 4. Create exit orders (if any)
if not trade.open_order_id:
self._get_exit_trade_entry(trade, row) # Place exit order if necessary
# 5. Process exit orders.
order = trade.select_order(trade.exit_side, is_open=True)
if order and self._get_order_filled(order.price, row):
order.close_bt_order(current_time, trade)
trade.open_order_id = None
sub_trade = order.safe_amount_after_fee != trade.amount
if sub_trade:
order.close_bt_order(current_time, trade)
trade.recalc_trade_from_orders()
else:
trade.close_date = current_time
trade.close(order.price, show_msg=False)
# logger.debug(f"{pair} - Backtesting exit {trade}")
open_trade_count -= 1
open_trades[pair].remove(trade)
LocalTrade.close_bt_trade(trade)
trades.append(trade)
self.wallets.update()
self.run_protections(
enable_protections, pair, current_time, trade.trade_direction)
open_trade_count_start = self.backtest_loop(
row, pair, current_time, end_date, max_open_trades, open_trade_count_start)
# Move time one configured time_interval ahead.
self.progress.increment()
current_time += timedelta(minutes=self.timeframe_min)
trades += self.handle_left_open(open_trades, data=data)
self.handle_left_open(LocalTrade.bt_trades_open_pp, data=data)
self.wallets.update()
results = trade_list_to_dataframe(trades)
results = trade_list_to_dataframe(LocalTrade.trades)
return {
'results': results,
'config': self.strategy.config,
@@ -1257,8 +1249,6 @@ class Backtesting:
start_date=min_date,
end_date=max_date,
max_open_trades=max_open_trades,
position_stacking=self.config.get('position_stacking', False),
enable_protections=self.config.get('enable_protections', False),
)
backtest_end_time = datetime.now(timezone.utc)
results.update({

View File

@@ -122,7 +122,6 @@ class Hyperopt:
else:
logger.debug('Ignoring max_open_trades (--disable-max-market-positions was used) ...')
self.max_open_trades = 0
self.position_stacking = self.config.get('position_stacking', False)
if HyperoptTools.has_space(self.config, 'sell'):
# Make sure use_exit_signal is enabled
@@ -258,6 +257,7 @@ class Hyperopt:
logger.debug("Hyperopt has 'protection' space")
# Enable Protections if protection space is selected.
self.config['enable_protections'] = True
self.backtesting.enable_protections = True
self.protection_space = self.custom_hyperopt.protection_space()
if HyperoptTools.has_space(self.config, 'buy'):
@@ -339,8 +339,6 @@ class Hyperopt:
start_date=self.min_date,
end_date=self.max_date,
max_open_trades=self.max_open_trades,
position_stacking=self.position_stacking,
enable_protections=self.config.get('enable_protections', False),
)
backtest_end_time = datetime.now(timezone.utc)
bt_results.update({

View File

@@ -12,7 +12,7 @@ import tabulate
from colorama import Fore, Style
from pandas import isna, json_normalize
from freqtrade.constants import FTHYPT_FILEVERSION, USERPATH_STRATEGIES, Config
from freqtrade.constants import FTHYPT_FILEVERSION, Config
from freqtrade.enums import HyperoptState
from freqtrade.exceptions import OperationalException
from freqtrade.misc import deep_merge_dicts, round_coin_value, round_dict, safe_value_fallback2
@@ -50,9 +50,8 @@ class HyperoptTools():
Get Strategy-location (filename) from strategy_name
"""
from freqtrade.resolvers.strategy_resolver import StrategyResolver
directory = Path(config.get('strategy_path', config['user_data_dir'] / USERPATH_STRATEGIES))
strategy_objs = StrategyResolver.search_all_objects(
directory, False, config.get('recursive_strategy_search', False))
config, False, config.get('recursive_strategy_search', False))
strategies = [s for s in strategy_objs if s['name'] == strategy_name]
if strategies:
strategy = strategies[0]

View File

@@ -408,10 +408,10 @@ def generate_strategy_stats(pairlist: List[str],
exit_reason_stats = generate_exit_reason_stats(max_open_trades=max_open_trades,
results=results)
left_open_results = generate_pair_metrics(pairlist, stake_currency=stake_currency,
starting_balance=start_balance,
results=results.loc[results['is_open']],
skip_nan=True)
left_open_results = generate_pair_metrics(
pairlist, stake_currency=stake_currency, starting_balance=start_balance,
results=results.loc[results['exit_reason'] == 'force_exit'], skip_nan=True)
daily_stats = generate_daily_stats(results)
trade_stats = generate_trading_stats(results)
best_pair = max([pair for pair in pair_results if pair['key'] != 'TOTAL'],

View File

@@ -2,6 +2,7 @@
This module contains the class to persist trades into SQLite
"""
import logging
from collections import defaultdict
from datetime import datetime, timedelta, timezone
from math import isclose
from typing import Any, Dict, List, Optional
@@ -255,6 +256,9 @@ class LocalTrade():
# Trades container for backtesting
trades: List['LocalTrade'] = []
trades_open: List['LocalTrade'] = []
# Copy of trades_open - but indexed by pair
bt_trades_open_pp: Dict[str, List['LocalTrade']] = defaultdict(list)
bt_open_open_trade_count: int = 0
total_profit: float = 0
realized_profit: float = 0
@@ -538,6 +542,8 @@ class LocalTrade():
"""
LocalTrade.trades = []
LocalTrade.trades_open = []
LocalTrade.bt_trades_open_pp = defaultdict(list)
LocalTrade.bt_open_open_trade_count = 0
LocalTrade.total_profit = 0
def adjust_min_max_rates(self, current_price: float, current_price_low: float) -> None:
@@ -1067,6 +1073,8 @@ class LocalTrade():
@staticmethod
def close_bt_trade(trade):
LocalTrade.trades_open.remove(trade)
LocalTrade.bt_trades_open_pp[trade.pair].remove(trade)
LocalTrade.bt_open_open_trade_count -= 1
LocalTrade.trades.append(trade)
LocalTrade.total_profit += trade.close_profit_abs
@@ -1074,9 +1082,17 @@ class LocalTrade():
def add_bt_trade(trade):
if trade.is_open:
LocalTrade.trades_open.append(trade)
LocalTrade.bt_trades_open_pp[trade.pair].append(trade)
LocalTrade.bt_open_open_trade_count += 1
else:
LocalTrade.trades.append(trade)
@staticmethod
def remove_bt_trade(trade):
LocalTrade.trades_open.remove(trade)
LocalTrade.bt_trades_open_pp[trade.pair].remove(trade)
LocalTrade.bt_open_open_trade_count -= 1
@staticmethod
def get_open_trades() -> List[Any]:
"""
@@ -1092,7 +1108,7 @@ class LocalTrade():
if Trade.use_db:
return Trade.query.filter(Trade.is_open.is_(True)).count()
else:
return len(LocalTrade.trades_open)
return LocalTrade.bt_open_open_trade_count
@staticmethod
def stoploss_reinitialization(desired_stoploss):

View File

@@ -26,6 +26,7 @@ class FreqaiModelResolver(IResolver):
initial_search_path = (
Path(__file__).parent.parent.joinpath("freqai/prediction_models").resolve()
)
extra_path = "freqaimodel_path"
@staticmethod
def load_freqaimodel(config: Config) -> IFreqaiModel:
@@ -50,7 +51,6 @@ class FreqaiModelResolver(IResolver):
freqaimodel_name,
config,
kwargs={"config": config},
extra_dir=config.get("freqaimodel_path"),
)
return freqaimodel

View File

@@ -42,6 +42,8 @@ class IResolver:
object_type_str: str
user_subdir: Optional[str] = None
initial_search_path: Optional[Path]
# Optional config setting containing a path (strategy_path, freqaimodel_path)
extra_path: Optional[str] = None
@classmethod
def build_search_paths(cls, config: Config, user_subdir: Optional[str] = None,
@@ -58,6 +60,9 @@ class IResolver:
for dir in extra_dirs:
abs_paths.insert(0, Path(dir).resolve())
if cls.extra_path and (extra := config.get(cls.extra_path)):
abs_paths.insert(0, Path(extra).resolve())
return abs_paths
@classmethod
@@ -183,9 +188,35 @@ class IResolver:
)
@classmethod
def search_all_objects(cls, directory: Path, enum_failed: bool,
def search_all_objects(cls, config: Config, enum_failed: bool,
recursive: bool = False) -> List[Dict[str, Any]]:
"""
Searches for valid objects
:param config: Config object
:param enum_failed: If True, will return None for modules which fail.
Otherwise, failing modules are skipped.
:param recursive: Recursively walk directory tree searching for strategies
:return: List of dicts containing 'name', 'class' and 'location' entries
"""
result = []
abs_paths = cls.build_search_paths(config, user_subdir=cls.user_subdir)
for path in abs_paths:
result.extend(cls._search_all_objects(path, enum_failed, recursive))
return result
@classmethod
def _build_rel_location(cls, directory: Path, entry: Path) -> str:
builtin = cls.initial_search_path == directory
return f"<builtin>/{entry.relative_to(directory)}" if builtin else str(
entry.relative_to(directory))
@classmethod
def _search_all_objects(
cls, directory: Path, enum_failed: bool, recursive: bool = False,
basedir: Optional[Path] = None) -> List[Dict[str, Any]]:
"""
Searches a directory for valid objects
:param directory: Path to search
:param enum_failed: If True, will return None for modules which fail.
@@ -204,7 +235,8 @@ class IResolver:
and not entry.name.startswith('__')
and not entry.name.startswith('.')
):
objects.extend(cls.search_all_objects(entry, enum_failed, recursive=recursive))
objects.extend(cls._search_all_objects(
entry, enum_failed, recursive, basedir or directory))
# Only consider python files
if entry.suffix != '.py':
logger.debug('Ignoring %s', entry)
@@ -217,5 +249,6 @@ class IResolver:
{'name': obj[0].__name__ if obj is not None else '',
'class': obj[0] if obj is not None else None,
'location': entry,
'location_rel': cls._build_rel_location(basedir or directory, entry),
})
return objects

View File

@@ -30,6 +30,7 @@ class StrategyResolver(IResolver):
object_type_str = "Strategy"
user_subdir = USERPATH_STRATEGIES
initial_search_path = None
extra_path = "strategy_path"
@staticmethod
def load_strategy(config: Config = None) -> IStrategy:

View File

@@ -89,6 +89,7 @@ async def api_start_backtest(bt_settings: BacktestRequest, background_tasks: Bac
lastconfig['enable_protections'] = btconfig.get('enable_protections')
lastconfig['dry_run_wallet'] = btconfig.get('dry_run_wallet')
ApiServer._bt.enable_protections = btconfig.get('enable_protections', False)
ApiServer._bt.strategylist = [strat]
ApiServer._bt.results = {}
ApiServer._bt.load_prior_backtest()

View File

@@ -1,13 +1,11 @@
import logging
from copy import deepcopy
from pathlib import Path
from typing import List, Optional
from fastapi import APIRouter, Depends, Query
from fastapi.exceptions import HTTPException
from freqtrade import __version__
from freqtrade.constants import USERPATH_STRATEGIES
from freqtrade.data.history import get_datahandler
from freqtrade.enums import CandleType, TradingMode
from freqtrade.exceptions import OperationalException
@@ -253,11 +251,9 @@ def plot_config(rpc: RPC = Depends(get_rpc)):
@router.get('/strategies', response_model=StrategyListResponse, tags=['strategy'])
def list_strategies(config=Depends(get_config)):
directory = Path(config.get(
'strategy_path', config['user_data_dir'] / USERPATH_STRATEGIES))
from freqtrade.resolvers.strategy_resolver import StrategyResolver
strategies = StrategyResolver.search_all_objects(
directory, False, config.get('recursive_strategy_search', False))
config, False, config.get('recursive_strategy_search', False))
strategies = sorted(strategies, key=lambda x: x['name'])
return {'strategies': [x['name'] for x in strategies]}

View File

@@ -1,3 +1,4 @@
import asyncio
import logging
from typing import Any, Dict
@@ -89,6 +90,8 @@ async def _process_consumer_request(
for _, message in analyzed_df.items():
response = WSAnalyzedDFMessage(data=message)
await channel.send(response.dict(exclude_none=True))
# Throttle the messages to 50/s
await asyncio.sleep(0.02)
@router.websocket("/message/ws")

View File

@@ -198,6 +198,10 @@ class ApiServer(RPCHandler):
logger.debug(f"Found message of type: {message.get('type')}")
# Broadcast it
await self._ws_channel_manager.broadcast(message)
# Limit messages per sec.
# Could cause problems with queue size if too low, and
# problems with network traffik if too high.
await asyncio.sleep(0.001)
except asyncio.CancelledError:
pass