-
Notifications
You must be signed in to change notification settings - Fork 11
/
get_cum_rewards.py
187 lines (158 loc) · 7.32 KB
/
get_cum_rewards.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
# -*- coding: utf-8 -*-
import os, pdb
import pandas as pd
import numpy as np
import gin
gin.enter_interactive_mode()
from joblib import Parallel, delayed
from utils.test import Out_sample_vs_gp
from utils.env import MarketEnv,RealMarketEnv, CashMarketEnv, ShortCashMarketEnv, MultiAssetCashMarketEnv, ShortMultiAssetCashMarketEnv
from agents.PPO import PPO
from utils.spaces import ActionSpace, ResActionSpace
from utils.common import readConfigYaml
from utils.plot import load_PPOmodel
import warnings
warnings.filterwarnings("ignore", category=DeprecationWarning)
# Read config ----------------------------------------------------------------
p = readConfigYaml(os.path.join(os.getcwd(), "config", "paramMultiTestOOS.yaml"))
def get_exp_length(modelpath):
# get the latest created folder "length"
all_subdirs = [
os.path.join(modelpath, d)
for d in os.listdir(modelpath)
if os.path.isdir(os.path.join(modelpath, d))
]
latest_subdir = max(all_subdirs, key=os.path.getmtime)
length = os.path.split(latest_subdir)[-1]
return length
def parallel_test(seed,test_class,train_agent,data_dir):
gin.parse_config_file(os.path.join(data_dir, "config.gin"), skip_unknown=True)
test_class.rnd_state = seed
# TODO temp
# gin.bind_parameter('%T_STUD', True)
gin.bind_parameter('%UNIVERSAL_TRAIN', True)
res_df = test_class.run_test(train_agent, return_output=True)
return res_df
N = 50
model_to_change = None #'202305120_GP_scratch_pt'
models_experiments = [('20230717_real_boot_universal_train_False_split_pct_0.8_rho_boot_0.4','universal_train_False_split_pct_0.8_rho_boot_0.4_seed_635')]
for me in models_experiments:
model = me[0]
experiment = me[1]
# Load parameters and get the path
query = gin.query_parameter
outputClass = p["outputClass"]
tag = p["algo"]
seed = p["seed"]
modelpath = "outputs/{}/{}".format(outputClass, model)
length = get_exp_length(modelpath)
data_dir = "outputs/{}/{}/{}/{}".format(outputClass, model, length, experiment)
gin.parse_config_file(os.path.join(data_dir, "config.gin"), skip_unknown=True)
p['N_test'] = gin.query_parameter('%LEN_SERIES')
rng = np.random.RandomState(query("%SEED"))
# Load the elements for producing the plot
if query("%MV_RES"):
action_space = ResActionSpace()
else:
action_space = ActionSpace()
if gin.query_parameter('%MULTIASSET'):
n_assets = len(gin.query_parameter('%HALFLIFE'))
n_factors = len(gin.query_parameter('%HALFLIFE')[0])
inputs = gin.query_parameter('%INPUTS')
if query("%INP_TYPE") == "f" or query("%INP_TYPE") == "alpha_f":
if 'sigma' in inputs and 'corr' in inputs:
input_shape = (int(n_factors*n_assets+1+ (n_assets**2 - n_assets)/2+n_assets+1),1)
else:
input_shape = (int(n_factors*n_assets+n_assets+1),1)
else:
if 'sigma' in inputs and 'corr' in inputs:
input_shape = (int(n_factors*n_assets+1+ (n_assets**2 - n_assets)/2+n_assets+1),1)
else:
input_shape = (int(n_factors*n_assets+n_assets+1),1)
else:
if query("%INP_TYPE") == "f" or query("%INP_TYPE") == "alpha_f":
if query("%TIME_DEPENDENT"):
input_shape = (len(query('%F_PARAM')) + 2,)
else:
input_shape = (len(query('%F_PARAM')) + 1,)
else:
# if query("%RHO_BOOT"):
# input_shape = (4,)
# else:
# input_shape = (3,)
input_shape = (3,)
train_agent = PPO(
input_shape=input_shape, action_space=action_space, rng=rng
)
if query('%LOAD_PRETRAINED_PATH'):
p['ep_ppo'] = None
else:
p['ep_ppo'] = 'best'
p['ep_ppo'] = 3000
if p['ep_ppo']:
train_agent.model = load_PPOmodel(data_dir, p['ep_ppo'], model=train_agent.model)
else:
train_agent.model = load_PPOmodel(data_dir, gin.query_parameter("%EPISODES"), model=train_agent.model)
if gin.query_parameter('%MULTIASSET'):
if 'Short' in str(gin.query_parameter('%ENV_CLS')):
env = ShortMultiAssetCashMarketEnv
else:
env = MultiAssetCashMarketEnv
else:
if query('%EXPERIMENT_TYPE') == 'Real':
env = RealMarketEnv
else:
env = MarketEnv
oos_test = Out_sample_vs_gp(
savedpath=None,
tag=tag[0],
experiment_type=query("%EXPERIMENT_TYPE"),
env_cls=env,
MV_res=query("%MV_RES"),
N_test=p['N_test'],
mv_solution=True
)
rng_seeds = np.random.RandomState(476)
seeds = rng_seeds.choice(100000,N)
# oos_test.rnd_state = 120
# res_df = oos_test.run_test(train_agent, return_output=True)
if query('%EXPERIMENT_TYPE') == 'Real':
pass
rewards = []
data = pd.read_csv('data/{}.csv'.format(gin.query_parameter('load_real_data.datafile')),index_col=0)
symbols = data.columns.get_level_values(0).unique()[1:]
for s in symbols:
gin.parse_config_file(os.path.join(data_dir, "config.gin"), skip_unknown=True)
oos_test.rnd_state = 34
gin.bind_parameter('%UNIVERSAL_TRAIN', True)
gin.bind_parameter('load_real_data.symbol',s)
res_df = oos_test.run_test(train_agent, return_output=True)
print(res_df['Reward_PPO'].cumsum().iloc[-1])
rewards.append(res_df)
else:
rewards = Parallel(n_jobs=p['cores'])(delayed(parallel_test)(
s, oos_test,train_agent,data_dir) for s in seeds)
if me[0] == model_to_change:
# create an HDF5 file and store the dataframes in it
with pd.HDFStore('outputs/full_results/{}_modified.h5'.format(model)) as store:
for i, df in enumerate(rewards):
store[f'df_{i+1}'] = df
# rewards_ppo = pd.concat(list(map(list, zip(*rewards)))[0],axis=1)
rewards_ppo = pd.concat([df['Reward_PPO'] for df in rewards], ignore_index=True, axis=1)
rewards_ppo.to_csv('outputs/cumrewards/{}_ppo_modified.csv'.format(model))
rewards_gp = pd.concat([df['OptReward'] for df in rewards], ignore_index=True, axis=1)
rewards_gp.to_csv('outputs/cumrewards/{}_gp_modified.csv'.format(model))
rewards_mv = pd.concat([df['MVReward'] for df in rewards], ignore_index=True, axis=1)
rewards_mv.to_csv('outputs/cumrewards/{}_mv_modified.csv'.format(model))
else:
# create an HDF5 file and store the dataframes in it
with pd.HDFStore('outputs/full_results/{}.h5'.format(model)) as store:
for i, df in enumerate(rewards):
store[f'df_{i+1}'] = df
# rewards_ppo = pd.concat(list(map(list, zip(*rewards)))[0],axis=1)
rewards_ppo = pd.concat([df['Reward_PPO'] for df in rewards], ignore_index=True, axis=1)
rewards_ppo.to_csv('outputs/cumrewards/{}_ppo.csv'.format(model))
rewards_gp = pd.concat([df['OptReward'] for df in rewards], ignore_index=True, axis=1)
rewards_gp.to_csv('outputs/cumrewards/{}_gp.csv'.format(model))
rewards_mv = pd.concat([df['MVReward'] for df in rewards], ignore_index=True, axis=1)
rewards_mv.to_csv('outputs/cumrewards/{}_mv.csv'.format(model))