learning-sidarthe/run_momentum_false_by_train_size_experiments.py at master · sailab-code/learning-sidarthe · GitHub

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
import os
import random

from params.params import SidartheParamGenerator
from populations import populations

from learning_models.sidarthe_extended import SidartheExtended
from experiments.sidarthe_experiment import SidartheExperiment
from experiments.sidarthe_extended_experiment import ExtendedSidartheExperiment

import multiprocessing as mp
import numpy as np
import itertools
import json

from utils.multiprocess_utils import ProcessPool

N_PROCESSES = 10
experiment_cls = ExtendedSidartheExperiment

os.environ['CUDA_VISIBLE_DEVICES'] = '-1'  # CHOOSE GPU HERE
if __name__ == "__main__":
    region = "Italy"
    t_step = 1.0
    train_size = 120
    val_len = 40

    der_1st_reg = 1e8

    n_epochs = 2000

    procs = []
    process_pool = ProcessPool(N_PROCESSES)
    mp.set_start_method('spawn')

    train_sizes = range(40, 120, 20)
    params_path = os.path.join(os.getcwd(), "runs", "momentum_train_size_exps", "sidarthe_extended", "Italy",
                               "initial_params.json")

    with open(params_path) as f:
        params_list = json.load(f)

    for initial_params in params_list:
        gen = SidartheParamGenerator()
        gen.init_from_base_params(initial_params)
        for train_size in train_sizes:
            experiment = experiment_cls(region, n_epochs=n_epochs, time_step=t_step,
                                        runs_directory="runs/momentum_false_train_size_exps", uuid_prefix=None)
            gen.extend(train_size)

            process_pool.start(target=experiment.run_exp,
                               kwargs={
                                   "initial_params": gen.params,
                                   "dataset_params": {"train_size": train_size, "val_len": val_len, "region": region},
                                   "model_params": {"der_1st_reg": der_1st_reg, "bound_loss_type": "step"},
                                   "train_params": {"momentum": False},
                               })

            process_pool.wait_for_empty_slot(timeout=5)

    process_pool.wait_for_all()