TextWorld-Neural-Algorithmic-Reasoning/train.py at master · AdeeshKolluru/TextWorld-Neural-Algorithmic-Reasoning · 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
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
"""
Usage:
    train.py [options] [--algorithms=ALGO]...

Options:
    -h --help                        Show this screen.
    --algorithms ALGO                Which algorithms to add. One of {AugmentingPath, BFS}
    --model-name NAME                Specific name of model
    --processor-type PROC            Type of processor. One of {MPNN, PNA, GAT}. [default: MPNN]
"""
from datetime import datetime

import time
import copy
import random

import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import torch_geometric
from torch_geometric.data import DataLoader

import numpy as np
from pprint import pprint
from docopt import docopt
import os

import models
from flow_datasets import SingleIterationDataset
from utils import get_print_info, get_print_format, interrupted, iterate_over, load_algorithms
from hyperparameters import get_hyperparameters


if __name__ == "__main__":

    args = docopt(__doc__)
    hyperparameters = get_hyperparameters()
    DEVICE = hyperparameters["device"]
    DIM_LATENT = hyperparameters["dim_latent"]
    DIM_EDGES = hyperparameters["dim_edges"]
    NAME = args["--model-name"] if args["--model-name"] is not None else datetime.now().strftime("%b-%d-%Y-%H-%M")

    processor = models.AlgorithmProcessor(DIM_LATENT, SingleIterationDataset, args["--processor-type"]).to(DEVICE)
    print("PARAMETERS", sum(p.numel() for p in processor.parameters()))
    print(list((name, p.numel()) for name, p in processor.named_parameters()))
    load_algorithms(args["--algorithms"], processor, True)
    # processor.reset_all_weights()
    params = list(processor.parameters())
    print(DEVICE)
    print(processor)
    augmenting_path_network = None
    for key, algorithm in processor.algorithms.items():
        if type(algorithm) == models.AugmentingPathNetwork:
            augmenting_path_network = algorithm
    print(augmenting_path_network)

    BATCH_SIZE = hyperparameters["batch_size"]
    PATIENCE_LIMIT = hyperparameters["patience_limit"]
    GROWTH_RATE = hyperparameters["growth_rate_sigmoid"]
    SIGMOID_OFFSET = hyperparameters["sigmoid_offset"]

    patience = 0
    last_mean = 0
    last_final = 0
    last_broken = 100
    last_loss = 0*1e9 if augmenting_path_network is not None else 1e9
    cnt = 0

    fmt = get_print_format()

    best_model = models.AlgorithmProcessor(DIM_LATENT, SingleIterationDataset, args["--processor-type"]).to(DEVICE)
    best_model.algorithms = nn.ModuleDict(processor.algorithms.items())
    best_model.load_state_dict(copy.deepcopy(processor.state_dict()))

    torch.set_printoptions(precision=20)

    with torch.autograd.profiler.profile(enabled=False, use_cuda=True) as prof:
        # for algorithm in processor.algorithms:
        #     algorithm.loader = DataLoader(algorithm.train_dataset, batch_size=BATCH_SIZE, shuffle=True, drop_last=False, num_workers=8)
        #     algorithm.val_loader = DataLoader(algorithm.val_dataset, batch_size=BATCH_SIZE, shuffle=False, drop_last=False, num_workers=8)
        optimizer = optim.Adam(params, lr=hyperparameters["lr"], weight_decay=hyperparameters["weight_decay"])
        for epoch in range(3000):# FIXME
            if interrupted():
                break
            # 0.0032
            processor.train()
            iterate_over(processor, optimizer)

            patience += 1
            print('Epoch {:4d}: \n'.format(epoch), end=' ')
            processor.eval()
            iterate_over(processor)
            # print("Mean/Last step acc", processor.algorithms[0].get_validation_accuracies())
            # total_loss = sum(processor.algorithms[0].get_validation_losses()) #TODO PRETTIER!

            if augmenting_path_network is None:
                total_loss = sum(processor.algorithms["BFS"].get_validation_losses())
                if (total_loss) < last_loss:
                    patience = 0
                    last_loss = (total_loss)
                    best_model.load_state_dict(copy.deepcopy(processor.state_dict()))
                print("Total Loss:", total_loss, "Patience:", patience)

            if augmenting_path_network is not None: # TODO prettier code plz
                (total_loss_dist, total_loss_pred, total_loss_term, find_min, total_loss,
                 mean_step_acc, final_step_acc, tnr, subtract_acc, broken_invariants,
                 broken_reachabilities, broken_flows, broken_all, len_broken) =\
                         get_print_info(processor.algorithms["AugmentingPath"])

                if get_hyperparameters()["calculate_termination_statistics"]: #DEPRECATED
                    print("Term precision:",
                            augmenting_path_network.true_positive/(augmenting_path_network.true_positive+augmenting_path_network.false_positive)
                            if
                            augmenting_path_network.true_positive+augmenting_path_network.false_positive
                            else 'N/A')
                    print("Term recall:",
                            augmenting_path_network.true_positive/(augmenting_path_network.true_positive+augmenting_path_network.false_negative)
                            if
                            augmenting_path_network.true_positive+augmenting_path_network.false_negative
                            else 'N/A')

                if (final_step_acc) > last_loss:
                    patience = 0
                    last_loss = (final_step_acc)
                    best_model.load_state_dict(copy.deepcopy(processor.state_dict()))

                print(fmt.format(
                    mean_step_acc,
                    final_step_acc,
                    tnr,
                    subtract_acc,
                    total_loss_dist,
                    total_loss_pred,
                    total_loss_term,
                    find_min,
                    total_loss,
                    sum(broken_invariants),
                    len_broken,
                    sum(broken_all),
                    len_broken,
                    sum(broken_reachabilities),
                    len_broken,
                    sum(broken_flows),
                    len_broken,
                    patience))

            os.makedirs(f"checkpoints/{NAME}", exist_ok=True)
            torch.save(processor.state_dict(), f'checkpoints/{NAME}/test_{NAME}_epoch_'+str(epoch)+'.pt')

            if patience >= PATIENCE_LIMIT:
                break

    torch.save(best_model.state_dict(), f'checkpoints/{NAME}/best_{NAME}.pt')