siamese_ged/test_learn.py at master · priba/siamese_ged · 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
import os, sys

# path

path_data = "/media/priba/PPAP/NeuralMessagePassing/data/IAM/Letter/"
path_experiments = "/media/priba/PPAP/SwitzerlandStay/nmp_ged/AndreasResults/experiments/"

# settings

dbs = ["LOW", "MED", "HIGH"]
layers = ["2", "3"]
edges = ["adj", "feat"]
hstates = ["2", "64"]

runs = 5
epochs = "300"
batch_size = "128"
lr = str(1e-2)
pipeline = "learn_representation"
distance = "Hd"

# params

params = []
for db in dbs:
    for edge in edges:
        for layer in layers:
            for hstate in hstates:
                for run in range(0, runs):
                    params.append({"db": db, "edge": edge, "layer": layer, "hstate": hstate, "run": str(run)})

# main

if len(sys.argv) <= 1:
    cnt = 0
    for param in params:
        print cnt, ":", param
        cnt += 1
    print "params: ", cnt
else:
    sge_task_id = int(sys.argv[1])
    param = params[sge_task_id - 1]
    db = param["db"]
    edge = param["edge"]
    layer = param["layer"]
    hstate = param["hstate"]
    run = param["run"]

    dir_db = path_data + db + "/"

    dir_param_old = pipeline + "/" + db + "/" + edge + "_SoftHd_l" + layer + "_h" + hstate + "/r" + run + "/"
    dir_param = pipeline + "/" + db + "/" + edge + "_" + distance + "_l" + layer + "_h" + hstate + "/r" + run + "/"

    dir_run = path_experiments + "run/" + dir_param
    dir_checkpoint = path_experiments + "checkpoint/" + dir_param_old + 'checkpoint.pth'
    dir_log = path_experiments + "log/" + dir_param
    if not os.path.isdir(dir_run):
        os.makedirs(dir_run)

    print "train NMP: " + dir_run + " .."
    cmd = "python train_learn_representation.py " + dir_db + " letters -t -l " + dir_checkpoint + " --log " + dir_log + " -lr " + lr + " --nlayers " + layer + " --hidden_size " + hstate + " -e " + epochs + " -b " + batch_size + " --representation " + edge + " --schedule 100 200 250 --distance " + distance + " > " + dir_run + "run.txt"
    print cmd
    os.system(cmd)
    print ".. done."