-
Notifications
You must be signed in to change notification settings - Fork 17
/
Copy pathla_runs.py
151 lines (133 loc) · 5.72 KB
/
la_runs.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
'''Bla bla
'''
import pandas as pd
import numpy as np
from numpy.random import shuffle
from sklearn.preprocessing import normalize
from pathlib import Path
from torchvision.utils import save_image
from sklearn.cluster import KMeans
from img_transforms import UnNormalizeTransform
from la_learner import LALearner
uu = UnNormalizeTransform()
eval_clusterer = KMeans(n_clusters=20)
def get_learner_1():
learner_1 = LALearner(run_label='simple test run',
raw_csv_toc='../../Desktop/Fungi/toc_full.csv', raw_csv_root='../../Desktop/Fungi',
loader_batch_size=64, iselector=[0,1,2,3,4,5,6,7,8,9],
dataset_type='grid basic idx',
lr_init=0.01,
temperature=0.07, k_nearest_neighbours=100, clustering_repeats=5, number_of_centroids=100,
memory_mixing=0.5, n_samples=360)
return learner_1
def get_learner_2():
v1 = list(range(759))
#v2 = list(range(759, 2429))
#shuffle(v1)
#shuffle(v2)
#vv = v1[:50] + v2[:100]
vv = v1[:50]
learner_2 = LALearner(run_label='simple test run',
raw_csv_toc='../../Desktop/Fungi/toc_full.csv', raw_csv_root='../../Desktop/Fungi',
loader_batch_size=128, iselector=vv,
dataset_type='grid basic idx',
selector=pd.IndexSlice[:, :, :, :, :, ['Cantharellaceae'], :, :, :],
lr_init=0.03, scheduler_step_size=5, scheduler_gamma=0.3,
temperature=0.07, k_nearest_neighbours=100, clustering_repeats=6, number_of_centroids=100,
memory_mixing=0.5, n_samples=1800)
return learner_2
def get_learner_2x():
v1 = list(range(759))
v2 = list(range(759, 2429))
shuffle(v1)
shuffle(v2)
vv = v1[:50] + v2[:150]
#vv = v1[:100]
learner_2 = LALearner(run_label='simple test run',
raw_csv_toc='../../Desktop/Fungi/toc_full.csv', raw_csv_root='../../Desktop/Fungi',
loader_batch_size=16, iselector=vv,
dataset_type='full basic idx',
selector=pd.IndexSlice[:, :, :, :, :, ['Cantharellaceae','Amanitaceae'], :, :, :],
lr_init=0.03, scheduler_step_size=5, scheduler_gamma=0.3,
temperature=0.07, k_nearest_neighbours=199, clustering_repeats=6, number_of_centroids=20,
memory_mixing=0.5, n_samples=200)
return learner_2
def get_learner_3():
v1 = list(range(759))
v2 = list(range(759, 2429))
shuffle(v1)
shuffle(v2)
vv = v1[:100] + v2[:300]
learner = LALearner(run_label='simple test run',
raw_csv_toc='../../Desktop/Fungi/toc_full.csv', raw_csv_root='../../Desktop/Fungi',
loader_batch_size=128, iselector=vv,
selector=pd.IndexSlice[:, :, :, :, :, ['Cantharellaceae', 'Amanitaceae'], :, :, :],
lr_init=0.03, scheduler_step_size=5, scheduler_gamma=0.3,
temperature=0.07, k_nearest_neighbours=1000, clustering_repeats=6, number_of_centroids=2000,
memory_mixing=0.5, n_samples=14400)
return learner
def saver_func(dloader, c_labels, root_dir, img_key):
Path(root_dir).mkdir(parents=False, exist_ok=False)
n_img = 0
for data in dloader:
image_tensor = data[img_key]
for image in image_tensor:
img_ = uu(image)
the_cluster = c_labels[n_img]
subfolder = '{}/cluster_{}'.format(root_dir, the_cluster)
Path(subfolder).mkdir(parents=False, exist_ok=True)
save_image(img_, '{}/{}.png'.format(subfolder, n_img))
n_img += 1
def train_simple():
learner_1 = get_learner_1()
learner_1.load_model('ae_learner_2_bigger')
learner_1.train(4)
learner_1.save_model('la_simple')
def eval_simple():
learner_2 = get_learner_2()
learner_2.load_model('la_bigger')
cluster_labels = learner_2.eval(clusterer=eval_clusterer.fit_predict)
saver_func(learner_2.dataloader, cluster_labels, './cluster_imgs', learner_2.dataset.getkeys.image)
def train_bigger():
learner_2 = get_learner_2()
learner_2.load_model('ae_learner_2_bigger')
learner_2.train(20)
learner_2.save_model('la_bigger_2')
def train_biggerx():
learner_2 = get_learner_2x()
learner_2.load_model('ae_kantflue_fullimage_wellconverged')
img_array = []
idx_array = []
for dd in learner_2.dataloader:
image = dd[learner_2.dataset.returnkey.image]
idx = dd[learner_2.dataset.returnkey.idx]
out = learner_2.model(image)
out_npy = normalize(out.detach().numpy(), axis=1)
img_array.append(out_npy)
idx_array.append(idx.detach().numpy())
print (idx.detach().numpy())
print (out_npy.shape)
img_vecs = np.concatenate(img_array, axis=0)
idx_vecs = np.concatenate(idx_array, axis=0)
print (img_vecs.shape)
learner_2.criterion.memory_bank.memory_mixing_rate=1.0
learner_2.criterion.memory_bank.update_memory(img_vecs, idx_vecs)
learner_2.criterion.memory_bank.memory_mixing_rate=0.5
learner_2.train(8)
learner_2.save_model('la_bigger_2x')
def eval_bigger():
learner_2 = get_learner_2x()
learner_2.load_model('la_bigger_2x')
cluster_labels = learner_2.eval(clusterer=eval_clusterer.fit_predict)
saver_func(learner_2.dataloader, cluster_labels, './cluster_imgs', learner_2.dataset.returnkey.image)
def train_verybig():
learner_3 = get_learner_3()
learner_3.load_model('ae_learner_2_bigger')
learner_3.train(20)
learner_3.save_model('la_verybig')
if __name__ == '__main__':
#train_simple()
#eval_simple()
train_biggerx()
eval_bigger()
#train_verybig()