-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtdvi.py
204 lines (169 loc) · 10.1 KB
/
tdvi.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
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
########################################################################################################################
# IMPORT #
########################################################################################################################
import torch
import sys
import os
import json
import time
import argparse
from umap.umap_ import find_ab_params
from singleVis.vis_models import vis_models as vmodels
from singleVis.losses import UmapLoss, ReconstructionLoss, SingleVisLoss, TemporalEdgeLoss, splittDVILoss
from singleVis.edge_dataset import DataHandler, DVIDataHandler, SplitTemporalDataHandler, create_dataloader
from singleVis.trainer import SingleVisTrainer, SplitTemporalTrainer
from singleVis.subsampling import DensityAwareSampling
from singleVis.data import NormalDataProvider
from singleVis.spatial_edge_constructor import SplitSpatialTemporalEdgeConstructor, SingleEpochSpatialEdgeConstructor
from singleVis.projector import DVIProjector
from singleVis.eval.evaluator import Evaluator
from singleVis.visualizer import visualizer
from config import load_cfg
'''
TODO:
1. weight in umap and temporal loss
2. density estimation
4. different negative sampling rate for umap and temporal loss
'''
########################################################################################################################
# DVI PARAMETERS #
########################################################################################################################
# """DVI with semantic temporal edges"""
VIS_METHOD = "tdvi"
########################################################################################################################
# LOAD PARAMETERS #
########################################################################################################################
parser = argparse.ArgumentParser(description='Process hyperparameters...')
parser.add_argument('--content_path', '-c', type=str)
parser.add_argument("--iteration", '-i', type=int)
parser.add_argument("--resume", '-r', type=int, default=-1)
args = parser.parse_args()
########################################################################################################################
# SETTING PARAMETERS #
########################################################################################################################
CONTENT_PATH = args.content_path
ITERATION = args.iteration
RESUME = args.resume
sys.path.append(CONTENT_PATH)
config = load_cfg(os.path.join(CONTENT_PATH, "config", f"{VIS_METHOD}.yaml"))
print(config)
SETTING = config.SETTING
CLASSES = config.CLASSES
DATASET = config.DATASET
PREPROCESS = config.VISUALIZATION.PREPROCESS
GPU_ID = config.GPU
EPOCH_START = config.EPOCH_START
EPOCH_END = config.EPOCH_END
EPOCH_PERIOD = config.EPOCH_PERIOD
EPOCH_NAME = config.EPOCH_NAME
# Training parameter (subject model)
TRAINING_PARAMETER = config.TRAINING
NET = TRAINING_PARAMETER.NET
LEN = TRAINING_PARAMETER.train_num
# Training parameter (visualization model)
VISUALIZATION_PARAMETER = config.VISUALIZATION
SAVE_BATCH_SIZE = VISUALIZATION_PARAMETER.SAVE_BATCH_SIZE
LAMBDA = VISUALIZATION_PARAMETER.LAMBDA
B_N_EPOCHS = VISUALIZATION_PARAMETER.BOUNDARY.B_N_EPOCHS
L_BOUND = VISUALIZATION_PARAMETER.BOUNDARY.L_BOUND
ENCODER_DIMS = VISUALIZATION_PARAMETER.ENCODER_DIMS
DECODER_DIMS = VISUALIZATION_PARAMETER.DECODER_DIMS
S_N_EPOCHS = VISUALIZATION_PARAMETER.S_N_EPOCHS
T_N_EPOCHS = VISUALIZATION_PARAMETER.T_N_EPOCHS
N_NEIGHBORS = VISUALIZATION_PARAMETER.N_NEIGHBORS
PATIENT = VISUALIZATION_PARAMETER.PATIENT
MAX_EPOCH = VISUALIZATION_PARAMETER.MAX_EPOCH
VIS_MODEL = VISUALIZATION_PARAMETER.VIS_MODEL
METRIC = VISUALIZATION_PARAMETER.METRIC
VIS_MODEL_NAME = f"{VIS_METHOD}"
EVALUATION_NAME = f"evaluation_{VIS_MODEL_NAME}"
# Define hyperparameters
DEVICE = torch.device("cuda:{}".format(GPU_ID) if torch.cuda.is_available() else "cpu")
import Model.model as subject_model
net = eval("subject_model.{}()".format(NET))
########################################################################################################################
# TRAINING SETTING #
########################################################################################################################
# Define data_provider
data_provider = NormalDataProvider(CONTENT_PATH, net, EPOCH_START, EPOCH_END, EPOCH_PERIOD, device=DEVICE, classes=CLASSES, epoch_name=EPOCH_NAME, verbose=1)
if PREPROCESS:
data_provider._meta_data_single(ITERATION, batch_size=SAVE_BATCH_SIZE)
if B_N_EPOCHS >0:
data_provider._estimate_boundary_single(LEN//10, l_bound=L_BOUND, n_epoch=ITERATION, batch_size=SAVE_BATCH_SIZE)
# Define visualization models
model = vmodels[VIS_MODEL](ENCODER_DIMS, DECODER_DIMS)
# Define Losses
negative_sample_rate = 5
min_dist = .1
_a, _b = find_ab_params(1.0, min_dist)
umap_loss_fn = UmapLoss(negative_sample_rate, _a, _b, repulsion_strength=1.0)
recon_loss_fn = ReconstructionLoss(beta=1.0)
temporal_loss_fn = TemporalEdgeLoss(negative_sample_rate, _a, _b, repulsion_strength=1.0)
single_loss_fn = SingleVisLoss(umap_loss_fn, recon_loss_fn, lambd=LAMBDA)
# Define Projector
projector = DVIProjector(vis_model=model, content_path=CONTENT_PATH, vis_model_name=VIS_MODEL_NAME, epoch_name=EPOCH_NAME, device=DEVICE, verbose=0)
# Define Visualizer
vis = visualizer(data_provider, projector, 200, "tab10")
# Define Evaluator
evaluator = Evaluator(data_provider, projector, metric=METRIC)
# Define training parameters
optimizer = torch.optim.Adam(model.parameters(), lr=.01, weight_decay=1e-5)
lr_scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=4, gamma=.1)
if ITERATION == EPOCH_START:
# Define Edge dataset
t0 = time.time()
sampler = DensityAwareSampling()
spatial_cons = SingleEpochSpatialEdgeConstructor(data_provider, EPOCH_START, S_N_EPOCHS, B_N_EPOCHS, N_NEIGHBORS, metric="euclidean", sampler=sampler)
edge_to, edge_from, probs, feature_vectors, attention = spatial_cons.construct()
# Construct two dataset and train on them separately
dataset = DVIDataHandler(edge_to, edge_from, feature_vectors, attention)
edge_loader = create_dataloader(dataset, S_N_EPOCHS, probs, len(edge_to))
# train
trainer = SingleVisTrainer(model, single_loss_fn, optimizer, lr_scheduler, edge_loader=edge_loader, DEVICE=DEVICE)
train_epoch, time_spent = trainer.train(PATIENT, MAX_EPOCH)
save_dir = os.path.join(data_provider.model_path, f"{EPOCH_NAME}_{EPOCH_START}")
trainer.save(save_dir=save_dir, file_name="{}".format(VIS_MODEL_NAME))
trainer.record_time(data_provider.model_path, "time_{}".format(VIS_MODEL_NAME), "training", EPOCH_START, (train_epoch, time_spent))
trainer.log(data_provider.content_path, EPOCH_START)
vis.savefig(EPOCH_START, f"{VIS_METHOD}_{EPOCH_START}.png")
evaluator.save_epoch_eval(EPOCH_START, 15, temporal_k=5, file_name="{}".format(EVALUATION_NAME))
else:
# load resume epoch/iteration
if RESUME < 0:
log_path = os.path.join(CONTENT_PATH, "log.json")
with open(log_path, "r") as f:
curr_log = json.load(f)
curr_log.sort()
RESUME = curr_log[-1]
print(f"Resuming from {RESUME} iterations...")
projector.load(RESUME)
# Define Criterion
criterion = splittDVILoss(umap_loss_fn, recon_loss_fn, temporal_loss_fn)
# Define training parameters
optimizer = torch.optim.Adam(model.parameters(), lr=.01, weight_decay=1e-5)
lr_scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=4, gamma=.1)
# Define Edge dataset
sampler = DensityAwareSampling()
spatial_cons = SplitSpatialTemporalEdgeConstructor(data_provider, projector, S_N_EPOCHS, B_N_EPOCHS, T_N_EPOCHS, N_NEIGHBORS, metric=METRIC, sampler=sampler)
t0 = time.time()
spatial_component, temporal_component = spatial_cons.construct(ITERATION, RESUME)
edge_to, edge_from, weights, feature_vectors, attention = spatial_component
edge_t_to, edge_t_from, weight_t, next_data, prev_data, prev_embedded, margins = temporal_component
t1 = time.time()
# two dataloaders for spatial and temporal datasets
spatial_dataset = DataHandler(edge_to, edge_from, feature_vectors, attention)
spatial_edge_loader = create_dataloader(spatial_dataset, S_N_EPOCHS, weights, len(edge_to))
temporal_dataset = SplitTemporalDataHandler(edge_t_to, edge_t_from, next_data, prev_data, prev_embedded, margins)
temporal_edge_loader = create_dataloader(temporal_dataset, T_N_EPOCHS, weight_t, len(edge_t_to))
########################################################################################################################
# TRAIN #
########################################################################################################################
trainer = SplitTemporalTrainer(model, criterion, optimizer, lr_scheduler, spatial_edge_loader=spatial_edge_loader, temporal_edge_loader=temporal_edge_loader, DEVICE=DEVICE)
train_epoch, time_spent = trainer.train(PATIENT, MAX_EPOCH)
save_dir = os.path.join(data_provider.model_path, "{}_{}".format(EPOCH_NAME, ITERATION))
trainer.save(save_dir=save_dir, file_name="{}".format(VIS_MODEL_NAME))
trainer.record_time(save_dir=data_provider.model_path, file_name="time_{}".format(VIS_MODEL_NAME), operation="training", iteration=str(ITERATION), t=(train_epoch, time_spent))
trainer.record_time(save_dir=data_provider.model_path, file_name="time_{}".format(VIS_MODEL_NAME), operation="complex", iteration=str(ITERATION), t=(t1-t0))
trainer.log(data_provider.content_path, ITERATION)
vis.savefig(ITERATION, f"{VIS_METHOD}_{ITERATION}.png")
evaluator.save_epoch_eval(ITERATION, 15, temporal_k=5, file_name="{}".format(EVALUATION_NAME))