-
Notifications
You must be signed in to change notification settings - Fork 0
/
train.py
323 lines (187 loc) · 11.9 KB
/
train.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
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
import warnings
from utils.mlflow_image_logging import MLFlowImageLogger
from train_val_functions import MapToLossFunction, ship_device, statistics
import network.end2end as network_utils
from typing import List
from tqdm import tqdm
import time
import torch
import torch.nn as nn
import os
import mlflow
import wandb
# ADDITIONAL CODE: get run from the current context
#run = Run.get_context()
class RoboFishNet():
def __init__(self, ver_name, unordered:bool, spatial_data:bool, merfish_data:bool, img_dir_timestamp:str,
groundtruth_emitters_count, scheduler, optimizer_fn, in_networks:int,
in_channels:int, outputs:List, gpus:int, num_tiles:int, batch_size:int,
lr:int, clip_value:int, network_opt:int, mcce_loss_opt:int,
loss_scale:List, checkpoint_dir:str, delta_loss=0.01, epochs=2, print_rate=260):
super().__init__()
self.scheduler = scheduler
self.optimizer_fn = optimizer_fn
self.outputs = outputs
self.epochs = epochs
self.print_rate = print_rate
self.checkpoint_dir = checkpoint_dir
self.lr = lr
self.clip_value = clip_value
if torch.cuda.is_available() and gpus>0:
self.target = torch.cuda.current_device()
else:
self.target = 'cpu'
self.th1 = 0.6
self.th2 = 0.9
self.delta_loss = delta_loss
self.total_classes = outputs
self.batch_size = ship_device(torch.tensor(batch_size), self.target)
self.mcce_loss_opt = mcce_loss_opt
self.loss_scale = loss_scale
self.mlflow_image_logger = MLFlowImageLogger(self.th1, self.th2, self.target)
self.train_eval_step_num = int((num_tiles*0.7)/batch_size)
self.val_eval_step_num = int((num_tiles*0.3)/batch_size)
mlflow.log_params({"optimizer_version_name": ver_name,
"unordered_data": unordered,
"spatial_data": spatial_data,
"merfish_data": merfish_data,
"dataset": img_dir_timestamp,
"groundtruth_emitters_count": groundtruth_emitters_count,
"mcce_loss_opt": mcce_loss_opt,
"learning_rate_mcce": lr,
"epochs": epochs,
"clipping_norm": clip_value,
"postprocess_th1": self.th1,
"postprocess_th2": self.th2,
"mcce_loss_weight": loss_scale,
"output_num": outputs,
"total_classes_count": outputs,
"codebook_gene_count": outputs - 1,
"delta_loss": delta_loss
})
wandb.config.update({"optimizer_version_name": ver_name,
"unordered_data": unordered,
"spatial_data": spatial_data,
"merfish_data": merfish_data,
"dataset": img_dir_timestamp,
"groundtruth_emitters_count": groundtruth_emitters_count,
"mcce_loss_opt": mcce_loss_opt,
"learning_rate_mcce": lr,
"epochs": epochs,
"clipping_norm": clip_value,
"postprocess_th1": self.th1,
"postprocess_th2": self.th2,
"mcce_loss_weight": loss_scale,
"output_num": outputs,
"total_classes_count": outputs,
"codebook_gene_count": outputs - 1,
"delta_loss": delta_loss
})
def __call__(self, model, train_dataloader, val_dataloader):
val_lowest = 99999999
val_seg_lowest = 99999999
val_dec_lowest = 99999999
for e in range(self.epochs):
train_step_count = 0
epoch_loss = 0
epoch_seg_loss = 0
epoch_dec_loss = 0
model.train()
weight_loss_seg, weight_loss_dec = network_utils.get_loss_weights(self.delta_loss, e, self.epochs)
# Always return 0, since we assume gradient_adjustment = True
dec_gradient_multiplier = network_utils.get_dec_gradient_multiplier()
network_utils.set_dec_gradient_multiplier(model, dec_gradient_multiplier)
for x, y in tqdm(train_dataloader):
x, groundtruth_labels_and_barcode = ship_device([x[:,0,:,:,:], y[:,0,:,:,:]], self.target)
groundtruth_labels = groundtruth_labels_and_barcode[:, 0, :, :]
train_step_count += 1
self.optimizer_fn.zero_grad()
classification_map, seg_mask = model(x)
train_loss, train_seg_loss, train_dec_loss = \
MapToLossFunction(self.target, self.mcce_loss_opt, classification_map,
seg_mask, groundtruth_labels_and_barcode, weight_loss_seg, weight_loss_dec)
""" Calling loss backward """
train_loss.backward()
nn.utils.clip_grad_norm_(model.parameters(), max_norm=self.clip_value, norm_type=2)
self.optimizer_fn.step()
epoch_loss += train_loss.item()
epoch_seg_loss += train_seg_loss.item()
epoch_dec_loss += train_dec_loss.item()
"""Print Update and Log image visualizations"""
if train_step_count % self.train_eval_step_num == 0:
train_denom = train_step_count
print('Epoch: {} | Avg Training Loss per {} steps: {}'.format(e, self.train_eval_step_num, epoch_loss/train_denom))
if train_step_count == self.train_eval_step_num:
"""Generate Statistics: Losses, Scores/Accuracy, Emitter Counts at the end of Train Epoch"""
statistics(self.batch_size, classification_map, groundtruth_labels, e, 'train')
if e%10 == 0:
"""Visualize Train Images at the end of Train Epoch"""
self.mlflow_image_logger(classification_map[0], groundtruth_labels[0], self.total_classes, e, 'train')
train_denom = train_step_count
mlflow.log_metric('train_loss', epoch_loss/train_denom)
mlflow.log_metric('train_seg_loss', epoch_seg_loss/train_denom)
mlflow.log_metric('train_dec_loss', epoch_dec_loss/train_denom)
wandb.log({'train_loss': epoch_loss/train_denom,
'epoch': e,
'train_seg_loss': epoch_seg_loss/train_denom,
'train_dec_loss': epoch_dec_loss/train_denom})
""" Turn on model evaluation mode."""
model.eval()
val_step_count = 0
total_val_loss = 0
total_seg_loss = 0
total_dec_loss = 0
avg_ji, avg_pre, avg_rec = 0, 0, 0
for x, y in tqdm(val_dataloader):
x, groundtruth_labels_and_barcode = ship_device([x[:,0,:,:,:], y[:,0,:,:,:]], self.target)
groundtruth_labels = groundtruth_labels_and_barcode[:, 0, :, :]
val_step_count += 1
classification_map, seg_mask = model(x)
val_loss, val_seg_loss, val_dec_loss = \
MapToLossFunction(self.target, self.mcce_loss_opt, classification_map,
seg_mask, groundtruth_labels_and_barcode, weight_loss_seg, weight_loss_dec)
total_val_loss += val_loss.item()
total_seg_loss += val_seg_loss.item()
total_dec_loss += val_dec_loss.item()
if val_step_count == self.val_eval_step_num:
"""Generate Statistics: Losses, Scores/Accuracy, Emitter Counts at the end of Val Epoch"""
avg_ji, avg_pre, avg_rec = statistics(self.batch_size, classification_map, groundtruth_labels, e, 'val')
if e%10 == 0:
"""Visualize Val Images at the end of Val Epoch"""
self.mlflow_image_logger(classification_map[0], groundtruth_labels[0], self.total_classes, e, 'val')
val_denom = val_step_count
val_loss = total_val_loss/val_denom
val_seg_loss = total_seg_loss/val_denom
val_dec_loss = total_dec_loss/val_denom
mlflow.log_metric('val_loss', val_loss)
mlflow.log_metric('val_seg_loss', val_seg_loss)
mlflow.log_metric('val_dec_loss', val_dec_loss)
wandb.log({'val_loss': val_loss,
'epoch': e,
'val_seg_loss': val_seg_loss,
'val_dec_loss': val_dec_loss})
if val_loss < val_lowest or val_seg_loss < val_seg_lowest or val_dec_loss < val_dec_lowest:
best_model = model
#mlflow.pytorch.log_state_dict(best_model.state_dict(),
# artifact_path=os.path.join(self.checkpoint_dir, f"{time.time()}_epoch{e}")
# )
try:
model_path = os.path.join(wandb.run.dir, f'model_epoch_{e}.pth')
torch.save(best_model.state_dict(), model_path)
artifact = wandb.Artifact('model', type='model')
artifact.add_file(model_path)
wandb.log_artifact(artifact)
mlflow.pytorch.log_model(best_model, f'model_epoch_{e}')
except:
print('PermissionError: [WinError 32] The process cannot access the file because it is being used by another process.')
print('Proceeding with updating best metrics and lowest validation loss.')
if val_loss < val_lowest:
val_lowest = val_loss
if val_seg_loss < val_seg_lowest:
val_seg_lowest = val_seg_loss
if val_dec_loss < val_dec_lowest:
val_dec_lowest = val_dec_loss
best_ji, best_precision, best_recall = avg_ji, avg_pre, avg_rec
# https://pytorch.org/docs/stable/optim.html
self.scheduler.step(val_seg_loss, val_dec_loss)
return best_model, best_ji, best_precision, best_recall