-
Notifications
You must be signed in to change notification settings - Fork 0
/
base_trainer.py
336 lines (296 loc) · 11.8 KB
/
base_trainer.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
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
import torch
import tqdm
import sys
from .utils import common_functions as c_f
from .utils import loss_tracker as l_t
from .utils.key_checker import KeyChecker, KeyCheckerDict
class BaseTrainer:
def __init__(
self,
models,
optimizers,
batch_size,
loss_funcs,
mining_funcs,
dataset,
iterations_per_epoch=None,
data_device=None,
dtype=None,
loss_weights=None,
sampler=None,
collate_fn=None,
lr_schedulers=None,
gradient_clippers=None,
freeze_these=(),
freeze_trunk_batchnorm=False,
label_hierarchy_level=0,
dataloader_num_workers=32,
data_and_label_getter=None,
dataset_labels=None,
set_min_label_to_zero=False,
end_of_iteration_hook=None,
end_of_epoch_hook=None,
):
self.models = models
self.optimizers = optimizers
self.batch_size = batch_size
self.loss_funcs = loss_funcs
self.mining_funcs = mining_funcs
self.dataset = dataset
self.iterations_per_epoch = iterations_per_epoch
self.data_device = data_device
self.dtype = dtype
self.sampler = sampler
self.collate_fn = collate_fn
self.lr_schedulers = lr_schedulers
self.gradient_clippers = gradient_clippers
self.freeze_these = freeze_these
self.freeze_trunk_batchnorm = freeze_trunk_batchnorm
self.label_hierarchy_level = label_hierarchy_level
self.dataloader_num_workers = dataloader_num_workers
self.loss_weights = loss_weights
self.data_and_label_getter = data_and_label_getter
self.dataset_labels = dataset_labels
self.set_min_label_to_zero = set_min_label_to_zero
self.end_of_iteration_hook = end_of_iteration_hook
self.end_of_epoch_hook = end_of_epoch_hook
self.loss_names = list(self.loss_funcs.keys())
self.custom_setup()
self.verify_dict_keys()
self.initialize_models()
self.initialize_data_device()
self.initialize_label_mapper()
self.initialize_loss_tracker()
self.initialize_loss_weights()
self.initialize_data_and_label_getter()
self.initialize_hooks()
self.initialize_lr_schedulers()
def custom_setup(self):
pass
def calculate_loss(self):
raise NotImplementedError
def update_loss_weights(self):
pass
def train(self, start_epoch=1, num_epochs=1):
self.initialize_dataloader()
for self.epoch in range(start_epoch, num_epochs + 1):
self.set_to_train()
c_f.LOGGER.info("TRAINING EPOCH %d" % self.epoch)
pbar = tqdm.tqdm(range(self.iterations_per_epoch))
for self.iteration in pbar:
self.forward_and_backward()
self.end_of_iteration_hook(self)
pbar.set_description("total_loss=%.5f" % self.losses["total_loss"])
self.step_lr_schedulers(end_of_epoch=False)
self.step_lr_schedulers(end_of_epoch=True)
self.zero_losses()
if self.end_of_epoch_hook(self) is False:
break
def initialize_dataloader(self):
c_f.LOGGER.info("Initializing dataloader")
self.dataloader = c_f.get_train_dataloader(
self.dataset,
self.batch_size,
self.sampler,
self.dataloader_num_workers,
self.collate_fn,
)
if not self.iterations_per_epoch:
self.iterations_per_epoch = len(self.dataloader)
c_f.LOGGER.info("Initializing dataloader iterator")
self.dataloader_iter = iter(self.dataloader)
c_f.LOGGER.info("Done creating dataloader iterator")
def forward_and_backward(self):
self.zero_losses()
self.zero_grad()
self.update_loss_weights()
self.calculate_loss(self.get_batch())
self.loss_tracker.update(self.loss_weights)
self.backward()
self.clip_gradients()
self.step_optimizers()
def zero_losses(self):
for k in self.losses.keys():
self.losses[k] = 0
def zero_grad(self):
for v in self.models.values():
v.zero_grad()
for v in self.optimizers.values():
v.zero_grad()
def get_batch(self):
self.dataloader_iter, curr_batch = c_f.try_next_on_generator(
self.dataloader_iter, self.dataloader
)
data, labels = self.data_and_label_getter(curr_batch)
labels = c_f.process_label(
labels, self.label_hierarchy_level, self.label_mapper
)
return self.maybe_do_batch_mining(data, labels)
def compute_embeddings(self, data):
trunk_output = self.get_trunk_output(data)
embeddings = self.get_final_embeddings(trunk_output)
return embeddings
def get_final_embeddings(self, base_output):
return self.models["embedder"](base_output)
def get_trunk_output(self, data):
data = c_f.to_device(data, device=self.data_device, dtype=self.dtype)
print(data.size())
sys.exit()
trunk_1 = self.models["trunk_1"](data)
data_fft = torch.fft.fft(data,dim=2)
real = torch.real(data_fft)
imag = torch.imag(data_fft)
norm = torch.sqrt(torch.pow(real,2)+torch.pow(imag,2))
angle = torch.atan2(real,imag)
data_fusion = torch.cat([norm,angle], dim=1)
trunk_2 = self.models["trunk_2"](data_fusion)
return self.models["trunk"](torch.cat([trunk_1,trunk_2], dim=1))
def maybe_mine_embeddings(self, embeddings, labels):
if "tuple_miner" in self.mining_funcs:
return self.mining_funcs["tuple_miner"](embeddings, labels)
return None
def maybe_do_batch_mining(self, data, labels):
if "subset_batch_miner" in self.mining_funcs:
with torch.no_grad():
self.set_to_eval()
embeddings = self.compute_embeddings(data)
idx = self.mining_funcs["subset_batch_miner"](embeddings, labels)
self.set_to_train()
data, labels = data[idx], labels[idx]
return data, labels
def backward(self):
self.losses["total_loss"].backward()
def get_global_iteration(self):
return self.iteration + self.iterations_per_epoch * (self.epoch - 1)
def step_lr_schedulers(self, end_of_epoch=False):
if self.lr_schedulers is not None:
for k, v in self.lr_schedulers.items():
if end_of_epoch and k.endswith(
self.allowed_lr_scheduler_key_suffixes["epoch"]
):
v.step()
elif not end_of_epoch and k.endswith(
self.allowed_lr_scheduler_key_suffixes["iteration"]
):
v.step()
def step_lr_plateau_schedulers(self, validation_info):
if self.lr_schedulers is not None:
for k, v in self.lr_schedulers.items():
if k.endswith(self.allowed_lr_scheduler_key_suffixes["plateau"]):
v.step(validation_info)
def step_optimizers(self):
for k, v in self.optimizers.items():
if c_f.regex_replace("_optimizer$", "", k) not in self.freeze_these:
v.step()
def clip_gradients(self):
if self.gradient_clippers is not None:
for v in self.gradient_clippers.values():
v()
def maybe_freeze_trunk_batchnorm(self):
if self.freeze_trunk_batchnorm:
self.models["trunk"].apply(c_f.set_layers_to_eval("BatchNorm"))
def initialize_data_device(self):
if self.data_device is None:
self.data_device = torch.device(
"cuda" if torch.cuda.is_available() else "cpu"
)
def initialize_label_mapper(self):
self.label_mapper = c_f.LabelMapper(
self.set_min_label_to_zero, self.dataset_labels
).map
def initialize_loss_tracker(self):
self.loss_tracker = l_t.LossTracker(self.loss_names)
self.losses = self.loss_tracker.losses
def initialize_data_and_label_getter(self):
if self.data_and_label_getter is None:
self.data_and_label_getter = c_f.return_input
def trainable_attributes(self):
return [self.models, self.loss_funcs]
def set_to_train(self):
for T in self.trainable_attributes():
for k, v in T.items():
if k in self.freeze_these:
c_f.set_requires_grad(v, requires_grad=False)
v.eval()
else:
v.train()
self.maybe_freeze_trunk_batchnorm()
def set_to_eval(self):
for T in self.trainable_attributes():
for v in T.values():
v.eval()
def initialize_loss_weights(self):
if self.loss_weights is None:
self.loss_weights = {k: 1 for k in self.loss_names}
def initialize_hooks(self):
if self.end_of_iteration_hook is None:
self.end_of_iteration_hook = c_f.return_input
if self.end_of_epoch_hook is None:
self.end_of_epoch_hook = c_f.return_input
def initialize_lr_schedulers(self):
if self.lr_schedulers is None:
self.lr_schedulers = {}
def initialize_models(self):
if "embedder" not in self.models:
self.models["embedder"] = c_f.Identity()
def verify_dict_keys(self):
self.allowed_lr_scheduler_key_suffixes = {
"iteration": "_scheduler_by_iteration",
"epoch": "_scheduler_by_epoch",
"plateau": "_scheduler_by_plateau",
}
self.set_schema()
self.schema.verify(self)
self.verify_freeze_these_keys()
def modify_schema(self):
pass
def set_schema(self):
self.schema = KeyCheckerDict(
{
"models": KeyChecker(["trunk", "embedder"], essential=["trunk"]),
"loss_funcs": KeyChecker(["metric_loss"]),
"mining_funcs": KeyChecker(
["subset_batch_miner", "tuple_miner"],
warn_empty=False,
important=[],
),
"loss_weights": KeyChecker(
self.loss_names, warn_empty=False, essential=self.loss_names
),
"optimizers": KeyChecker(
lambda s, d: c_f.append_map(
d["models"].keys + d["loss_funcs"].keys, "_optimizer"
),
important=c_f.append_map(self.models.keys(), "_optimizer"),
),
"lr_schedulers": KeyChecker(
lambda s, d: [
x + y
for y in self.allowed_lr_scheduler_key_suffixes.values()
for x in d["models"].keys + d["loss_funcs"].keys
],
warn_empty=False,
important=[],
),
"gradient_clippers": KeyChecker(
lambda s, d: c_f.append_map(
d["models"].keys + d["loss_funcs"].keys, "_grad_clipper"
),
warn_empty=False,
important=[],
),
}
)
self.modify_schema()
def verify_freeze_these_keys(self):
allowed_keys = self.schema["models"].keys + self.schema["loss_funcs"].keys
for k in self.freeze_these:
assert k in allowed_keys, "freeze_these keys must be one of {}".format(
", ".join(allowed_keys)
)
if k + "_optimizer" in self.optimizers.keys():
c_f.LOGGER.warning(
"You have passed in an optimizer for {}, but are freezing its parameters.".format(
k
)
)