-
Notifications
You must be signed in to change notification settings - Fork 1
/
train.py
206 lines (168 loc) · 7.04 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
import logging
import time
import numpy as np
import torch
from torch.optim.lr_scheduler import ExponentialLR
from data_loader import Batch, create_dataloaders, get_level_sizes, load_ontology
from loss import LabelSmoothing, SimpleLossCompute
from model import make_model
from optimizer import DenseSparseAdam
class TrainState:
"""Track number of steps, examples, and tokens processed"""
step: int = 0 # Steps in the current epoch
accum_step: int = 0 # Number of gradient accumulation steps
samples: int = 0 # total # of examples used
tokens: int = 0 # total # of tokens processed
def run_train_epoch(data_iter, model, loss_fn, optimizer, scheduler, accum_iter, size_epoch, train_state):
"""
Train a single epoch
:param data_iter: torch.utils.data.DataLoader: iterator over train batches
:param model: model instance
:param loss_fn: SimpleLossCompute (see loss.py): generate predictions and compute loss
:param optimizer: DenseSparseAdam optimizer
:param scheduler: learning rate scheduler (ExponentialLR)
:param accum_iter: int; gradient accumulation steps
:param size_epoch: dataset_size / batch_size
:param train_state: struct for tracking training state (see above)
:return: tuple: (average loss, train_state)
"""
start = time.time()
total_tokens = 0
total_loss = 0
tokens = 0
n_accum = 0
for i, batch in enumerate(data_iter):
out = model.forward(batch.src, batch.tgt, batch.src_mask, batch.tgt_mask)
loss, loss_node = loss_fn(out, batch.tgt_y, batch.ntokens)
loss_node.backward()
if i % accum_iter == 0:
optimizer.step()
optimizer.zero_grad(set_to_none=True)
n_accum += 1
train_state.accum_step += 1
scheduler.step()
total_loss += loss
total_tokens += batch.ntokens
if i % 40 == 1:
lr = optimizer.param_groups[0]["lr"]
elapsed = time.time() - start
logging.info(f"Epoch Done: {(100 * i / size_epoch):.2f} | "
f"Acc Steps: {n_accum:3d} | "
f"Loss: {(loss / batch.ntokens):6.2f} | "
f"T/Sec: {(tokens / elapsed):7.1f} | "
f"LR: {lr:6.1e} | ")
tokens = 0
train_state.step += 1
train_state.samples += batch.src.shape[0]
train_state.tokens += batch.ntokens
del loss
del loss_node
return total_loss / total_tokens, train_state
def run_eval(data_iter, model, loss_fn):
"""
Evaluate model on the validation set
:param data_iter: torch.utils.data.DataLoader: iterator over validation batches
:param model: model instance
:param loss_fn: SimpleLossCompute (see loss.py): generate predictions and compute loss
:return: tuple: average loss, prec@5, ndcg@5
"""
start = time.time()
total_loss = 0
total_tokens = 0
for batch in data_iter:
out = model.forward(batch.src, batch.tgt, batch.src_mask, batch.tgt_mask)
loss, loss_node = loss_fn(out, batch.tgt_y, batch.ntokens, valid=True)
total_loss += loss.item()
total_tokens += batch.ntokens
avg_loss = total_loss / total_tokens
elapsed = time.time() - start
logging.info(f"Result Eval | Time Elapsed : {elapsed:.3f} | Loss: {avg_loss}")
return avg_loss
def train_model(vocab_src, vocab_tgt, tokenizer_src, config):
"""
:param vocab_src: instance of torchtext.vocab.Vocab, mapping between tokens and their ids
:param vocab_tgt: instance of torchtext.vocab.Vocab, mapping between labels and their ids
:param tokenizer_src: function for text tokenization
:param config: config
:return:
"""
gpu = 0
logging.info(f"Train worker process using GPU")
torch.cuda.set_device(gpu)
pad_idx = vocab_tgt["<blank>"]
emb_src_init = np.load(config["Paths"]["embedding_src"])
print(emb_src_init.shape)
label2level = load_ontology(config["Paths"]["ontology"]) # {str: int}
level_sizes = get_level_sizes(config["Prediction"].getint("max_level"), vocab_tgt, label2level)
# create model instance
model = make_model(vocab_src, vocab_tgt, config, emb_src_init=emb_src_init)
model.cuda(gpu)
model_name = config["Model"]["name"]
# loss function
criterion = LabelSmoothing(
level_sizes=level_sizes,
padding_idx=pad_idx,
criterion=config["Train"]["loss"],
smoothing=config["Train"].getfloat("smoothing")
)
criterion.cuda(gpu)
loss_fn = SimpleLossCompute(criterion)
# optimizer & scheduler
optimizer = DenseSparseAdam(
model.parameters(),
lr=config["Train"].getfloat("base_lr"),
betas=(config["Train"].getfloat("beta1"), config["Train"].getfloat("beta2")),
eps=config["Train"].getfloat("eps"),
weight_decay=config["Train"].getfloat("weight_decay"),
)
lr_scheduler = ExponentialLR(
optimizer=optimizer,
gamma=config["Train"].getfloat("gamma")
)
# X_dataloader is instance of torch.utils.data.DataLoader, yields batched (src, tgt) pairs, where
# src: 2D tensor of shape (batch_size, max_padding_src) with document tokens ids
# tgt: 2D tensor of shape (batch_size, max_padding_tgt) with relevant labels ids
train_dataloader, valid_dataloader, test_dataloader = create_dataloaders(
gpu,
vocab_src,
vocab_tgt,
tokenizer_src,
config,
)
epoch_size = len(train_dataloader.dataset) / config["DataLoader"].getint("batch_size_train")
nepochs = config["Train"].getint("num_epochs")
# simple struct for tracking training state
train_state = TrainState()
best_loss = float('inf')
for epoch in range(nepochs):
model.train()
logging.info(f"[GPU{gpu}] Epoch {epoch} Training ====")
_, train_state = run_train_epoch(
(Batch(b[0], b[1], pad_idx) for b in train_dataloader),
model,
loss_fn,
optimizer,
lr_scheduler,
accum_iter=config["Train"].getint("accum_iter"),
size_epoch=epoch_size,
train_state=train_state,
)
# evaluate model on validation dataset
logging.info(f"Epoch {epoch} Evaluation ====")
model.eval()
with torch.no_grad():
valid_loss = run_eval(
(Batch(b[0], b[1], pad_idx) for b in valid_dataloader),
model=model,
loss_fn=loss_fn,
)
if valid_loss < best_loss:
best_loss = valid_loss
logging.info(f"New Best Score, saving ...")
file_path = f"models/{model_name}.pt"
torch.save({
'epoch': epoch,
'model_state_dict': model.state_dict(),
'optimizer_state_dict': optimizer.state_dict(),
}, file_path)
torch.cuda.empty_cache()