-
Notifications
You must be signed in to change notification settings - Fork 0
/
train_captioner.py
213 lines (170 loc) · 8.09 KB
/
train_captioner.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
# using https://github.com/rmokady/CLIP_prefix_caption
import torch
import torch.nn as nn
import pytorch_lightning as pl
from pytorch_lightning.callbacks import ModelCheckpoint, LearningRateMonitor
from pytorch_lightning.loggers import WandbLogger
from transformers import GPT2LMHeadModel, get_linear_schedule_with_warmup
from dataset import PrefixingDataset
from models.mapping_networks import MLP, TransformerMapper
import yaml
import argparse
from typing import Optional
class CaptionModel(pl.LightningModule):
def __init__(self,
model,
optimizer,
scheduler,
prefix_length: int,
config: dict
):
super(CaptionModel, self).__init__()
self.save_hyperparameters(ignore=['model', 'optimizer', 'scheduler'])
self.model = model
self.optimizer = optimizer
self.scheduler = scheduler
self.prefix_length = prefix_length
self.config = config
def forward(self, tokens, prefix, mask):
return self.model(tokens, prefix.float(), mask.float())
def loss_fn(self, outputs, tokens):
logits = outputs.logits[:, self.prefix_length - 1: -1]
return torch.nn.functional.cross_entropy(logits.reshape(-1, logits.shape[-1]), tokens.flatten(), ignore_index=0)
def training_step(self, batch, batch_idx):
tokens, mask, prefix = batch
output = self(tokens, prefix, mask)
loss = self.loss_fn(output, tokens)
self.log('train_loss', loss, on_step=True, on_epoch=True, prog_bar=True)
return loss
def validation_step(self, batch, batch_idx):
tokens, mask, prefix = batch
output = self(tokens, prefix, mask)
loss = self.loss_fn(output, tokens)
self.log('val_loss', loss, on_step=True, on_epoch=True, prog_bar=True)
return loss
def configure_optimizers(self):
return [self.optimizer], [{'scheduler': self.scheduler, 'interval': 'step'}]
class ClipCaptionModel(nn.Module):
def get_dummy_token(self, batch_size: int, device: torch.device) -> torch.Tensor:
return torch.zeros(batch_size, self.prefix_length, dtype=torch.int64, device=device)
def forward(self, tokens: torch.Tensor, prefix: torch.Tensor, mask: Optional[torch.Tensor] = None,
labels: Optional[torch.Tensor] = None):
embedding_text = self.gpt.transformer.wte(tokens)
prefix_projections = self.clip_project(prefix).view(-1, self.prefix_length, self.gpt_embedding_size)
embedding_cat = torch.cat((prefix_projections, embedding_text), dim=1)
if labels is not None:
dummy_token = self.get_dummy_token(tokens.shape[0], tokens.device)
labels = torch.cat((dummy_token, tokens), dim=1)
out = self.gpt(inputs_embeds=embedding_cat, labels=labels, attention_mask=mask)
return out
def __init__(self, prefix_length: int, clip_length: Optional[int] = None, prefix_size: int = 512,
num_layers: int = 8, mapping_type: str = 'mlp', gpt2_type: str = 'gpt2'):
super(ClipCaptionModel, self).__init__()
self.prefix_length = prefix_length
self.gpt = GPT2LMHeadModel.from_pretrained(gpt2_type)
self.gpt_embedding_size = self.gpt.transformer.wte.weight.shape[1]
if mapping_type == 'mlp':
self.clip_project = MLP((prefix_size, (self.gpt_embedding_size * prefix_length) // 2,
self.gpt_embedding_size * prefix_length))
else:
self.clip_project = TransformerMapper(prefix_size, self.gpt_embedding_size, prefix_length,
clip_length, num_layers)
class ClipCaptionPrefix(ClipCaptionModel):
def parameters(self, recurse: bool = True):
return self.clip_project.parameters()
def train(self, mode: bool = True):
super(ClipCaptionPrefix, self).train(mode)
self.gpt.eval()
return self
def main(config, savename):
logger = WandbLogger(
entity='slavaheroes',
project='brain-captioner',
name=f'captioner_{config["gpt2_type"]}_prefix_{config["prefix_length"]}_{savename}',
)
# load dataset
train_ds = PrefixingDataset(
data_path=config['data_path'],
captions_path=config['captions_path'],
idx_path=config['train_idx_path'],
prefix_length=config['prefix_length'],
gpt2_type=config['gpt2_type'],
normalize_prefix=config['normalize_prefix']
)
test_ds = PrefixingDataset(
data_path=config['data_path'],
captions_path=config['captions_path'],
idx_path=config['test_idx_path'],
prefix_length=config['prefix_length'],
gpt2_type=config['gpt2_type'],
normalize_prefix=config['normalize_prefix']
)
train_loader = torch.utils.data.DataLoader(train_ds, batch_size=config['batch_size'], shuffle=True, num_workers=config['num_workers'], drop_last=False)
test_loader = torch.utils.data.DataLoader(test_ds, batch_size=1, shuffle=False, num_workers=0)
print("Length of train loader is ", len(train_loader))
print("Length of test loader is ", len(test_loader))
config['mapping_network']['prefix_length'] = config['prefix_length']
# load mapping network
if config['mapping_network']['only_prefix']:
# train only mapping network
model = ClipCaptionPrefix(
prefix_length=config['mapping_network']['prefix_length'],
clip_length=config['mapping_network']['clip_length'],
prefix_size=config['mapping_network']['prefix_size'],
num_layers=config['mapping_network']['num_layers'],
mapping_type=config['mapping_network']['mapping_type'],
gpt2_type=config['gpt2_type']
)
else:
# train both captioner and mapping network
model = ClipCaptionModel(
prefix_length=config['mapping_network']['prefix_length'],
clip_length=config['mapping_network']['clip_length'],
prefix_size=config['mapping_network']['prefix_size'],
num_layers=config['mapping_network']['num_layers'],
mapping_type=config['mapping_network']['mapping_type'],
gpt2_type=config['gpt2_type']
)
# load optimizer
optimizer = torch.optim.AdamW(model.parameters(), lr=config['lr'])
scheduler = get_linear_schedule_with_warmup(optimizer,
num_warmup_steps=config['num_warmup_steps'],
num_training_steps=config['epochs']*len(train_loader))
print("Number of training steps: ", config['epochs']*len(train_loader))
pl_module = CaptionModel(
model=model,
optimizer=optimizer,
scheduler=scheduler,
prefix_length=config['prefix_length'],
config=config
)
callbacks = [
LearningRateMonitor(logging_interval='step'),
ModelCheckpoint(
dirpath=f'./checkpoints/captioner_{config["gpt2_type"]}_prefix_{config["prefix_length"]}_{savename}',
filename='orig_dinov2_captioner_{epoch:02d}_{val_loss:.5f}',
save_top_k=1,
monitor='val_loss',
mode='min',
verbose=True
)
]
trainer = pl.Trainer(
accelerator='gpu',
devices=[6],
max_epochs=config['epochs'],
logger=logger,
callbacks=callbacks,
accumulate_grad_batches=2,
log_every_n_steps=1,
val_check_interval=0.25,
)
trainer.fit(pl_module, train_loader, test_loader)
if __name__=='__main__':
parser = argparse.ArgumentParser(description='Train a captioner')
parser.add_argument('--config', type=str, default='config.yaml', help='Path to the config file')
parser.add_argument('--savename', type=str, default='captioner', help='Name of the model')
args = parser.parse_args()
config = yaml.load(open(args.config, 'r'), Loader=yaml.Loader)
pl.seed_everything(config['seed'])
main(config, args.savename)