-
Notifications
You must be signed in to change notification settings - Fork 0
/
train.py
150 lines (116 loc) · 5.45 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
import torch
import torch.nn as nn
from tqdm import tqdm
import numpy as np
from pathlib import Path
from model import build_decoder
from torch.utils.data import DataLoader
from dataset import SpikeDataset
import warnings
from config import get_config, latest_weights_file_path, get_weights_file_path
import pytorch_ssim
import wandb
def load_dataloader(config):
train_ds = SpikeDataset(config["neuron_ranges"], config["stimuli_list"], categ='train', seed = config["seed"] )
test_ds = SpikeDataset(config["neuron_ranges"], config["stimuli_list"], categ='test', seed = config["seed"] )
train_dataloader = DataLoader(train_ds, config["batch_size"], shuffle=True)
test_dataloader = DataLoader(test_ds, config["batch_size"], shuffle=True)
return train_dataloader, test_dataloader
def ssim(img1, img2, device):
l = []
for _ in range(5):
l.append(torch.tensor(1).to(device))
for _ in range(10):
l.append(torch.tensor(2).to(device))
for _ in range(30):
l.append(torch.tensor(5).to(device))
all_loss=0
for i in range(45):
img_1=img1[:,:,i,:,:]
img_2=img2[:,:,i,:,:]
ssim_loss = pytorch_ssim.SSIM(window_size=4, device=device).to(device)
all_loss+=ssim_loss(img_1, img_2, device)*l[i].to(device)
return (sum(l)-all_loss)/sum(l)
def run_test(model, test_dataloader, loss_fn1, loss_fn2, global_step, device):
model.eval()
loss = 0
with torch.no_grad():
for batch in test_dataloader:
batch_input = batch["input"].to(device)
batch_output = batch["output"].to(device)
model_output = model.decode(batch_input)
loss += loss_fn1(model_output, batch_output, device)+loss_fn2(model_output, batch_output).item()
print(f'********* Test Loss: {loss} **********')
# log the loss
wandb.log({'test/loss': loss, 'global_step': global_step})
return loss
def train_model(config):
device = torch.device(config["device"] if torch.cuda.is_available() else 'cpu')
print(f'Using device {device}')
Path(config['model_folder']).mkdir(parents=True, exist_ok=True)
Path(config['loss_folder']).mkdir(parents=True, exist_ok=True)
train_dataloader, test_dataloader = load_dataloader(config)
model = build_decoder(neuron_dim=config["num_neurons"]).to(device)
optimizer = torch.optim.Adam(model.parameters(), lr=config['lr'], eps=1e-9)
scheduler =torch.optim.lr_scheduler.StepLR(optimizer, step_size=3, gamma=0.85, last_epoch=-1)
loss_fn1 = ssim
loss_fn2 = nn.MSELoss()
initial_epoch = 0
global_step = 0
if config['pre_train']:
model_filename = latest_weights_file_path(config) if config['pre_train']=='latest' else get_weights_file_path(config, config['preload'])
print(f'Preloading model {model_filename}')
state = torch.load(model_filename)
model.load_state_dict(state[['model_state_dict']])
initial_epoch = state['epoch']+1
optimizer.load_state_dict(state['optimizer_state_dict'])
scheduler.load_state_dict(state['scheduler_state_dict'])
global_step = state['global_step']
del state
# define our custom x axis metric
wandb.define_metric("global_step")
# define which metrics will be plotted against it
wandb.define_metric("test/*", step_metric="global_step")
wandb.define_metric("train/*", step_metric="global_step")
train_loss = []
test_loss = []
for epoch in range(initial_epoch, config['num_epochs']):
torch.cuda.empty_cache()
model.train()
batch_iterator = tqdm(train_dataloader, desc=f'Processing epoch {epoch:02d}')
temp_loss = 0
for batch in batch_iterator:
batch_input = batch["input"].to(device) #torch.Size([64, 45, 140])
batch_output = batch["output"].to(device) #torch.Size([64, 3, 45, 32, 32])
model_output = model.decode(batch_input) # [bs, C3, D45, H32, W32]
loss = loss_fn1(model_output, batch_output, device)+loss_fn2(model_output, batch_output)
batch_iterator.set_postfix({f"loss": f"{loss.item():6.3f}"})
# log the loss
wandb.log({'train/loss': loss.item(), 'global_step': global_step})
loss.backward()
optimizer.step()
optimizer.zero_grad(set_to_none=True)
global_step += 1
temp_loss = temp_loss+ loss.item()
scheduler.step()
train_loss.append(temp_loss)
test_loss.append(run_test(model, test_dataloader, loss_fn1, loss_fn2, global_step, device))
# Save the model
model_filename = get_weights_file_path(config, f'{epoch:02d}')
torch.save({
'epoch': epoch,
'global_step': global_step,
'optimizer_state_dict': optimizer.state_dict(),
'model_state_dict': model.state_dict(), # Save the weights of model
'scheduler_state_dict': scheduler.state_dict()
}, model_filename)
np.save(str(Path('.') / config['loss_folder'] / config['model_basename'] / 'train_loss.npy'), train_loss)
np.save(str(Path('.') / config['loss_folder'] / config['model_basename'] / 'test_loss.npy'), test_loss)
if __name__ == '__main__':
warnings.filterwarnings('ignore')
config = get_config()
wandb.init(
project="neural_decoding",
config=config
)
train_model(config)