def test_get_dataset_frame(name, split, num_frames, size, dataset_type,
record_set_type):
dataset = core.get_dataset(
name,
DATA_ROOT,
split=split,
num_frames=num_frames,
size=size,
dataset_type=dataset_type,
record_set_type=record_set_type,
)
for i, (frames, label) in enumerate(dataset):
if not (i < MAX_ITERS):
break
print(i, frames.shape, label)
assert label < cfg.num_classes_dict[name]
assert frames.shape == torch.Size((3, num_frames, size, size))
def test_get_heatvol_dataset(name, split, num_frames, size, dataset_type,
record_set_type):
dataset = core.get_dataset(
name,
DATA_ROOT,
split=split,
num_frames=num_frames,
size=size,
dataset_type=dataset_type,
record_set_type=record_set_type,
)
print(f'Dataset {name}-{split} has len: {len(dataset)}')
for i, (frames, label, heatvol, volmask) in enumerate(dataset):
if not (i < MAX_ITERS):
break
print(i, frames.shape, label, heatvol.shape, volmask)
assert label < cfg.num_classes_dict[name]
assert frames.shape == torch.Size((3, num_frames, size, size))
wandb.watch(model)
# setting device
device = torch.device(
'cuda') if torch.cuda.is_available() else torch.device('cpu')
# 多GPU
print(torch.cuda.device_count())
if device == 'cuda' and torch.cuda.device_count() > 1:
model = torch.nn.DataParallel(model, device_ids=[0, 1])
print("using device", device)
model.to(device)
train_dataset = get_dataset(name="boolq",
tokenizer=tokenizer,
split='train')
test_dataset = get_dataset(name="boolq",
tokenizer=tokenizer,
split='validation')
train_dataloader = DataLoader(train_dataset, batch_size=6, shuffle=True)
test_dataloader = DataLoader(test_dataset, batch_size=6, shuffle=True)
# Prepare optimizer and schedule (linear warmup and decay)
no_decay = ['bias', 'LayerNorm.weight']
optimizer_grouped_parameters = [{
'params': [
p for n, p in model.named_parameters()
if not any(nd in n for nd in no_decay)
],
wandb.watch(model)
# setting device
device = torch.device(
'cuda') if torch.cuda.is_available() else torch.device('cpu')
# 多GPU
print(torch.cuda.device_count())
if torch.cuda.device_count() > 1:
model = torch.nn.DataParallel(model, device_ids=[0, 1])
print("using device", device)
model.to(device)
train_dataset = get_dataset(name="snli",
tokenizer=tokenizer,
split='train')
test_dataset = get_dataset(name="snli", tokenizer=tokenizer, split='test')
train_dataloader = DataLoader(train_dataset, batch_size=16, shuffle=True)
test_dataloader = DataLoader(test_dataset, batch_size=1, shuffle=True)
# Prepare optimizer and schedule (linear warmup and decay)
no_decay = ['bias', 'LayerNorm.weight']
optimizer_grouped_parameters = [
{
"params": [
p for n, p in model.named_parameters()
if not any(nd in n for nd in no_decay)
],
"weight_decay":