You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
2012 lines
102 KiB
2012 lines
102 KiB
_create_mixer
|
|
Pretrained= False
|
|
default_Cfgs= {'url': '', 'num_classes': 2, 'input_size': (3, 224, 224), 'pool_size': None, 'crop_pct': 0.875, 'interpolation': 'bicubic', 'fixed_input_size': True, 'mean': (0.5, 0.5, 0.5), 'std': (0.5, 0.5, 0.5), 'first_conv': 'stem.proj', 'classifier': 'head'}
|
|
dataset_len= 288
|
|
True
|
|
<class 'torch.utils.data.dataloader.DataLoader'>
|
|
dataset_len= 32
|
|
False
|
|
<class 'torch.utils.data.dataloader.DataLoader'>
|
|
------Training-------
|
|
batch= 0
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931,
|
|
0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931,
|
|
0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931,
|
|
0.6931, 0.6931, 0.6931, 0.6931, 0.6931], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931,
|
|
0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931,
|
|
0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931,
|
|
0.6931, 0.6931, 0.6931, 0.6931, 0.6931], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.6931, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 1
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.6927, 0.6926, 0.6926, 0.6926, 0.6926, 0.6926, 0.6926, 0.6926, 0.6926,
|
|
0.6926, 0.6926, 0.6926, 0.6926, 0.6926, 0.6926, 0.6926, 0.6926, 0.6926,
|
|
0.6926, 0.6926, 0.6926, 0.6925, 0.6926, 0.6926, 0.6926, 0.6926, 0.6926,
|
|
0.6926, 0.6926, 0.6925, 0.6926, 0.6937], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931,
|
|
0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931,
|
|
0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931,
|
|
0.6931, 0.6931, 0.6931, 0.6931, 0.6931], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.6927, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 2
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.6916, 0.6916, 0.6918, 0.6917, 0.6918, 0.6916, 0.6916, 0.6918, 0.6918,
|
|
0.6917, 0.6917, 0.6917, 0.6917, 0.6917, 0.6916, 0.6917, 0.6917, 0.6916,
|
|
0.6947, 0.6917, 0.6917, 0.6917, 0.6917, 0.6917, 0.6917, 0.6917, 0.6918,
|
|
0.6916, 0.6916, 0.6916, 0.6917, 0.6918], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931,
|
|
0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931,
|
|
0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931, 0.6931,
|
|
0.6931, 0.6931, 0.6931, 0.6931, 0.6931], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.6919, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 3
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.6903, 0.6907, 0.6903, 0.6904, 0.6907, 0.6906, 0.6907, 0.6903, 0.6904,
|
|
0.6905, 0.6903, 0.6959, 0.6904, 0.6906, 0.6905, 0.6906, 0.6905, 0.6905,
|
|
0.6906, 0.6904, 0.6906, 0.6905, 0.6906, 0.6960, 0.6903, 0.6907, 0.6905,
|
|
0.6906, 0.6905, 0.6907, 0.6907, 0.6907], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.6911, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 4
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.6894, 0.6894, 0.6889, 0.6894, 0.6889, 0.6975, 0.6893, 0.6889, 0.6891,
|
|
0.6892, 0.6896, 0.6973, 0.6976, 0.6890, 0.6893, 0.6889, 0.6888, 0.6889,
|
|
0.6894, 0.6888, 0.6974, 0.6974, 0.6890, 0.6896, 0.6972, 0.6887, 0.6892,
|
|
0.6889, 0.6890, 0.6888, 0.6893, 0.6892], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.6909, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 5
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.6876, 0.6878, 0.6880, 0.6876, 0.6877, 0.6878, 0.6880, 0.6989, 0.6879,
|
|
0.6873, 0.6884, 0.6877, 0.6876, 0.6876, 0.6879, 0.6877, 0.6876, 0.6882,
|
|
0.6875, 0.6875, 0.6877, 0.6881, 0.6877, 0.6991, 0.6880, 0.6875, 0.6881,
|
|
0.6987, 0.6874, 0.6877, 0.6880, 0.6884], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.6893, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 6
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.6862, 0.6858, 0.6861, 0.7007, 0.6857, 0.6864, 0.6866, 0.6860, 0.6861,
|
|
0.6862, 0.6862, 0.6863, 0.6864, 0.6863, 0.6857, 0.6863, 0.6859, 0.6862,
|
|
0.6857, 0.7008, 0.6860, 0.6869, 0.6860, 0.7005, 0.6861, 0.6857, 0.6866,
|
|
0.7008, 0.6857, 0.6856, 0.6855, 0.6864], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.6884, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 7
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.6848, 0.7023, 0.7028, 0.6848, 0.6849, 0.6848, 0.7029, 0.6839, 0.6837,
|
|
0.6838, 0.6837, 0.6848, 0.6839, 0.6842, 0.6838, 0.6844, 0.6843, 0.6847,
|
|
0.6836, 0.6840, 0.7021, 0.6838, 0.6836, 0.6836, 0.7027, 0.6849, 0.7023,
|
|
0.7023, 0.6842, 0.6848, 0.6843, 0.6844], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.6887, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 8
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.6833, 0.6820, 0.6809, 0.7045, 0.6832, 0.6825, 0.6827, 0.6829, 0.7038,
|
|
0.7045, 0.6832, 0.6833, 0.6839, 0.6829, 0.6834, 0.6840, 0.6829, 0.7037,
|
|
0.6833, 0.6829, 0.6827, 0.6828, 0.6829, 0.6821, 0.6829, 0.6834, 0.6828,
|
|
0.6823, 0.6835, 0.6822, 0.6825, 0.6829], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.6863, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
------Validating-------
|
|
output = tensor([[0.5123, 0.4877],
|
|
[0.5117, 0.4883],
|
|
[0.5118, 0.4882],
|
|
[0.5118, 0.4882],
|
|
[0.5124, 0.4876],
|
|
[0.5122, 0.4878],
|
|
[0.5119, 0.4881],
|
|
[0.5127, 0.4873],
|
|
[0.5123, 0.4877],
|
|
[0.5121, 0.4879],
|
|
[0.5120, 0.4880],
|
|
[0.5121, 0.4879],
|
|
[0.5122, 0.4878],
|
|
[0.5125, 0.4875],
|
|
[0.5128, 0.4872],
|
|
[0.5126, 0.4874],
|
|
[0.5124, 0.4876],
|
|
[0.5124, 0.4876],
|
|
[0.5124, 0.4876],
|
|
[0.5124, 0.4876],
|
|
[0.5123, 0.4877],
|
|
[0.5122, 0.4878],
|
|
[0.5120, 0.4880],
|
|
[0.5124, 0.4876],
|
|
[0.5120, 0.4880],
|
|
[0.5119, 0.4881],
|
|
[0.5117, 0.4883],
|
|
[0.5119, 0.4881],
|
|
[0.5118, 0.4882],
|
|
[0.5119, 0.4881],
|
|
[0.5117, 0.4883],
|
|
[0.5124, 0.4876]], device='cuda:0')
|
|
target= tensor([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 1, 1, 1, 1, 1, 1, 1], device='cuda:0')
|
|
eval_loss= tensor(0.7054, device='cuda:0')
|
|
eval_acc1= tensor(0., device='cuda:0')
|
|
------Training-------
|
|
batch= 0
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.6801, 0.6811, 0.6807, 0.6810, 0.6798, 0.6802, 0.6811, 0.6811, 0.6799,
|
|
0.6807, 0.7065, 0.6818, 0.6802, 0.6824, 0.6808, 0.6806, 0.6810, 0.6797,
|
|
0.6803, 0.6806, 0.7064, 0.6804, 0.6800, 0.7054, 0.6813, 0.6809, 0.6811,
|
|
0.6816, 0.6820, 0.6803, 0.6800, 0.6813], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932, 0.6932,
|
|
0.6932, 0.6932, 0.6932, 0.6932, 0.6932], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.6841, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 1
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.4440, 0.4431, 0.4423, 1.0512, 0.4354, 0.4477, 0.4348, 0.4372, 0.4326,
|
|
0.4353, 0.4288, 1.0622, 0.4329, 0.4263, 0.4323, 0.4431, 0.4356, 0.4367,
|
|
0.4443, 0.4378, 0.4332, 0.4242, 0.4547, 0.4266, 0.4205, 0.4291, 0.4446,
|
|
0.4455, 0.4256, 0.4609, 0.4424, 0.4271], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.7349, 0.7352, 0.7356, 0.7406, 0.7383, 0.7334, 0.7386, 0.7376, 0.7395,
|
|
0.7384, 0.7412, 0.7432, 0.7394, 0.7422, 0.7396, 0.7352, 0.7383, 0.7378,
|
|
0.7348, 0.7373, 0.7393, 0.7432, 0.7308, 0.7421, 0.7449, 0.7410, 0.7346,
|
|
0.7343, 0.7426, 0.7285, 0.7355, 0.7419], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5018, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 2
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([1.2373, 0.3479, 1.2251, 0.3466, 0.3493, 0.3496, 0.3549, 0.3486, 0.3413,
|
|
0.3564, 0.3540, 0.3417, 0.3526, 0.3504, 0.3545, 0.3465, 0.3580, 0.3578,
|
|
0.3433, 0.3538, 0.3425, 0.3502, 1.2251, 1.2219, 0.3512, 0.3503, 0.3460,
|
|
0.3590, 0.3446, 0.3562, 0.3484, 1.2262], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.7900, 0.7863, 0.7864, 0.7872, 0.7854, 0.7852, 0.7815, 0.7858, 0.7910,
|
|
0.7805, 0.7821, 0.7907, 0.7831, 0.7846, 0.7818, 0.7873, 0.7795, 0.7796,
|
|
0.7896, 0.7823, 0.7901, 0.7847, 0.7864, 0.7855, 0.7841, 0.7847, 0.7877,
|
|
0.7788, 0.7887, 0.7806, 0.7860, 0.7868], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5170, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 3
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3278, 0.3256, 0.3231, 0.3233, 0.3251, 0.3253, 0.3262, 0.3278, 0.3243,
|
|
0.3225, 0.3259, 0.3254, 0.3295, 0.3260, 0.3256, 0.3262, 0.3250, 0.3260,
|
|
0.3236, 0.3283, 0.3268, 0.3250, 0.3323, 0.3233, 0.3267, 0.3259, 0.3223,
|
|
0.3228, 0.3229, 0.3225, 0.3246, 0.3310], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8013, 0.8030, 0.8051, 0.8049, 0.8034, 0.8033, 0.8026, 0.8013, 0.8041,
|
|
0.8055, 0.8028, 0.8032, 0.8000, 0.8027, 0.8031, 0.8026, 0.8035, 0.8027,
|
|
0.8046, 0.8009, 0.8021, 0.8035, 0.7978, 0.8049, 0.8022, 0.8028, 0.8057,
|
|
0.8053, 0.8052, 0.8056, 0.8038, 0.7988], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.3733, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 4
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3171, 0.3167, 0.3177, 0.3159, 0.3175, 1.3035, 0.3192, 0.3186, 0.3171,
|
|
0.3171, 0.3168, 0.3180, 0.3172, 1.3040, 0.3188, 0.3205, 0.3203, 0.3178,
|
|
1.3052, 0.3161, 0.3166, 0.3167, 1.3039, 1.3045, 0.3166, 0.3163, 1.3048,
|
|
1.3016, 1.3018, 0.3161, 0.3171, 0.3150], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8100, 0.8104, 0.8095, 0.8110, 0.8096, 0.8102, 0.8083, 0.8087, 0.8100,
|
|
0.8100, 0.8103, 0.8092, 0.8099, 0.8103, 0.8086, 0.8071, 0.8073, 0.8094,
|
|
0.8107, 0.8108, 0.8104, 0.8104, 0.8103, 0.8105, 0.8104, 0.8107, 0.8106,
|
|
0.8096, 0.8097, 0.8108, 0.8100, 0.8117], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5885, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 5
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3142, 0.3149, 0.3155, 0.3143, 1.3105, 0.3147, 1.3086, 0.3151, 1.3093,
|
|
1.3094, 0.3152, 0.3146, 0.3145, 0.3144, 0.3151, 0.3150, 1.3094, 0.3147,
|
|
0.3158, 0.3148, 0.3149, 0.3150, 0.3157, 0.3144, 0.3144, 0.3155, 0.3151,
|
|
0.3148, 0.3144, 0.3150, 0.3159, 0.3147], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8125, 0.8118, 0.8114, 0.8123, 0.8124, 0.8120, 0.8118, 0.8117, 0.8120,
|
|
0.8120, 0.8116, 0.8122, 0.8122, 0.8122, 0.8117, 0.8118, 0.8121, 0.8120,
|
|
0.8111, 0.8120, 0.8119, 0.8118, 0.8112, 0.8123, 0.8123, 0.8113, 0.8117,
|
|
0.8119, 0.8123, 0.8118, 0.8110, 0.8120], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5045, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 6
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3139, 0.3137, 0.3137, 0.3138, 0.3138, 1.3116, 0.3138, 0.3139, 0.3140,
|
|
0.3139, 1.3113, 0.3137, 0.3137, 0.3140, 0.3139, 0.3138, 0.3143, 0.3143,
|
|
0.3139, 0.3140, 0.3138, 0.3138, 0.3138, 0.3140, 0.3139, 1.3122, 0.3137,
|
|
0.3138, 0.3141, 0.3138, 1.3120, 0.3139], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8127, 0.8129, 0.8129, 0.8128, 0.8128, 0.8127, 0.8128, 0.8127, 0.8126,
|
|
0.8127, 0.8126, 0.8129, 0.8129, 0.8126, 0.8127, 0.8128, 0.8124, 0.8124,
|
|
0.8127, 0.8127, 0.8128, 0.8128, 0.8128, 0.8126, 0.8127, 0.8129, 0.8129,
|
|
0.8128, 0.8126, 0.8128, 0.8129, 0.8127], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4760, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 7
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3135, 0.3134, 0.3135, 0.3136, 0.3136, 0.3135, 0.3139, 0.3135, 0.3134,
|
|
0.3134, 0.3134, 0.3135, 0.3136, 0.3135, 0.3134, 0.3135, 1.3126, 0.3137,
|
|
0.3135, 0.3138, 0.3134, 0.3135, 0.3138, 0.3134, 0.3135, 0.3136, 0.3134,
|
|
0.3135, 1.3127, 0.3135, 0.3134, 0.3135], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8130, 0.8131, 0.8131, 0.8130, 0.8130, 0.8131, 0.8127, 0.8131, 0.8131,
|
|
0.8131, 0.8131, 0.8131, 0.8130, 0.8131, 0.8131, 0.8131, 0.8131, 0.8129,
|
|
0.8131, 0.8128, 0.8131, 0.8130, 0.8128, 0.8131, 0.8131, 0.8130, 0.8131,
|
|
0.8131, 0.8131, 0.8130, 0.8131, 0.8131], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4197, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 8
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3135, 0.3134, 0.3134, 0.3133, 0.3134, 1.3129, 1.3130, 0.3133, 0.3135,
|
|
0.3133, 0.3134, 0.3134, 0.3134, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3130, 0.3134, 0.3133, 0.3134, 0.3134, 0.3134, 0.3134, 0.3133, 0.3133,
|
|
0.3134, 0.3134, 0.3134, 0.3134, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8131, 0.8132, 0.8132, 0.8132, 0.8131, 0.8131, 0.8132, 0.8132, 0.8130,
|
|
0.8132, 0.8131, 0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8132,
|
|
0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8131, 0.8131, 0.8132, 0.8132,
|
|
0.8131, 0.8131, 0.8131, 0.8132, 0.8132], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4477, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
------Validating-------
|
|
output = tensor([[9.9987e-01, 1.3297e-04],
|
|
[9.9981e-01, 1.9078e-04],
|
|
[9.9982e-01, 1.7943e-04],
|
|
[9.9982e-01, 1.7781e-04],
|
|
[9.9988e-01, 1.2462e-04],
|
|
[9.9986e-01, 1.3894e-04],
|
|
[9.9983e-01, 1.7157e-04],
|
|
[9.9989e-01, 1.0836e-04],
|
|
[9.9987e-01, 1.3451e-04],
|
|
[9.9985e-01, 1.4906e-04],
|
|
[9.9984e-01, 1.6087e-04],
|
|
[9.9984e-01, 1.5916e-04],
|
|
[9.9985e-01, 1.4523e-04],
|
|
[9.9988e-01, 1.1763e-04],
|
|
[9.9990e-01, 9.5136e-05],
|
|
[9.9989e-01, 1.0943e-04],
|
|
[9.9987e-01, 1.3151e-04],
|
|
[9.9987e-01, 1.2863e-04],
|
|
[9.9988e-01, 1.2315e-04],
|
|
[9.9988e-01, 1.2338e-04],
|
|
[9.9986e-01, 1.3644e-04],
|
|
[9.9986e-01, 1.4150e-04],
|
|
[9.9984e-01, 1.6069e-04],
|
|
[9.9988e-01, 1.2354e-04],
|
|
[9.9984e-01, 1.5766e-04],
|
|
[9.9983e-01, 1.7292e-04],
|
|
[9.9981e-01, 1.9048e-04],
|
|
[9.9982e-01, 1.7739e-04],
|
|
[9.9981e-01, 1.8760e-04],
|
|
[9.9983e-01, 1.7279e-04],
|
|
[9.9981e-01, 1.9314e-04],
|
|
[9.9987e-01, 1.2820e-04]], device='cuda:0')
|
|
target= tensor([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 1, 1, 1, 1, 1, 1, 1], device='cuda:0')
|
|
eval_loss= tensor(1.3130, device='cuda:0')
|
|
eval_acc1= tensor(0., device='cuda:0')
|
|
------Training-------
|
|
batch= 0
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([1.3131, 1.3131, 0.3133, 0.3133, 0.3133, 1.3130, 0.3133, 0.3133, 0.3134,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3134, 0.3133, 0.3133, 1.3131, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3134, 0.3133, 0.3133, 0.3133, 1.3131], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8131,
|
|
0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8132,
|
|
0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8132,
|
|
0.8132, 0.8132, 0.8132, 0.8132, 0.8132], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5039, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 1
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 1.3132, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3132, 0.3133, 0.3133,
|
|
0.3133, 1.3132, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8132, 0.8133, 0.8132,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8132, 0.8132, 0.8132, 0.8133, 0.8133,
|
|
0.8132, 0.8133, 0.8133, 0.8132, 0.8133, 0.8132, 0.8132, 0.8132, 0.8133,
|
|
0.8132, 0.8133, 0.8133, 0.8132, 0.8132], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 2
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3132,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8132, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 3
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 4
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 1.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5039, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 5
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 1.3133, 0.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5601, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 6
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.3633, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 7
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4758, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 8
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4195, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
------Validating-------
|
|
output = tensor([[1.0000e+00, 3.4283e-08],
|
|
[1.0000e+00, 6.0943e-08],
|
|
[1.0000e+00, 5.2414e-08],
|
|
[1.0000e+00, 5.2234e-08],
|
|
[1.0000e+00, 2.9134e-08],
|
|
[1.0000e+00, 3.5439e-08],
|
|
[1.0000e+00, 5.2968e-08],
|
|
[1.0000e+00, 2.4628e-08],
|
|
[1.0000e+00, 3.3891e-08],
|
|
[1.0000e+00, 3.9833e-08],
|
|
[1.0000e+00, 4.6143e-08],
|
|
[1.0000e+00, 4.5376e-08],
|
|
[1.0000e+00, 3.9571e-08],
|
|
[1.0000e+00, 2.7552e-08],
|
|
[1.0000e+00, 1.9134e-08],
|
|
[1.0000e+00, 2.4356e-08],
|
|
[1.0000e+00, 3.3188e-08],
|
|
[1.0000e+00, 3.1166e-08],
|
|
[1.0000e+00, 2.8297e-08],
|
|
[1.0000e+00, 2.8663e-08],
|
|
[1.0000e+00, 3.4569e-08],
|
|
[1.0000e+00, 3.7047e-08],
|
|
[1.0000e+00, 4.4655e-08],
|
|
[1.0000e+00, 2.7564e-08],
|
|
[1.0000e+00, 4.3321e-08],
|
|
[1.0000e+00, 5.1786e-08],
|
|
[1.0000e+00, 6.0503e-08],
|
|
[1.0000e+00, 5.4386e-08],
|
|
[1.0000e+00, 5.9361e-08],
|
|
[1.0000e+00, 5.2789e-08],
|
|
[1.0000e+00, 6.3719e-08],
|
|
[1.0000e+00, 3.1184e-08]], device='cuda:0')
|
|
target= tensor([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 1, 1, 1, 1, 1, 1, 1], device='cuda:0')
|
|
eval_loss= tensor(1.3133, device='cuda:0')
|
|
eval_acc1= tensor(0., device='cuda:0')
|
|
------Training-------
|
|
batch= 0
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 1.3133, 1.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5039, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 1
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 1.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5039, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 2
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 1.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 3
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4758, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 4
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 5
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 6
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4758, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 7
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 8
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4195, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
------Validating-------
|
|
output = tensor([[1.0000e+00, 3.4280e-08],
|
|
[1.0000e+00, 6.0938e-08],
|
|
[1.0000e+00, 5.2409e-08],
|
|
[1.0000e+00, 5.2230e-08],
|
|
[1.0000e+00, 2.9132e-08],
|
|
[1.0000e+00, 3.5436e-08],
|
|
[1.0000e+00, 5.2964e-08],
|
|
[1.0000e+00, 2.4626e-08],
|
|
[1.0000e+00, 3.3888e-08],
|
|
[1.0000e+00, 3.9830e-08],
|
|
[1.0000e+00, 4.6140e-08],
|
|
[1.0000e+00, 4.5372e-08],
|
|
[1.0000e+00, 3.9568e-08],
|
|
[1.0000e+00, 2.7549e-08],
|
|
[1.0000e+00, 1.9132e-08],
|
|
[1.0000e+00, 2.4354e-08],
|
|
[1.0000e+00, 3.3185e-08],
|
|
[1.0000e+00, 3.1164e-08],
|
|
[1.0000e+00, 2.8295e-08],
|
|
[1.0000e+00, 2.8660e-08],
|
|
[1.0000e+00, 3.4566e-08],
|
|
[1.0000e+00, 3.7044e-08],
|
|
[1.0000e+00, 4.4652e-08],
|
|
[1.0000e+00, 2.7561e-08],
|
|
[1.0000e+00, 4.3318e-08],
|
|
[1.0000e+00, 5.1782e-08],
|
|
[1.0000e+00, 6.0497e-08],
|
|
[1.0000e+00, 5.4381e-08],
|
|
[1.0000e+00, 5.9356e-08],
|
|
[1.0000e+00, 5.2785e-08],
|
|
[1.0000e+00, 6.3714e-08],
|
|
[1.0000e+00, 3.1181e-08]], device='cuda:0')
|
|
target= tensor([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 1, 1, 1, 1, 1, 1, 1], device='cuda:0')
|
|
eval_loss= tensor(1.3133, device='cuda:0')
|
|
eval_acc1= tensor(0., device='cuda:0')
|
|
------Training-------
|
|
batch= 0
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 1.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4758, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 1
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5039, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 2
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 1.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4758, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 3
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4758, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 4
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.3914, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 5
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 1.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5320, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 6
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4758, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 7
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.3914, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 8
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
------Validating-------
|
|
output = tensor([[1.0000e+00, 3.4279e-08],
|
|
[1.0000e+00, 6.0936e-08],
|
|
[1.0000e+00, 5.2408e-08],
|
|
[1.0000e+00, 5.2228e-08],
|
|
[1.0000e+00, 2.9131e-08],
|
|
[1.0000e+00, 3.5435e-08],
|
|
[1.0000e+00, 5.2962e-08],
|
|
[1.0000e+00, 2.4625e-08],
|
|
[1.0000e+00, 3.3888e-08],
|
|
[1.0000e+00, 3.9828e-08],
|
|
[1.0000e+00, 4.6138e-08],
|
|
[1.0000e+00, 4.5371e-08],
|
|
[1.0000e+00, 3.9567e-08],
|
|
[1.0000e+00, 2.7549e-08],
|
|
[1.0000e+00, 1.9132e-08],
|
|
[1.0000e+00, 2.4353e-08],
|
|
[1.0000e+00, 3.3184e-08],
|
|
[1.0000e+00, 3.1163e-08],
|
|
[1.0000e+00, 2.8294e-08],
|
|
[1.0000e+00, 2.8660e-08],
|
|
[1.0000e+00, 3.4565e-08],
|
|
[1.0000e+00, 3.7043e-08],
|
|
[1.0000e+00, 4.4650e-08],
|
|
[1.0000e+00, 2.7560e-08],
|
|
[1.0000e+00, 4.3316e-08],
|
|
[1.0000e+00, 5.1780e-08],
|
|
[1.0000e+00, 6.0495e-08],
|
|
[1.0000e+00, 5.4379e-08],
|
|
[1.0000e+00, 5.9354e-08],
|
|
[1.0000e+00, 5.2783e-08],
|
|
[1.0000e+00, 6.3712e-08],
|
|
[1.0000e+00, 3.1180e-08]], device='cuda:0')
|
|
target= tensor([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 1, 1, 1, 1, 1, 1, 1], device='cuda:0')
|
|
eval_loss= tensor(1.3133, device='cuda:0')
|
|
eval_acc1= tensor(0., device='cuda:0')
|
|
------Training-------
|
|
batch= 0
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5039, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 1
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 2
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 1.3133, 1.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5039, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 3
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4195, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 4
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 5
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4195, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 6
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 1.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4758, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 7
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 1.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5039, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 8
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 1.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
------Validating-------
|
|
output = tensor([[1.0000e+00, 3.4278e-08],
|
|
[1.0000e+00, 6.0935e-08],
|
|
[1.0000e+00, 5.2407e-08],
|
|
[1.0000e+00, 5.2227e-08],
|
|
[1.0000e+00, 2.9130e-08],
|
|
[1.0000e+00, 3.5434e-08],
|
|
[1.0000e+00, 5.2961e-08],
|
|
[1.0000e+00, 2.4625e-08],
|
|
[1.0000e+00, 3.3887e-08],
|
|
[1.0000e+00, 3.9828e-08],
|
|
[1.0000e+00, 4.6137e-08],
|
|
[1.0000e+00, 4.5370e-08],
|
|
[1.0000e+00, 3.9566e-08],
|
|
[1.0000e+00, 2.7548e-08],
|
|
[1.0000e+00, 1.9131e-08],
|
|
[1.0000e+00, 2.4353e-08],
|
|
[1.0000e+00, 3.3183e-08],
|
|
[1.0000e+00, 3.1162e-08],
|
|
[1.0000e+00, 2.8294e-08],
|
|
[1.0000e+00, 2.8659e-08],
|
|
[1.0000e+00, 3.4565e-08],
|
|
[1.0000e+00, 3.7043e-08],
|
|
[1.0000e+00, 4.4650e-08],
|
|
[1.0000e+00, 2.7560e-08],
|
|
[1.0000e+00, 4.3316e-08],
|
|
[1.0000e+00, 5.1779e-08],
|
|
[1.0000e+00, 6.0495e-08],
|
|
[1.0000e+00, 5.4379e-08],
|
|
[1.0000e+00, 5.9353e-08],
|
|
[1.0000e+00, 5.2782e-08],
|
|
[1.0000e+00, 6.3711e-08],
|
|
[1.0000e+00, 3.1179e-08]], device='cuda:0')
|
|
target= tensor([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 1, 1, 1, 1, 1, 1, 1], device='cuda:0')
|
|
eval_loss= tensor(1.3133, device='cuda:0')
|
|
eval_acc1= tensor(0., device='cuda:0')
|
|
------Training-------
|
|
batch= 0
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 1
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 2
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4758, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 3
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 1.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4195, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 4
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 1.3133, 1.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5320, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 5
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 1.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5320, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 6
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 7
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 1.3133, 0.3133,
|
|
0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4758, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 8
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.3914, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
------Validating-------
|
|
output = tensor([[1.0000e+00, 3.4278e-08],
|
|
[1.0000e+00, 6.0935e-08],
|
|
[1.0000e+00, 5.2407e-08],
|
|
[1.0000e+00, 5.2227e-08],
|
|
[1.0000e+00, 2.9130e-08],
|
|
[1.0000e+00, 3.5434e-08],
|
|
[1.0000e+00, 5.2961e-08],
|
|
[1.0000e+00, 2.4625e-08],
|
|
[1.0000e+00, 3.3887e-08],
|
|
[1.0000e+00, 3.9828e-08],
|
|
[1.0000e+00, 4.6137e-08],
|
|
[1.0000e+00, 4.5370e-08],
|
|
[1.0000e+00, 3.9566e-08],
|
|
[1.0000e+00, 2.7548e-08],
|
|
[1.0000e+00, 1.9131e-08],
|
|
[1.0000e+00, 2.4353e-08],
|
|
[1.0000e+00, 3.3183e-08],
|
|
[1.0000e+00, 3.1162e-08],
|
|
[1.0000e+00, 2.8294e-08],
|
|
[1.0000e+00, 2.8659e-08],
|
|
[1.0000e+00, 3.4564e-08],
|
|
[1.0000e+00, 3.7043e-08],
|
|
[1.0000e+00, 4.4649e-08],
|
|
[1.0000e+00, 2.7560e-08],
|
|
[1.0000e+00, 4.3316e-08],
|
|
[1.0000e+00, 5.1779e-08],
|
|
[1.0000e+00, 6.0495e-08],
|
|
[1.0000e+00, 5.4378e-08],
|
|
[1.0000e+00, 5.9353e-08],
|
|
[1.0000e+00, 5.2782e-08],
|
|
[1.0000e+00, 6.3711e-08],
|
|
[1.0000e+00, 3.1179e-08]], device='cuda:0')
|
|
target= tensor([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 1, 1, 1, 1, 1, 1, 1], device='cuda:0')
|
|
eval_loss= tensor(1.3133, device='cuda:0')
|
|
eval_acc1= tensor(0., device='cuda:0')
|
|
------Training-------
|
|
batch= 0
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4758, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 1
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 1.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5039, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 2
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
1.3133, 1.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5320, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 3
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4195, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 4
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 1.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4758, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 5
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 6
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 1.3133, 1.3133, 1.3133, 0.3133,
|
|
0.3133, 1.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5320, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 7
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.3914, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 8
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.3914, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
------Validating-------
|
|
output = tensor([[1.0000e+00, 3.4278e-08],
|
|
[1.0000e+00, 6.0935e-08],
|
|
[1.0000e+00, 5.2407e-08],
|
|
[1.0000e+00, 5.2227e-08],
|
|
[1.0000e+00, 2.9130e-08],
|
|
[1.0000e+00, 3.5434e-08],
|
|
[1.0000e+00, 5.2961e-08],
|
|
[1.0000e+00, 2.4625e-08],
|
|
[1.0000e+00, 3.3887e-08],
|
|
[1.0000e+00, 3.9828e-08],
|
|
[1.0000e+00, 4.6137e-08],
|
|
[1.0000e+00, 4.5370e-08],
|
|
[1.0000e+00, 3.9566e-08],
|
|
[1.0000e+00, 2.7548e-08],
|
|
[1.0000e+00, 1.9131e-08],
|
|
[1.0000e+00, 2.4353e-08],
|
|
[1.0000e+00, 3.3183e-08],
|
|
[1.0000e+00, 3.1162e-08],
|
|
[1.0000e+00, 2.8294e-08],
|
|
[1.0000e+00, 2.8659e-08],
|
|
[1.0000e+00, 3.4564e-08],
|
|
[1.0000e+00, 3.7042e-08],
|
|
[1.0000e+00, 4.4649e-08],
|
|
[1.0000e+00, 2.7560e-08],
|
|
[1.0000e+00, 4.3316e-08],
|
|
[1.0000e+00, 5.1779e-08],
|
|
[1.0000e+00, 6.0494e-08],
|
|
[1.0000e+00, 5.4378e-08],
|
|
[1.0000e+00, 5.9353e-08],
|
|
[1.0000e+00, 5.2782e-08],
|
|
[1.0000e+00, 6.3711e-08],
|
|
[1.0000e+00, 3.1179e-08]], device='cuda:0')
|
|
target= tensor([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 1, 1, 1, 1, 1, 1, 1], device='cuda:0')
|
|
eval_loss= tensor(1.3133, device='cuda:0')
|
|
eval_acc1= tensor(0., device='cuda:0')
|
|
------Training-------
|
|
batch= 0
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([1.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5320, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 1
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4195, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 2
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4195, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 3
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 1.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5039, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 4
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4195, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 5
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 1.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 1.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5320, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 6
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([1.3133, 1.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4758, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 7
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 8
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 1.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4195, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
------Validating-------
|
|
output = tensor([[1.0000e+00, 3.4278e-08],
|
|
[1.0000e+00, 6.0935e-08],
|
|
[1.0000e+00, 5.2407e-08],
|
|
[1.0000e+00, 5.2227e-08],
|
|
[1.0000e+00, 2.9130e-08],
|
|
[1.0000e+00, 3.5434e-08],
|
|
[1.0000e+00, 5.2961e-08],
|
|
[1.0000e+00, 2.4625e-08],
|
|
[1.0000e+00, 3.3887e-08],
|
|
[1.0000e+00, 3.9828e-08],
|
|
[1.0000e+00, 4.6137e-08],
|
|
[1.0000e+00, 4.5370e-08],
|
|
[1.0000e+00, 3.9566e-08],
|
|
[1.0000e+00, 2.7548e-08],
|
|
[1.0000e+00, 1.9131e-08],
|
|
[1.0000e+00, 2.4353e-08],
|
|
[1.0000e+00, 3.3183e-08],
|
|
[1.0000e+00, 3.1162e-08],
|
|
[1.0000e+00, 2.8294e-08],
|
|
[1.0000e+00, 2.8659e-08],
|
|
[1.0000e+00, 3.4564e-08],
|
|
[1.0000e+00, 3.7042e-08],
|
|
[1.0000e+00, 4.4649e-08],
|
|
[1.0000e+00, 2.7560e-08],
|
|
[1.0000e+00, 4.3316e-08],
|
|
[1.0000e+00, 5.1779e-08],
|
|
[1.0000e+00, 6.0494e-08],
|
|
[1.0000e+00, 5.4378e-08],
|
|
[1.0000e+00, 5.9353e-08],
|
|
[1.0000e+00, 5.2782e-08],
|
|
[1.0000e+00, 6.3711e-08],
|
|
[1.0000e+00, 3.1179e-08]], device='cuda:0')
|
|
target= tensor([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 1, 1, 1, 1, 1, 1, 1], device='cuda:0')
|
|
eval_loss= tensor(1.3133, device='cuda:0')
|
|
eval_acc1= tensor(0., device='cuda:0')
|
|
------Training-------
|
|
batch= 0
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 1.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4758, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 1
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4195, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 2
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 1.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5039, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 3
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4195, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 4
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4758, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 5
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 6
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 7
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 1.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 1.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5039, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 8
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 1.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4758, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
------Validating-------
|
|
output = tensor([[1.0000e+00, 3.4278e-08],
|
|
[1.0000e+00, 6.0935e-08],
|
|
[1.0000e+00, 5.2407e-08],
|
|
[1.0000e+00, 5.2227e-08],
|
|
[1.0000e+00, 2.9130e-08],
|
|
[1.0000e+00, 3.5434e-08],
|
|
[1.0000e+00, 5.2961e-08],
|
|
[1.0000e+00, 2.4625e-08],
|
|
[1.0000e+00, 3.3887e-08],
|
|
[1.0000e+00, 3.9828e-08],
|
|
[1.0000e+00, 4.6137e-08],
|
|
[1.0000e+00, 4.5370e-08],
|
|
[1.0000e+00, 3.9566e-08],
|
|
[1.0000e+00, 2.7548e-08],
|
|
[1.0000e+00, 1.9131e-08],
|
|
[1.0000e+00, 2.4353e-08],
|
|
[1.0000e+00, 3.3183e-08],
|
|
[1.0000e+00, 3.1162e-08],
|
|
[1.0000e+00, 2.8294e-08],
|
|
[1.0000e+00, 2.8659e-08],
|
|
[1.0000e+00, 3.4564e-08],
|
|
[1.0000e+00, 3.7042e-08],
|
|
[1.0000e+00, 4.4649e-08],
|
|
[1.0000e+00, 2.7560e-08],
|
|
[1.0000e+00, 4.3316e-08],
|
|
[1.0000e+00, 5.1779e-08],
|
|
[1.0000e+00, 6.0494e-08],
|
|
[1.0000e+00, 5.4378e-08],
|
|
[1.0000e+00, 5.9353e-08],
|
|
[1.0000e+00, 5.2782e-08],
|
|
[1.0000e+00, 6.3711e-08],
|
|
[1.0000e+00, 3.1179e-08]], device='cuda:0')
|
|
target= tensor([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 1, 1, 1, 1, 1, 1, 1], device='cuda:0')
|
|
eval_loss= tensor(1.3133, device='cuda:0')
|
|
eval_acc1= tensor(0., device='cuda:0')
|
|
------Training-------
|
|
batch= 0
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 1.3133, 0.3133, 1.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5039, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 1
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4195, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 2
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.3914, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 3
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4195, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 4
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 1.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
1.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5039, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 5
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 6
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 1.3133, 0.3133, 0.3133, 1.3133, 1.3133, 0.3133, 1.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 1.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5320, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 7
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 1.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.5039, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
batch= 8
|
|
torch.Size([32, 16, 256])
|
|
torch.Size([32, 2])
|
|
torch.Size([32])
|
|
LabelSmoothingCrossEntropy()
|
|
confidence, nll_loss, self.smoothing, smooth_loss
|
|
0.9 tensor([0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 0.3133, 1.3133, 0.3133, 0.3133, 0.3133, 0.3133, 0.3133,
|
|
0.3133, 0.3133, 1.3133, 0.3133, 0.3133], device='cuda:0',
|
|
grad_fn=<SqueezeBackward1>) 0.1 tensor([0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133, 0.8133,
|
|
0.8133, 0.8133, 0.8133, 0.8133, 0.8133], device='cuda:0',
|
|
grad_fn=<NegBackward0>)
|
|
loss= tensor(0.4476, device='cuda:0', grad_fn=<MeanBackward0>)
|
|
------Validating-------
|
|
output = tensor([[1.0000e+00, 3.4278e-08],
|
|
[1.0000e+00, 6.0935e-08],
|
|
[1.0000e+00, 5.2407e-08],
|
|
[1.0000e+00, 5.2227e-08],
|
|
[1.0000e+00, 2.9130e-08],
|
|
[1.0000e+00, 3.5434e-08],
|
|
[1.0000e+00, 5.2961e-08],
|
|
[1.0000e+00, 2.4625e-08],
|
|
[1.0000e+00, 3.3887e-08],
|
|
[1.0000e+00, 3.9828e-08],
|
|
[1.0000e+00, 4.6137e-08],
|
|
[1.0000e+00, 4.5370e-08],
|
|
[1.0000e+00, 3.9566e-08],
|
|
[1.0000e+00, 2.7548e-08],
|
|
[1.0000e+00, 1.9131e-08],
|
|
[1.0000e+00, 2.4353e-08],
|
|
[1.0000e+00, 3.3183e-08],
|
|
[1.0000e+00, 3.1162e-08],
|
|
[1.0000e+00, 2.8294e-08],
|
|
[1.0000e+00, 2.8659e-08],
|
|
[1.0000e+00, 3.4564e-08],
|
|
[1.0000e+00, 3.7042e-08],
|
|
[1.0000e+00, 4.4649e-08],
|
|
[1.0000e+00, 2.7560e-08],
|
|
[1.0000e+00, 4.3316e-08],
|
|
[1.0000e+00, 5.1779e-08],
|
|
[1.0000e+00, 6.0494e-08],
|
|
[1.0000e+00, 5.4378e-08],
|
|
[1.0000e+00, 5.9353e-08],
|
|
[1.0000e+00, 5.2782e-08],
|
|
[1.0000e+00, 6.3711e-08],
|
|
[1.0000e+00, 3.1179e-08]], device='cuda:0')
|
|
target= tensor([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 1, 1, 1, 1, 1, 1, 1], device='cuda:0')
|
|
eval_loss= tensor(1.3133, device='cuda:0')
|
|
eval_acc1= tensor(0., device='cuda:0')
|