loaderr

Traceback (most recent call last):

File "/ai/zhdata/lyp/multiyolov5_point_608_736/train_608_736.py", line 718, in <module>

train(hyp, opt, device, tb_writer)

File "/ai/zhdata/lyp/multiyolov5_point_608_736/train_608_736.py", line 166, in train

ema.ema.load_state_dict(ckpt['ema'].float().state_dict())

File "/ai/zhdata/lyp/conda/anaconda3/envs/mmd3.0/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1482, in load_state_dict

raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(

RuntimeError: Error(s) in loading state_dict for Model:

Missing key(s) in state_dict: "model.24.m.0.cv1.conv.weight", "model.24.m.0.cv1.bn.weight", "model.24.m.0.cv1.bn.bias", "model.24.m.0.cv1.bn.running_mean", "model.24.m.0.cv1.bn.running_var", "model.24.m.0.cv2.conv.weight", "model.24.m.0.cv2.bn.weight", "model.24.m.0.cv2.bn.bias", "model.24.m.0.cv2.bn.running_mean", "model.24.m.0.cv2.bn.running_var", "model.24.m.0.cv3.conv.weight", "model.24.m.0.cv3.bn.weight", "model.24.m.0.cv3.bn.bias", "model.24.m.0.cv3.bn.running_mean", "model.24.m.0.cv3.bn.running_var", "model.24.m.0.m.0.cv1.conv.weight", "model.24.m.0.m.0.cv1.bn.weight", "model.24.m.0.m.0.cv1.bn.bias", "model.24.m.0.m.0.cv1.bn.running_mean", "model.24.m.0.m.0.cv1.bn.running_var", "model.24.m.0.m.0.cv2.conv.weight", "model.24.m.0.m.0.cv2.bn.weight", "model.24.m.0.m.0.cv2.bn.bias", "model.24.m.0.m.0.cv2.bn.running_mean", "model.24.m.0.m.0.cv2.bn.running_var", "model.24.m.2.m.0.cv1.conv.weight", "model.24.m.2.m.0.cv1.bn.weight", "model.24.m.2.m.0.cv1.bn.bias", "model.24.m.2.m.0.cv1.bn.running_mean", "model.24.m.2.m.0.cv1.bn.running_var", "model.24.m.2.m.0.cv2.conv.weight", "model.24.m.2.m.0.cv2.bn.weight", "model.24.m.2.m.0.cv2.bn.bias", "model.24.m.2.m.0.cv2.bn.running_mean", "model.24.m.2.m.0.cv2.bn.running_var", "model.24.m.4.cv1.conv.weight", "model.24.m.4.cv1.bn.weight", "model.24.m.4.cv1.bn.bias", "model.24.m.4.cv1.bn.running_mean", "model.24.m.4.cv1.bn.running_var", "model.24.m.4.cv2.conv.weight", "model.24.m.4.cv2.bn.weight", "model.24.m.4.cv2.bn.bias", "model.24.m.4.cv2.bn.running_mean", "model.24.m.4.cv2.bn.running_var", "model.24.m.4.cv3.conv.weight", "model.24.m.4.cv3.bn.weight", "model.24.m.4.cv3.bn.bias", "model.24.m.4.cv3.bn.running_mean", "model.24.m.4.cv3.bn.running_var", "model.24.m.4.m.0.cv1.conv.weight", "model.24.m.4.m.0.cv1.bn.weight", "model.24.m.4.m.0.cv1.bn.bias", "model.24.m.4.m.0.cv1.bn.running_mean", "model.24.m.4.m.0.cv1.bn.running_var", "model.24.m.4.m.0.cv2.conv.weight", "model.24.m.4.m.0.cv2.bn.weight", "model.24.m.4.m.0.cv2.bn.bias", "model.24.m.4.m.0.cv2.bn.running_mean", "model.24.m.4.m.0.cv2.bn.running_var", "model.24.m.5.weight", "model.24.m.5.bias", "model.24.decoder1.3.cv1.conv.weight", "model.24.decoder1.3.cv1.bn.weight", "model.24.decoder1.3.cv1.bn.bias", "model.24.decoder1.3.cv1.bn.running_mean", "model.24.decoder1.3.cv1.bn.running_var", "model.24.decoder1.3.cv2.conv.weight", "model.24.decoder1.3.cv2.bn.weight", "model.24.decoder1.3.cv2.bn.bias", "model.24.decoder1.3.cv2.bn.running_mean", "model.24.decoder1.3.cv2.bn.running_var", "model.24.decoder1.3.cv3.conv.weight", "model.24.decoder1.3.cv3.bn.weight", "model.24.decoder1.3.cv3.bn.bias", "model.24.decoder1.3.cv3.bn.running_mean", "model.24.decoder1.3.cv3.bn.running_var", "model.24.decoder1.3.m.0.cv1.conv.weight", "model.24.decoder1.3.m.0.cv1.bn.weight", "model.24.decoder1.3.m.0.cv1.bn.bias", "model.24.decoder1.3.m.0.cv1.bn.running_mean", "model.24.decoder1.3.m.0.cv1.bn.running_var", "model.24.decoder1.3.m.0.cv2.conv.weight", "model.24.decoder1.3.m.0.cv2.bn.weight", "model.24.decoder1.3.m.0.cv2.bn.bias", "model.24.decoder1.3.m.0.cv2.bn.running_mean", "model.24.decoder1.3.m.0.cv2.bn.running_var", "model.24.decoder1.5.cv1.conv.weight", "model.24.decoder1.5.cv1.bn.weight", "model.24.decoder1.5.cv1.bn.bias", "model.24.decoder1.5.cv1.bn.running_mean", "model.24.decoder1.5.cv1.bn.running_var", "model.24.decoder1.5.cv2.conv.weight", "model.24.decoder1.5.cv2.bn.weight", "model.24.decoder1.5.cv2.bn.bias", "model.24.decoder1.5.cv2.bn.running_mean", "model.24.decoder1.5.cv2.bn.running_var", "model.24.decoder1.5.cv3.conv.weight", "model.24.decoder1.5.cv3.bn.weight", "model.24.decoder1.5.cv3.bn.bias", "model.24.decoder1.5.cv3.bn.running_mean", "model.24.decoder1.5.cv3.bn.running_var", "model.24.decoder1.5.m.0.cv1.conv.weight", "model.24.decoder1.5.m.0.cv1.bn.weight", "model.24.decoder1.5.m.0.cv1.bn.bias", "model.24.decoder1.5.m.0.cv1.bn.running_mean", "model.24.decoder1.5.m.0.cv1.bn.running_var", "model.24.decoder1.5.m.0.cv2.conv.weight", "model.24.decoder1.5.m.0.cv2.bn.weight", "model.24.decoder1.5.m.0.cv2.bn.bias", "model.24.decoder1.5.m.0.cv2.bn.running_mean", "model.24.decoder1.5.m.0.cv2.bn.running_var", "model.24.decoder1.6.weight", "model.24.decoder1.6.bias", "model.24.m32.2.cv1.conv.weight", "model.24.m32.2.cv1.bn.weight", "model.24.m32.2.cv1.bn.bias", "model.24.m32.2.cv1.bn.running_mean", "model.24.m32.2.cv1.bn.running_var", "model.24.m32.2.cv2.conv.weight", "model.24.m32.2.cv2.bn.weight", "model.24.m32.2.cv2.bn.bias", "model.24.m32.2.cv2.bn.running_mean", "model.24.m32.2.cv2.bn.running_var", "model.24.m32.2.cv3.conv.weight", "model.24.m32.2.cv3.bn.weight", "model.24.m32.2.cv3.bn.bias", "model.24.m32.2.cv3.bn.running_mean", "model.24.m32.2.cv3.bn.running_var", "model.24.m32.2.m.0.cv1.conv.weight", "model.24.m32.2.m.0.cv1.bn.weight", "model.24.m32.2.m.0.cv1.bn.bias", "model.24.m32.2.m.0.cv1.bn.running_mean", "model.24.m32.2.m.0.cv1.bn.running_var", "model.24.m32.2.m.0.cv2.conv.weight", "model.24.m32.2.m.0.cv2.bn.weight", "model.24.m32.2.m.0.cv2.bn.bias", "model.24.m32.2.m.0.cv2.bn.running_mean", "model.24.m32.2.m.0.cv2.bn.running_var", "model.24.m16.0.conv.weight", "model.24.m16.0.bn.weight", "model.24.m16.0.bn.bias", "model.24.m16.0.bn.running_mean", "model.24.m16.0.bn.running_var".

Unexpected key(s) in state_dict: "model.24.m.0.conv.weight", "model.24.m.0.bn.weight", "model.24.m.0.bn.bias", "model.24.m.0.bn.running_mean", "model.24.m.0.bn.running_var", "model.24.m.0.bn.num_batches_tracked", "model.24.m.1.cv1.conv.weight", "model.24.m.1.cv1.bn.weight", "model.24.m.1.cv1.bn.bias", "model.24.m.1.cv1.bn.running_mean", "model.24.m.1.cv1.bn.running_var", "model.24.m.1.cv1.bn.num_batches_tracked", "model.24.m.1.cv2.conv.weight", "model.24.m.1.cv2.bn.weight", "model.24.m.1.cv2.bn.bias", "model.24.m.1.cv2.bn.running_mean", "model.24.m.1.cv2.bn.running_var", "model.24.m.1.cv2.bn.num_batches_tracked", "model.24.m.1.cv3.conv.weight", "model.24.m.1.cv3.bn.weight", "model.24.m.1.cv3.bn.bias", "model.24.m.1.cv3.bn.running_mean", "model.24.m.1.cv3.bn.running_var", "model.24.m.1.cv3.bn.num_batches_tracked", "model.24.m.1.m.0.cv1.conv.weight", "model.24.m.1.m.0.cv1.bn.weight", "model.24.m.1.m.0.cv1.bn.bias", "model.24.m.1.m.0.cv1.bn.running_mean", "model.24.m.1.m.0.cv1.bn.running_var", "model.24.m.1.m.0.cv1.bn.num_batches_tracked", "model.24.m.1.m.0.cv2.conv.weight", "model.24.m.1.m.0.cv2.bn.weight", "model.24.m.1.m.0.cv2.bn.bias", "model.24.m.1.m.0.cv2.bn.running_mean", "model.24.m.1.m.0.cv2.bn.running_var", "model.24.m.1.m.0.cv2.bn.num_batches_tracked", "model.24.m.2.m.cv1.conv.weight", "model.24.m.2.m.cv1.bn.weight", "model.24.m.2.m.cv1.bn.bias", "model.24.m.2.m.cv1.bn.running_mean", "model.24.m.2.m.cv1.bn.running_var", "model.24.m.2.m.cv1.bn.num_batches_tracked", "model.24.m.2.m.cv2.conv.weight", "model.24.m.2.m.cv2.bn.weight", "model.24.m.2.m.cv2.bn.bias", "model.24.m.2.m.cv2.bn.running_mean", "model.24.m.2.m.cv2.bn.running_var", "model.24.m.2.m.cv2.bn.num_batches_tracked", "model.24.m.3.weight", "model.24.m.3.bias", "model.24.decoder1.2.cv1.conv.weight", "model.24.decoder1.2.cv1.bn.weight", "model.24.decoder1.2.cv1.bn.bias", "model.24.decoder1.2.cv1.bn.running_mean", "model.24.decoder1.2.cv1.bn.running_var", "model.24.decoder1.2.cv1.bn.num_batches_tracked", "model.24.decoder1.2.cv2.conv.weight", "model.24.decoder1.2.cv2.bn.weight", "model.24.decoder1.2.cv2.bn.bias", "model.24.decoder1.2.cv2.bn.running_mean", "model.24.decoder1.2.cv2.bn.running_var", "model.24.decoder1.2.cv2.bn.num_batches_tracked", "model.24.decoder1.2.cv3.conv.weight", "model.24.decoder1.2.cv3.bn.weight", "model.24.decoder1.2.cv3.bn.bias", "model.24.decoder1.2.cv3.bn.running_mean", "model.24.decoder1.2.cv3.bn.running_var", "model.24.decoder1.2.cv3.bn.num_batches_tracked", "model.24.decoder1.2.m.cv1.conv.weight", "model.24.decoder1.2.m.cv1.bn.weight", "model.24.decoder1.2.m.cv1.bn.bias", "model.24.decoder1.2.m.cv1.bn.running_mean", "model.24.decoder1.2.m.cv1.bn.running_var", "model.24.decoder1.2.m.cv1.bn.num_batches_tracked", "model.24.decoder1.2.m.cv2.conv.weight", "model.24.decoder1.2.m.cv2.bn.weight", "model.24.decoder1.2.m.cv2.bn.bias", "model.24.decoder1.2.m.cv2.bn.running_mean", "model.24.decoder1.2.m.cv2.bn.running_var", "model.24.decoder1.2.m.cv2.bn.num_batches_tracked", "model.24.decoder1.3.weight", "model.24.decoder1.3.bias", "model.24.m8.1.cv1.conv.weight", "model.24.m8.1.cv1.bn.weight", "model.24.m8.1.cv1.bn.bias", "model.24.m8.1.cv1.bn.running_mean", "model.24.m8.1.cv1.bn.running_var", "model.24.m8.1.cv1.bn.num_batches_tracked", "model.24.m8.1.cv2.conv.weight", "model.24.m8.1.cv2.bn.weight", "model.24.m8.1.cv2.bn.bias", "model.24.m8.1.cv2.bn.running_mean", "model.24.m8.1.cv2.bn.running_var", "model.24.m8.1.cv2.bn.num_batches_tracked", "model.24.m8.1.cv3.conv.weight", "model.24.m8.1.cv3.bn.weight", "model.24.m8.1.cv3.bn.bias", "model.24.m8.1.cv3.bn.running_mean", "model.24.m8.1.cv3.bn.running_var", "model.24.m8.1.cv3.bn.num_batches_tracked", "model.24.m8.1.m.0.cv1.conv.weight", "model.24.m8.1.m.0.cv1.bn.weight", "model.24.m8.1.m.0.cv1.bn.bias", "model.24.m8.1.m.0.cv1.bn.running_mean", "model.24.m8.1.m.0.cv1.bn.running_var", "model.24.m8.1.m.0.cv1.bn.num_batches_tracked", "model.24.m8.1.m.0.cv2.conv.weight", "model.24.m8.1.m.0.cv2.bn.weight", "model.24.m8.1.m.0.cv2.bn.bias", "model.24.m8.1.m.0.cv2.bn.running_mean", "model.24.m8.1.m.0.cv2.bn.running_var", "model.24.m8.1.m.0.cv2.bn.num_batches_tracked", "model.24.m32.1.cv1.conv.weight", "model.24.m32.1.cv1.bn.weight", "model.24.m32.1.cv1.bn.bias", "model.24.m32.1.cv1.bn.running_mean", "model.24.m32.1.cv1.bn.running_var", "model.24.m32.1.cv1.bn.num_batches_tracked", "model.24.m32.1.cv2.conv.weight", "model.24.m32.1.cv2.bn.weight", "model.24.m32.1.cv2.bn.bias", "model.24.m32.1.cv2.bn.running_mean", "model.24.m32.1.cv2.bn.running_var", "model.24.m32.1.cv2.bn.num_batches_tracked", "model.24.m32.1.cv3.conv.weight", "model.24.m32.1.cv3.bn.weight", "model.24.m32.1.cv3.bn.bias", "model.24.m32.1.cv3.bn.running_mean", "model.24.m32.1.cv3.bn.running_var", "model.24.m32.1.cv3.bn.num_batches_tracked", "model.24.m32.1.m.0.cv1.conv.weight", "model.24.m32.1.m.0.cv1.bn.weight", "model.24.m32.1.m.0.cv1.bn.bias", "model.24.m32.1.m.0.cv1.bn.running_mean", "model.24.m32.1.m.0.cv1.bn.running_var", "model.24.m32.1.m.0.cv1.bn.num_batches_tracked", "model.24.m32.1.m.0.cv2.conv.weight", "model.24.m32.1.m.0.cv2.bn.weight", "model.24.m32.1.m.0.cv2.bn.bias", "model.24.m32.1.m.0.cv2.bn.running_mean", "model.24.m32.1.m.0.cv2.bn.running_var", "model.24.m32.1.m.0.cv2.bn.num_batches_tracked", "model.24.m16.0.cv1.conv.weight", "model.24.m16.0.cv1.bn.weight", "model.24.m16.0.cv1.bn.bias", "model.24.m16.0.cv1.bn.running_mean", "model.24.m16.0.cv1.bn.running_var", "model.24.m16.0.cv1.bn.num_batches_tracked", "model.24.m16.0.cv2.conv.weight", "model.24.m16.0.cv2.bn.weight", "model.24.m16.0.cv2.bn.bias", "model.24.m16.0.cv2.bn.running_mean", "model.24.m16.0.cv2.bn.running_var", "model.24.m16.0.cv2.bn.num_batches_tracked", "model.24.m16.0.cv3.conv.weight", "model.24.m16.0.cv3.bn.weight", "model.24.m16.0.cv3.bn.bias", "model.24.m16.0.cv3.bn.running_mean", "model.24.m16.0.cv3.bn.running_var", "model.24.m16.0.cv3.bn.num_batches_tracked", "model.24.m16.0.m.0.cv1.conv.weight", "model.24.m16.0.m.0.cv1.bn.weight", "model.24.m16.0.m.0.cv1.bn.bias", "model.24.m16.0.m.0.cv1.bn.running_mean", "model.24.m16.0.m.0.cv1.bn.running_var", "model.24.m16.0.m.0.cv1.bn.num_batches_tracked", "model.24.m16.0.m.0.cv2.conv.weight", "model.24.m16.0.m.0.cv2.bn.weight", "model.24.m16.0.m.0.cv2.bn.bias", "model.24.m16.0.m.0.cv2.bn.running_mean", "model.24.m16.0.m.0.cv2.bn.running_var", "model.24.m16.0.m.0.cv2.bn.num_batches_tracked".

size mismatch for model.24.m.2.cv1.conv.weight: copying a param with shape torch.Size([128, 256, 1, 1]) from checkpoint, the shape in current model is torch.Size([32, 128, 1, 1]).

size mismatch for model.24.m.2.cv1.bn.weight: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([32]).

size mismatch for model.24.m.2.cv1.bn.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([32]).

size mismatch for model.24.m.2.cv1.bn.running_mean: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([32]).

size mismatch for model.24.m.2.cv1.bn.running_var: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([32]).

size mismatch for model.24.m.2.cv2.conv.weight: copying a param with shape torch.Size([128, 256, 1, 1]) from checkpoint, the shape in current model is torch.Size([32, 128, 1, 1]).

size mismatch for model.24.m.2.cv2.bn.weight: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([32]).

size mismatch for model.24.m.2.cv2.bn.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([32]).

size mismatch for model.24.m.2.cv2.bn.running_mean: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([32]).

size mismatch for model.24.m.2.cv2.bn.running_var: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([32]).

size mismatch for model.24.m.2.cv3.conv.weight: copying a param with shape torch.Size([256, 320, 1, 1]) from checkpoint, the shape in current model is torch.Size([64, 64, 1, 1]).

size mismatch for model.24.m.2.cv3.bn.weight: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.m.2.cv3.bn.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.m.2.cv3.bn.running_mean: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.m.2.cv3.bn.running_var: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.decoder1.1.cv1.conv.weight: copying a param with shape torch.Size([128, 256, 1, 1]) from checkpoint, the shape in current model is torch.Size([64, 256, 1, 1]).

size mismatch for model.24.decoder1.1.cv1.bn.weight: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.decoder1.1.cv1.bn.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.decoder1.1.cv1.bn.running_mean: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.decoder1.1.cv1.bn.running_var: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.decoder1.1.cv2.conv.weight: copying a param with shape torch.Size([128, 256, 1, 1]) from checkpoint, the shape in current model is torch.Size([64, 256, 1, 1]).

size mismatch for model.24.decoder1.1.cv2.bn.weight: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.decoder1.1.cv2.bn.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.decoder1.1.cv2.bn.running_mean: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.decoder1.1.cv2.bn.running_var: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.decoder1.1.cv3.conv.weight: copying a param with shape torch.Size([256, 256, 1, 1]) from checkpoint, the shape in current model is torch.Size([128, 128, 1, 1]).

size mismatch for model.24.decoder1.1.cv3.bn.weight: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([128]).

size mismatch for model.24.decoder1.1.cv3.bn.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([128]).

size mismatch for model.24.decoder1.1.cv3.bn.running_mean: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([128]).

size mismatch for model.24.decoder1.1.cv3.bn.running_var: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([128]).

size mismatch for model.24.decoder1.1.m.0.cv1.conv.weight: copying a param with shape torch.Size([128, 128, 1, 1]) from checkpoint, the shape in current model is torch.Size([64, 64, 1, 1]).

size mismatch for model.24.decoder1.1.m.0.cv1.bn.weight: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.decoder1.1.m.0.cv1.bn.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.decoder1.1.m.0.cv1.bn.running_mean: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.decoder1.1.m.0.cv1.bn.running_var: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.decoder1.1.m.0.cv2.conv.weight: copying a param with shape torch.Size([128, 128, 3, 3]) from checkpoint, the shape in current model is torch.Size([64, 64, 3, 3]).

size mismatch for model.24.decoder1.1.m.0.cv2.bn.weight: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.decoder1.1.m.0.cv2.bn.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.decoder1.1.m.0.cv2.bn.running_mean: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64]).

size mismatch for model.24.decoder1.1.m.0.cv2.bn.running_var: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([64]).