Error when increase batch size

[Phan-Trung-Thuan]

Hello, when I increase batch size to 64 in utilities/argument_funcs.py, I got this error. Note that I'm running on kaggle gpu
Setting:

dataset_dir: ./dataset/
input_dir_music: ./dataset/vevo_chord/lab_v2_norm
input_dir_video: ./dataset/vevo_vis
output_dir: ./saved_models
weight_modulus: 1
print_modulus: 1

n_workers: 4
force_cpu: False
tensorboard: False

continue_weights: None
continue_epoch: None

lr: None
ce_smoothing: None
batch_size: 64
epochs: 100

rpr: True
max_sequence_midi: 2048
max_sequence_video: 300
max_sequence_chord: 300
n_layers: 6
num_heads: 8
d_model: 512

dim_feedforward: 1024
dropout: 0.1
is_video: True

Traceback (most recent call last):
File "/kaggle/working/Video2Music/train.py", line 283, in
main(vm, False)
File "/kaggle/working/Video2Music/train.py", line 193, in main
train_metric_dict = eval_model(model, train_loader,
File "/kaggle/working/Video2Music/utilities/run_model_vevo.py", line 266, in eval_model
y= model(x,
File "/opt/conda/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
return forward_call(*input, **kwargs)
File "/kaggle/working/Video2Music/model/video_music_transformer.py", line 77, in forward
feature_key_padded = torch.full((x.shape[0], x.shape[1], 1), feature_key.item())
ValueError: only one element tensors can be converted to Python scalars