Getting zero accuracy on running in "main" mode

[satyamr1]

I am getting zero accuracy when I run

python mainNN.py --mode main

Do I need to run it on some other mode first to train the model?

I already had AudioSet downloaded with me so is this problem anyhow dependent on the variation in data downloading procedure?

I am running the code on small number of files and not the whole dataset

I am using Google Colab.

Using batch size: 64 Using lr: 0.000025 Loaded dataloader and loss function. Optimizer loaded. Epoch: 0, Subepoch: 0, Loss: 0.782957, Valloss: 0.804846, batch_size: 64, acc: 0.000000, valacc: 0.000000 Epoch: 0, Subepoch: 10, Loss: 0.670178, Valloss: 0.813841, batch_size: 64, acc: 0.000000, valacc: 0.000000 Epoch: 0, Subepoch: 20, Loss: 0.767996, Valloss: 0.848316, batch_size: 64, acc: 0.000000, valacc: 0.000000 Epoch: 0, Subepoch: 30, Loss: 0.774370, Valloss: 0.745560, batch_size: 64, acc: 0.000000, valacc: 0.000000 Epoch: 0, Subepoch: 40, Loss: 0.789229, Valloss: 0.763325, batch_size: 64, acc: 0.000000, valacc: 0.000000 Epoch: 0, Subepoch: 50, Loss: 0.721946, Valloss: 0.747408, batch_size: 64, acc: 0.000000, valacc: 0.000000 Epoch: 1, Subepoch: 0, Loss: 0.685583, Valloss: 0.750512, batch_size: 64, acc: 0.000000, valacc: 0.000000 Epoch: 1, Subepoch: 10, Loss: 0.820786, Valloss: 0.750623, batch_size: 64, acc: 0.000000, valacc: 0.000000 Epoch: 1, Subepoch: 20, Loss: 0.732163, Valloss: 0.698070, batch_size: 64, acc: 0.000000, valacc: 0.000000 Epoch: 1, Subepoch: 30, Loss: 0.753967, Valloss: 0.691314, batch_size: 64, acc: 0.000000, valacc: 0.000000

Please help me out. Thanks