Advice for machine learning task with chemical data

2

[batch, 1, length_of_spectra] is the correct shape for 1D convolution. Normalization and regularization is the proper way too.
Regarding your loss becoming NaN after a while, try applying this function to your model before you update weights

def sanitize_grads(module):
	for param in module.parameters():
		if param.grad is not None:
			torch.nan_to_num(param.grad, nan=0, posinf=1e5, neginf=-1e5, out=param.grad)
1 Like