Class-balanced-loss-pytorch

Class-balanced-loss-pytorch copied to clipboard

why modulator?

2

Hi, i'm interested in your work! Now i have problem, why your implement code of focal loss use "modulator = torch.exp(-gamma * labels * logits - gamma * torch.log(1 +...

lai199508

could you explain why you are using binary cross-entropy

1

Hi, Inside the CBloss you are using binary cross-entropy, so why not using cross-entropy ? Could you explain? Thank you

xysong1201

Line 87 typo

2

original: cb_loss = F.binary_cross_entropy_with_logits(input = logits,target = labels_one_hot, weights = weights) update: cb_loss = F.binary_cross_entropy_with_logits(input = logits,target = labels_one_hot, weight = weights) - pytorch optimizer takes keyword 'weight', not 'weights'...

BenDrewry

'weight' keyword for sigmoid loss spelled as 'weights'

Anurag14

Update class_balanced_loss.py

1

The typo of weight parameter causes an error. It should be "weight" and not "weights".

krittaprot

shouldn't it be F.cross_entropy instead?

Starethics

Make it available on GPU device

2

Hi, I made the following modifications to make this loss function available on both CPU and GPU devices: 1. Check the device of `labels ` and `logits`; 2. Define the...

JosieHong