cp_examples/mip_finetune/mip_model.py [189:204]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    def loss(self, output, target):
        counts = 0
        loss = 0
        for i in range(len(output)):
            pos_weights, _ = filter_nans(self.pos_weights, target[i])
            loss_fn = torch.nn.BCEWithLogitsLoss(
                pos_weight=pos_weights, reduction="sum"
            )
            bind_logits, bind_labels = filter_nans(output[i], target[i])
            loss = loss + loss_fn(bind_logits, bind_labels)
            counts = counts + bind_labels.numel()
        counts = 1 if counts == 0 else counts
        loss = loss / counts
        return loss

    def training_step(self, batch, batch_idx):
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



cp_examples/sip_finetune/sip_finetune.py [149:166]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    def loss(self, output, target):
        counts = 0
        loss = 0
        for i in range(len(output)):
            pos_weights, _ = filter_nans(self.pos_weights, target[i])
            loss_fn = torch.nn.BCEWithLogitsLoss(
                pos_weight=pos_weights, reduction="sum"
            )
            bind_logits, bind_labels = filter_nans(output[i], target[i])
            loss = loss + loss_fn(bind_logits, bind_labels)
            counts = counts + bind_labels.numel()

        counts = 1 if counts == 0 else counts
        loss = loss / counts

        return loss

    def training_step(self, batch, batch_idx):
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -