No validation loss ?

[ssbagalkar]

I have been trying to reproduce some of the results as well as trying to train on a custom dataset. While, the training runs without any issues, there is no validation loss reported, either on command line or on tensorboard. A little digging helped me to understand that inside pysot/models/model_builder.py only the total_loss, cls_loss and loc_loss is reported. Can we add validation losses to this code? If so, how and can anyone give me any tips. I would love to contribute as well

def forward(self, data):
    """ only used in training
    """
    template = data['template'].cuda()
    search = data['search'].cuda()
    label_cls = data['label_cls'].cuda()
    label_loc = data['label_loc'].cuda()
    label_loc_weight = data['label_loc_weight'].cuda()

    # get feature
    zf = self.backbone(template)
    xf = self.backbone(search)
    if cfg.MASK.MASK:
        zf = zf[-1]
        self.xf_refine = xf[:-1]
        xf = xf[-1]
    if cfg.ADJUST.ADJUST:
        zf = self.neck(zf)
        xf = self.neck(xf)
    cls, loc = self.rpn_head(zf, xf)

    # get loss
    cls = self.log_softmax(cls)
    cls_loss = select_cross_entropy_loss(cls, label_cls)
    loc_loss = weight_l1_loss(loc, label_loc, label_loc_weight)

    outputs = {}
    outputs['total_loss'] = cfg.TRAIN.CLS_WEIGHT * cls_loss + \
        cfg.TRAIN.LOC_WEIGHT * loc_loss
    outputs['cls_loss'] = cls_loss
    outputs['loc_loss'] = loc_loss

    if cfg.MASK.MASK:
        # TODO
        mask, self.mask_corr_feature = self.mask_head(zf, xf)
        mask_loss = None
        outputs['total_loss'] += cfg.TRAIN.MASK_WEIGHT * mask_loss
        outputs['mask_loss'] = mask_loss
    return outputs

[ZhiyuanChen]

It is the forward function, so if you call it in training, you get training loss, and if you call it in validation, you get validation loss