NVlabs
Result per class returns nan
Hi. Thanks for your Code and great community!
I am training a custom dataset. The training goes fine, but when i do the inference, the result per class returns nan like this:
+------------+-----+-----+ | Class | IoU | Acc | +------------+-----+-----+ | table | 0.0 | 0.0 | | pc | 0.0 | nan | | phone | 0.0 | nan | | background | 0.0 | nan | +------------+-----+-----+
My parameters are: """"optimizer = dict(delete=True, type='AdamW', lr=0.00006, betas=(0.9, 0.999), weight_decay=0.01, paramwise_cfg=dict(custom_keys={'pos_block': dict(decay_mult=0.), 'norm': dict(decay_mult=0.), 'head': dict(lr_mult=10.) }))
lr_config = dict(delete=True, policy='poly', warmup='linear', warmup_iters=1500, warmup_ratio=1e-6, power=1.0, min_lr=0.0, by_epoch=False)""""
What might be the cause?
