Rick Yang

And what about `catchAll` listener in Hubot? ```coffee module.exports = (robot) -> robot.catchAll (msg) -> msg.send "I don't know how to react to: #{msg.message.text}." ```

w32 & w48 not work, either. PyTorch 1.12, Python 3.9 on Paperspace ```python learn = get_segmentation_learner(dls=dls, number_classes=2, segmentation_type="Semantic Segmentation", architecture_name="hrnet", backbone_name="hrnet_w32", splitter=segmentron_splitter, loss_func=CustomLoss(), metrics=[Dice, foreground_acc, JaccardCoeff], wd=1e-3).to_fp16() ``` ``` RuntimeError...

complete error message. [hrnet-w32-error.txt](https://github.com/WaterKnight1998/SemTorch/files/10846031/hrnet-w32-error.txt)

After deleting the previous cache file, the notebook can load the hrnet_w32 weights. It seems the PTH cache will always be `~/.cache/torch/checkpoints`.