Yunguan Fu

> Hi, @mathpluscode. Nice to see you here. Would the example in [#205 (comment)](https://github.com/fepegar/torchio/issues/205#issuecomment-647703444) work? > > If the reading time is another bottleneck, we could implement a reader that...

> Hi, @mathpluscode. Nice to see you here. Would the example in [#205 (comment)](https://github.com/fepegar/torchio/issues/205#issuecomment-647703444) work? > > If the reading time is another bottleneck, we could implement a reader that...

Could you elaborate a bit regarding this issue, thx! - What is the minimal example we want to demo? - Which data/model we would like to use?

I got the same confusion as well, the source comes from `_infer_shape`. ## Description [MaxPool](https://github.com/deepmind/dm-haiku/blob/main/haiku/_src/pool.py#L149) is calling [max_pool](https://github.com/deepmind/dm-haiku/blob/9c9624a58a700a8f2827ff073fbe9e65badc8606/haiku/_src/pool.py#L74), which eventually calls [_infer_shape](https://github.com/deepmind/dm-haiku/blob/9c9624a58a700a8f2827ff073fbe9e65badc8606/haiku/_src/pool.py#L32). But in `MaxPool` the argument description is only...

Hi, @Destinycjk 你遇到的问题可能是因为模型较大 你可以尝试减少模型大小 https://github.com/mathpluscode/ImgX-DiffSeg/blob/277eee4da54c1d1ab46eaee75b1b287f2ddd199a/imgx/conf/task/seg.yaml#L10 也可以直接修改 `debug=True` 这个会自动减小模型大小和训练数据 https://github.com/mathpluscode/ImgX-DiffSeg/blob/277eee4da54c1d1ab46eaee75b1b287f2ddd199a/imgx/conf/config.yaml#L8 Debug config的定义如下 https://github.com/mathpluscode/ImgX-DiffSeg/blob/277eee4da54c1d1ab46eaee75b1b287f2ddd199a/imgx/run_train.py#L25 希望可以帮到你 谢谢