flywheel1412

> I fixed this by switching pytorch versions > `conda install pytorch==1.1.0 torchvision==0.3.0 cudatoolkit=10.0 -c pytorch` fix it, tks

我是升级后出现这个问题的，问题解决了吗？

i fix it by modify this [line](https://github.com/baudm/parseq/blob/8734d7323de479050ee2ef6e0268c944cab537bb/strhub/models/parseq/system.py#L127) as fellow ``` tgt_in[:, j] = p_i.argmax(-1).squeeze() ```

+1

可以把匹配过程放到cpu上做

加上这句试试 `torch.cuda.empty_cache()`

tkx for your reply, where cloud i found the benchmark table of `timm`'s models? which cloud be a better guide for everyone to choose models

https://pan.baidu.com/s/1OgxPSSzUhaC9mPltIpp2pg#list/path=%2F author put the url of weights in README