horcham
horcham
Train with O0 might work. You could set the amp_level and amp_level_infer to O0, and try to train the model. `O2` amp_level might cause unsteady training. You could try to...
Can it train normally with svtr_tiny.yaml instead of svtr_tiny_ch.yaml?
It seems you train ic15 dataset on svtr_tiny_ch, but ic15 does not contain chinese characters. How about using svtr_tiny.yaml instead?
Would you mind offering us more information, so that we could locate the problems faster? - your GPU information - cuda/cudnn version - mindspore version
In the log, the warnings occur with "here are no valid values in the ndarray", maybe some Data produce large or overflow gradient while training, and cause nan problems. To...
We are sorry for keeping your long waiting. Some Python 3.7 environment may meet this problem when multiple python3 environment exists. You could try the following steps to solve this...