Ezra-Yu

Sorry, we currently have no plans to support such a task.

try to modify your head config, for example [resnet50 head](https://github.com/open-mmlab/mmclassification/blob/master/configs/_base_/models/resnest50.py) ``` head=dict( type='LinearClsHead', num_classes=1000, in_channels=2048, loss=dict( type='LabelSmoothLoss', label_smooth_val=0.1, num_classes=1000, reduction='mean', loss_weight=1.0), topk=(1, 5), cal_acc=True)) ```

First, check the annotations of the dataset. you can use this [tool](https://mmclassification.readthedocs.io/en/latest/tools/visualization.html#pipeline-visualization).

sure, you can do that, but you have to implement your own [head](https://github.com/open-mmlab/mmclassification/tree/master/mmcls/models/heads), [loss](https://github.com/open-mmlab/mmclassification/tree/master/mmcls/models/losses) and [regressor](https://github.com/open-mmlab/mmclassification/tree/master/mmcls/models/classifiers). Also, the evaluation method needs to be rewrited.

It seems there is an error in `evalutionHook` config, your input of the model is the shape of `[32, 224, 225, 5]`, And excepted input shep is `[32, 224, 224,...

Yes, you can modify a hrnet_cifra.py from resnet101_cifra.py. The easiest way is to modify [the head](https://github.com/open-mmlab/mmclassification/blob/7c5ddb1e5bee68d52ff8b5622cdbd75c02792c07/configs/_base_/models/resnet101_cifar.py#L11).

If we add all cifra dataset, it will add a lot of configs, which is very messy, but very few people use it. All here only supports resnet_cifra.py as an...

When the number of categories of the model is greater than the number of categories in the dataset, it can be trained and no error will be reported.

You can try the multi-task classification task, refer to #808

For K-Fold cross-validation, all images should be placed in the same folder. You can refer to [the K-fold PR](https://github.com/open-mmlab/mmclassification/pull/563) for help, there is a detailed example config in PR desciption...