qkv_conv = Conv2D(c*3, kernel_size=3, padding='same', groups=c*3, use_bias=False) project_out = Conv2D(filters=c, kernel_size=1, use_bias=False) temperature = tf.Variable([[[1.]]for_inrange(self.num_heads)]], shape=[None, self.num_heads, 1, 1], trainable=True)#--- ...
Hi i solve with before you build datasets, you must run this: import mmdet mmdet.datasets.coco.CocoDataset.CLASSES=('class_1','class2') you can adjust it with the dataset you have :) The solution provided here may be a more elegant one, as it only requires modifications in the configu...
{{ message }} This repository has been archived by the owner on Sep 18, 2024. It is now read-only. microsoft / nni Public archive Notifications You must be signed in to change notification settings Fork 1.8k Star 14.1k Code ...