模型使用了Local Response Normalization 即mindspore.ops.LRN算子
r1.5 api地址mindspore.ops.LRN
【截图信息】
网络部分代码:
self.LRN = P.LRN()
x = self.conv1(x) x = self.relu(x) x = self.LRN(x) x = self.max_pool2d(x) x = self.conv2(x) x = self.relu(x) x = self.LRN(x) x = self.max_pool2d(x) x = self.conv3(x) x = self.relu(x) x = self.conv4(x) x = self.relu(x) x = self.conv5(x) x = self.relu(x)
转换模型代码:
param_dict = load_checkpoint(args.ckpt_file) load_param_into_net(network, param_dict) input_arr = Tensor(np.zeros([args.batch_size, 3, cfg.image_height, cfg.image_width]), ms.float32) export(network, input_arr, file_name=args.file_name, file_format=args.file_format)
报错信息
能导出MindIR的话,可以试一下用MindIR推理。 可以参考下这个:https://mindspore.cn/tutorials/zh-CN/master/inference.html