[错误1]
在model.fit()调用之后会出现如下错误说明:
Cannot interpret feed_dict key as Tensor: Tensor Tensor("conv2d_1_input:0", shape=(?, 256, 196, 3), ,dtype=float32) is not an element of this graph
[错误原因]:model底层tensorflow的session中还有数据
[解决方法]:在调用模型前面加一句 keras.backend.clear_session()
[错误2]
InternalError: Dst tensor is not initialized.
[错误原因]:GPU内存不够
[解决方法]:增加GPU内存