The implementation of tf.nn.conv2d() is written in C++, which invokes optimized code using Eigen or the cuDNN library (for GPU).
確定! 回上一頁