conv_general_dilated实现

AI算法网奇

于 2021-02-14 11:26:33 发布

阅读量2k

点赞数

分类专栏：深度学习宝典

本文链接：https://blog.csdn.net/jacke121/article/details/113806678

版权

深度学习宝典专栏收录该内容

998 篇文章 717 订阅 ¥29.90 ¥99.00

订阅专栏

超级会员免费看

本文探讨了TensorFlow中conv_general_dilated的实现，参考了一位GitHub用户的代码实现，并指出可能存在的一些问题。

摘要由CSDN通过智能技术生成

参考：

https://github.com/DarrenZhang01/TensorFlow_GSoC/blob/2fd360c8b1b8c9106044034f6a8b5c2734db9c3d/tf_jax_stax/tf_conv_general.py

结果应该还不对：


"""
This file contains a general conv operation for TensorFlow.
Zhibo Zhang, 2020.06.06
"""
import torch.nn as nn
import numpy as np
import tensorflow as tf



# # Translate dimension numbers from number representations into string
# #   representations
# #
# # For example, in a 1D case, "NWC" means "# batches", "spatial dimensions" and
# #   "# Channels (features)", so the acco