【文件属性】：
文件名称：pixel-cnn-pp:OpenAI的PixelCNN ++的Pytorch实现
文件大小：3.07MB
文件格式：ZIP
更新时间：2021-05-02 11:36:35
Python
PixelCNN ++
Pytorch实现
主要工作
预训练模型可
我保留了代码结构，以方便与正式代码进行比较。
该代码在测试集上达到2.95 BPD，而在正式的tensorflow实现上达到2.92 BPD。
运行代码
python main.py
与官方实施的差异
没有数据相关的权重初始化
没有用于测试集评估的过去模型的指数移动平均值
接触
该存储库不再维护。 如果需要，可以随时提出问题，但是响应速度可能很慢。
【文件预览】：
pixel-cnn-pp-master
----main.py(7KB)
----images()
--------pcnn_lr:0.00050_nr-resnet3_nr-filters160_29.png(60KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_77.png(59KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_26.png(61KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_47.png(55KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_74.png(58KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_86.png(58KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_119.png(58KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_122.png(56KB)
--------pcnn_lr:0.00050_nr-resnet3_nr-filters160_39.png(61KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_56.png(58KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_95.png(54KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_50.png(59KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_98.png(56KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_20.png(55KB)
--------pcnn_lr:0.00100_nr-resnet5_nr-filters160_0.png(61KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_17.png(57KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_137.png(55KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_5.png(61KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_38.png(57KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_92.png(56KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_68.png(58KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_104.png(57KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_11.png(57KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_113.png(59KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_107.png(57KB)
--------pcnn_lr:0.00050_nr-resnet3_nr-filters160_49.png(61KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_29.png(59KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_41.png(56KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_116.png(57KB)
--------pcnn_lr:0.00050_nr-resnet3_nr-filters160_19.png(58KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_80.png(57KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_110.png(58KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_125.png(54KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_134.png(55KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_143.png(55KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_44.png(60KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_53.png(57KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_2.png(61KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_101.png(57KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_35.png(58KB)
--------pcnn_lr:0.00050_nr-resnet3_nr-filters160_9.png(63KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_140.png(56KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_71.png(59KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_83.png(57KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_128.png(55KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_62.png(59KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_65.png(55KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_59.png(60KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_23.png(61KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_8.png(57KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_89.png(56KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_32.png(59KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_131.png(57KB)
--------pcnn_lr:0.00020_nr-resnet5_nr-filters160_14.png(59KB)
----utils.py(11KB)
----readme.md(1KB)
----model.py(7KB)
----license.md(1KB)
----layers.py(6KB)