Resnet-from-scratch
Implement a simple Resnet according to the tutorial
- 根据教程使用pytorch实现Resnet
- 微调了教程中的一些代码,比如validationset 实际和 trainingset来自一个batch
- 使用pytorch自带的dataloader下载CIFAR-10数据
- 添加了部分注释,标识了layer的输出shape
- 添加了visualization part
2022年10月21日更新
Resnet_labml.py
文件结合了Labmlai的writer库进行实现.
If you have an error about decoder , you can put the save_info
function in labml\internal\experiment\experiment_run.py
:
def save_info(self):
...
if self.diff is not None:
with open(str(self.diff_path), "w") as f:
f.write(self.diff)
change into :
def save_info(self):
...
if self.diff is not None:
with open(str(self.diff_path), "w",encoding='utf-8') as f:
f.write(self.diff)
2022年10月21日更新
Resnet_board.py
文件结合加入了tensor board进行可视化实现.