JAVA

第六章错误

5

作者吹嘘自己多牛，希望你更新一下代码，这个错误太多了，那么你服务也要走点心。别这么马虎，就这么不管，

BUG：AttributeError: Can't pickle local object 'GeneratorEnqueuer.start..data_generator_task'

This code is not reproducible. Why there are so many stars? If you load it from vgg16, the effect of map is poor.

1

This code is not reproducible. Why there are so many stars? If you load it from vgg16, the effect of map is poor.

Result

1

Epoch：【100】 train accuracy：0.9999 test accuracy：0.9487

Errors in the testing process

4

The training is normal. The following problems are encountered in the testing process. wrong with vl_nnbnorm vl::impl::dispatch_cudnn::operator(): BatchNormForwardWithMomentCudnn::operator(): cuDNN error [cudnn: "d:\\matlab\\r2018b\\bin\\matconvnet-1.0-beta25\\matlab\\src\\bits\\nnbnorm_cudnn.cu":155 (CUDNN_STATUS_BAD_PARAM)]

InvalidArgumentError (see above for traceback): Restoring from checkpoint failed. This is most likely due to a mismatch between the current graph and the graph from the checkpoint. Please ensure that you have not altered the graph expected based on the checkpoint. Original error: Assign requires shapes of both tensors to match. lhs shape= [1,1,1024,1024] rhs shape= [1,1,4096,4096] [[node save/Assign_29 (defined at E:\Infrared_Deeplearning\SSD-AbsoluteCoord\utility\scaffolds.py:86) ]]

2

InvalidArgumentError (see above for traceback): Restoring from checkpoint failed. This is most likely due to a mismatch between the current graph and the graph from the checkpoint. Please ensure that...

Memory Explode

6

This code is poorly written and can be tuned, but memory will continue to increase and will not be released until the memory explodes.

14k times------loss suddenly increase

3

Why does the value of loss suddenly increase and explode when training 14,000 times or so? The value of loss increased from 1.2 to 16 000

和unet对比感觉是没有提升啊

请问，要这个尺寸shape2=[1024,2048]的目的是什么，没明白这个含义