ethanhe42/channel-pruning

关于Process Process-3报错

canteen-man opened this issue · 2 comments

canteen-man commented

我尝试用cifar10数据集训练剪枝，更改了数据集的读入路径，将VGG全连接层改为512个：
stage0 freeze
temp/bn_vgg.prototxt
including last conv layer!
run for 500 batches nFeatsPerBatch 100
Extracting conv1_1 (50000, 64)
Extracting conv1_2 (50000, 64)
Extracting conv2_1 (50000, 128)
Extracting conv2_2 (50000, 128)
Extracting conv3_1 (50000, 256)
Extracting conv3_2 (50000, 256)
Extracting conv3_3 (50000, 256)
Extracting conv4_1 (50000, 512)
Extracting conv4_2 (50000, 512)
Extracting conv4_3 (50000, 512)
Extracting conv5_1 (50000, 512)
Extracting conv5_2 (50000, 512)
Extracting conv5_3 (50000, 512)
Acc 49.820
wrote memory data layer to temp/mem_bn_vgg.prototxt
freezing imgs to temp/frozen500.pickle

stage1 speed3.0

loading imgs from temp/frozen500.pickle
loaded
Extracting X relu1_1 From Y conv1_2 stride 1
Acc 49.820
Process Process-3:
################################################################################
在进行到Process-3时报错
Traceback (most recent call last):
File "/usr/lib/python3.4/multiprocessing/process.py", line 254, in _bootstrap
self.run()
File "/usr/lib/python3.4/multiprocessing/process.py", line 93, in run
self._target(*self._args, **self._kwargs)
File "/home/wu/channel-pruning/lib/worker.py", line 21, in job
ret = target(**kwargs)
File "train.py", line 75, in solve
WPQ, new_pt = net.R3()
File "/home/wu/channel-pruning/lib/net.py", line 1359, in R3
V, H, VHr, b = VH_decompose(weights, rank=rank, DEBUG=DEBUG, X=X, Y=Y)
File "/home/wu/channel-pruning/lib/decompose.py", line 101, in VH_decompose
V, sigmaVH, H = svd(VH)
File "/home/wu/channel-pruning/lib/decompose.py", line 155, in svd
return scipy.linalg.svd(x, full_matrices=False, lapack_driver='gesvd')
TypeError: svd() got an unexpected keyword argument 'lapack_driver'
不知道该如何解决？

ethanhe42 commented
canteen-man commented

@yihui-he 感谢您的指点，问题解决了。您源代码下边就注释着已经删除lapack_driver='gesvd'的函数。