What are the Training Arguments for ImageNet Pre-Trained Model? about condensenet HOT 5 OPEN

@LeighDavis Hi, same questions here, did you find the answer? Thanks

from condensenet.

@LeighDavis, Hi, I find the training setting for imagenet in script.sh, it is

python main.py --model condensenet -b 256 -j 20 /PATH/TO/IMAGENET --epochs 120 --stages 4-6-8-10-8 --growth 8-16-32-64-128 --group-1x1 4 --group-3x3 4 --condense-factor 4 --bottleneck 4 --resume --group-lasso 0.00001 --gpu 0,1,2,3

from condensenet.

Hi @lizhenstat, thanks for following up. Yes, I had tried those arguments from that training command from script.sh file but I kept getting similar shape errors. Did these arguments work for you?

from condensenet.

@LeighDavis Hi, to evaluate from the pre-trained converted mode, you can use the following code

python main.py --model condensenet_converted -b 64 -j 20 /PATH/TO/IMAGENET \
--stages 4-6-8-10-8 --growth 8-16-32-64-128 --gpu 0 --resume \
--evaluate-from /PATH/TO/CONVERTED/MODEL

Your log shows the mismatch of the parameters in the same layer
Since the converted model is rearranged to a group convolution, and on ImageNet, the condense factor is taking
the value of 2, which is equivalent to pruning 50% of the total filters.

I evaluate condensenet-74 on my one 1080-ti successfully.

from condensenet.

@lizhenstat Thanks for addressing @LeighDavis 's questions. I currently have no ImageNet installed on my machine, so I am really glad you can help!

from condensenet.