Code for PackNet: Adding Multiple Tasks to a Single Network by Iterative Pruning

hello，

I run this command：

python main.py --mode eval --dataset cubs_cropped \
  --loadname ../checkpoints/csf_0.75,0.75,-1_vgg16_0.5-nobias-nobn_1.pt

everything is ok, but test other model,

python main.py --mode eval --dataset cubs_cropped \
  --loadname ../checkpoints/csf_0.75,0.75,-1_vgg16bn_0.5-nobias-nobn_1.pt

or

python main.py --mode eval --dataset cubs_cropped \
  --loadname ../checkpoints/csf_0.75,0.75,-1_densenet121_0.5-nobias-nobn_1.pt

problem happened.

result = self.forward(*input, **kwargs)
File "pytorch_test/ijcai/packnet/src/networks.py", line 78, in forward
     x = self.shared(x)
File "pytorch_test/env3/lib/python3.5/site-packages/torch/nn/modules/module.py",
    result = self.forward(*input, **kwargs)
File "/home/linning/pytorch_test/env3/lib/python3.5/site-packages/torch/nn/modules/batchnorm.py", line 49, 
     self.training or not self.track_running_stats, self.momentum, self.eps)
AttributeError: 'BatchNorm2d' object has no attribute 'track_running_stats'

My PyTorch version is : 0.4.0, python version is 3.5 Could you help me fixed it, thank you! @arunmallya

hi @arunmallya ，I try to run script ./scripts/run_lwf.sh csf ../checkpoints/imagenet/vgg16.pt 2,3

seems vgg16.pt file worked with pytorch 0.2.0.post3, and can download from https://uofi.box.com/s/zap2p03tnst9dfisad4u0sfupc0y1fxt

is there any way to upgrade to newer pytorch version, such as pytorch 1.6

Thanks

Hello, thank you very much for sharing. But I ran into some problems while running your code. I would really appreciate it if you could help me.

1. I'm running

Python main. Py - mode eval - dataset cubs_cropped - loadname/home/qiuyu/packnet - master/checkpoints/csf_0,0.75, 75-1 _vgg16_0. 5 - nobias - nobn_1. Pt

In this code, the program appeared "killed" and then stopped running. I wonder if it was because of my computer configuration. I used i5 and gpu=2070. When you run this program, the computer gets really stuck.

2. I'm running @@@@ Bash run_all_baseline_finetuning. Sh

This code, the program is reporting an error, Traceback (most recent call last): File "main.py", line 393, in The main () File "main.py", line 377, in main The save = True, savename = args. Save_prefix) File "main.py", line 215, in train Self. Do_epoch (epoch_idx, optimizer) File "main.py", line 172, in do_epoch For batch, label in TQDM (self.train_data_loader, desc='Epoch: %d '% (epoch_idx)): The File "/ home/qiuyu/anaconda3 / envs/pytorch/lib/python3.6 / site - packages/TQDM _tqdm. Py", line 1017, in iter For obj in iterable: The File "/ home/qiuyu/anaconda3 / envs/pytorch/lib/python3.6 / site - packages/torch/utils/data/dataloader py", line 301, in iter Return DataLoaderIter (self) The File "/ home/qiuyu/anaconda3 / envs/pytorch/lib/python3.6 / site - packages/torch/utils/data/dataloader py", line 158, in init Baron tart () The File "/ home/qiuyu/anaconda3 / envs/pytorch/lib/python3.6 / multiprocessing/process. Py", line 105, in the start Self. _popen = self. _popen (self) The File "/ home/qiuyu/anaconda3 / envs/pytorch/lib/python3.6 / multiprocessing/context. Py", line 223, in _Popen Return _default_context. Get_context (). The Process. _Popen (process_obj) The File "/ home/qiuyu/anaconda3 / envs/pytorch/lib/python3.6 / multiprocessing/context. Py", line 277, in _Popen Return Popen (process_obj) The File "/ home/qiuyu/anaconda3 / envs/pytorch/lib/python3.6 multiprocessing/popen_fork py", line 19, in init Self. _launch (process_obj) The File "/ home/qiuyu/anaconda3 / envs/pytorch/lib/python3.6 multiprocessing/popen_fork py", line 66, in _launch The self. The pid = OS. The fork () OSError: [Errno 12] no allocate memory

What's the reason for this? Looking forward to your reply. Thank you very much. @

Hello

Please, I have a problem concerning the optimization of YOLOX model by Pytorch Pruning (Global & Local) I got the same inference time after the optimization, I don't understand why ?

is it that the prune model by Pytoch does not remove the tensors which have equal to 0 after the equalization or there is another thing

any help

thanks in advance

I am new in this task and read your code briefly。 Does line 137 in src/prune.py : ' self.current_masks = self.previous_masks' should be 'self.current_masks = mask'?

Hi @arunmallya ：

Running Platform: GoogleColab Environment: Python3.6 torch==0.2.0.post3 torchvision==0.1.9 torchnet, tqdm I have changed "raw_input()" to "input()" in network.py. When I try to run the program, the following error occurs.

Traceback (most recent call last): File "main.py", line 393, in main() File "main.py", line 331, in main model = ckpt['model'] KeyError: 'model'

The file structure below the project is the same as your GitHub commit, where the file structure in checkpoints is shown below.

The contents of the imagenet folder are shown below.

The files vgg16.pt, vgg16bn.pt, resnet50.pt, densenet121.pt were downloaded in the same environment with the following code. vgg16 = torchvision.models.vgg16(pretrained=True) torch.save(vgg16.state_dict(), "/content/drive/MyDrive/packnet/vgg16.pt"). The other model files are downloaded in the same way.

How should I fix this error? Is there a mistake in the way I download the model? Or is the problem with the network.py file?

One more question. Is there a requirement for the order in which the .sh files are run? I am talking about the following .sh files. run_all_baseline_finetuning.sh run_all_lwf.sh run_all_sequence.sh

Thanks