Sparsify Callback

Use the sparsifier in fastai Callback system

from fastai.vision.all import *

path = untar_data(URLs.PETS)
files = get_image_files(path/"images")

def label_func(f): return f[0].isupper()

dls = ImageDataLoaders.from_name_func(path, files, label_func, item_tfms=Resize(64))

The most important part of our Callback happens in before_batch. There, we first compute the sparsity of our network according to our schedule and then we remove the parameters accordingly.

import timm
pretrained_resnet_34 = timm.create_model('resnet34', pretrained=True)

learn = Learner(dls, pretrained_resnet_34, metrics=accuracy)
learn.fc = nn.Linear(512, 2)
sp_cb = SparsifyCallback(sparsity=50, granularity='filter', context='global', criteria=large_final, schedule=one_cycle)
learn.fit_one_cycle(5, cbs=sp_cb)

Pruning of filter until a sparsity of [50]%
Saving Weights at epoch 0
Sparsity at the end of epoch 0: [1.96]%
Sparsity at the end of epoch 1: [20.07]%
Sparsity at the end of epoch 2: [45.86]%
Sparsity at the end of epoch 3: [49.74]%
Sparsity at the end of epoch 4: [50.0]%
Final Sparsity: [50.0]%
Sparsity in Conv2d 1: 50.00%
Sparsity in Conv2d 7: 53.12%
Sparsity in Conv2d 12: 50.47%
Sparsity in Conv2d 16: 50.00%
Sparsity in Conv2d 21: 50.53%
Sparsity in Conv2d 25: 50.00%
Sparsity in Conv2d 30: 50.41%
Sparsity in Conv2d 35: 50.00%
Sparsity in Conv2d 40: 50.24%
Sparsity in Conv2d 44: 50.00%
Sparsity in Conv2d 47: 50.00%
Sparsity in Conv2d 52: 50.24%
Sparsity in Conv2d 56: 50.00%
Sparsity in Conv2d 61: 50.38%
Sparsity in Conv2d 65: 50.00%
Sparsity in Conv2d 70: 50.11%
Sparsity in Conv2d 75: 50.00%
Sparsity in Conv2d 80: 50.12%
Sparsity in Conv2d 84: 50.00%
Sparsity in Conv2d 87: 50.00%
Sparsity in Conv2d 92: 50.09%
Sparsity in Conv2d 96: 50.00%
Sparsity in Conv2d 101: 50.10%
Sparsity in Conv2d 105: 50.00%
Sparsity in Conv2d 110: 50.03%
Sparsity in Conv2d 114: 50.00%
Sparsity in Conv2d 119: 50.11%
Sparsity in Conv2d 123: 50.00%
Sparsity in Conv2d 128: 50.12%
Sparsity in Conv2d 133: 50.00%
Sparsity in Conv2d 138: 50.08%
Sparsity in Conv2d 142: 50.00%
Sparsity in Conv2d 145: 50.01%
Sparsity in Conv2d 150: 50.14%
Sparsity in Conv2d 154: 50.00%
Sparsity in Conv2d 159: 50.28%

epoch	train_loss	valid_loss	accuracy	time
0	0.434834	0.498019	0.758457	00:07
1	0.409158	0.432628	0.809202	00:06
2	0.338707	0.371978	0.832206	00:07
3	0.274749	0.368416	0.854533	00:06
4	0.237638	0.373818	0.849797	00:06

learn.model.conv1.weight

Parameter containing:
tensor([[[[ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          ...,
          [ 0.0000e+00,  0.0000e+00, -0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [-0.0000e+00, -0.0000e+00, -0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00, -0.0000e+00, -0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00]],

         [[ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          ...,
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00, -0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00, -0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00]],

         [[ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          ...,
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [-0.0000e+00, -0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00, -0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00]]],


        [[[ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          ...,
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00]],

         [[ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00, -0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00, -0.0000e+00],
          ...,
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00, -0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00]],

         [[ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
           -0.0000e+00, -0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
           -0.0000e+00, -0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ..., -0.0000e+00,
           -0.0000e+00, -0.0000e+00],
          ...,
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
           -0.0000e+00, -0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00, -0.0000e+00, -0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00]]],


        [[[ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [-0.0000e+00, -0.0000e+00, -0.0000e+00,  ...,  0.0000e+00,
           -0.0000e+00,  0.0000e+00],
          ...,
          [-0.0000e+00, -0.0000e+00, -0.0000e+00,  ..., -0.0000e+00,
           -0.0000e+00, -0.0000e+00],
          [-0.0000e+00, -0.0000e+00, -0.0000e+00,  ..., -0.0000e+00,
           -0.0000e+00, -0.0000e+00],
          [-0.0000e+00, -0.0000e+00, -0.0000e+00,  ..., -0.0000e+00,
           -0.0000e+00, -0.0000e+00]],

         [[ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ..., -0.0000e+00,
           -0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ..., -0.0000e+00,
           -0.0000e+00,  0.0000e+00],
          ...,
          [-0.0000e+00, -0.0000e+00, -0.0000e+00,  ..., -0.0000e+00,
           -0.0000e+00, -0.0000e+00],
          [-0.0000e+00, -0.0000e+00, -0.0000e+00,  ..., -0.0000e+00,
           -0.0000e+00, -0.0000e+00],
          [-0.0000e+00, -0.0000e+00, -0.0000e+00,  ..., -0.0000e+00,
           -0.0000e+00, -0.0000e+00]],

         [[ 0.0000e+00,  0.0000e+00,  0.0000e+00,  ...,  0.0000e+00,
            0.0000e+00,  0.0000e+00],
          [ 0.0000e+00,  0.0000e+00, -0.0000e+00,  ..., -0.0000e+00,
           -0.0000e+00, -0.0000e+00],
          [-0.0000e+00, -0.0000e+00, -0.0000e+00,  ..., -0.0000e+00,
           -0.0000e+00, -0.0000e+00],
          ...,
          [-0.0000e+00, -0.0000e+00, -0.0000e+00,  ..., -0.0000e+00,
           -0.0000e+00, -0.0000e+00],
          [-0.0000e+00, -0.0000e+00, -0.0000e+00,  ..., -0.0000e+00,
           -0.0000e+00, -0.0000e+00],
          [-0.0000e+00, -0.0000e+00, -0.0000e+00,  ..., -0.0000e+00,
           -0.0000e+00, -0.0000e+00]]],


        ...,


        [[[-7.4296e-03, -7.5463e-03, -8.3205e-03,  ..., -7.1928e-02,
            1.5917e-01,  1.2489e-01],
          [ 1.0137e-01, -5.4320e-03, -7.5936e-03,  ..., -4.2560e-01,
           -1.0976e-02,  1.9817e-01],
          [ 1.2293e-01,  2.8018e-01,  3.2716e-01,  ..., -6.7938e-01,
           -4.9350e-01, -1.6077e-02],
          ...,
          [-1.8629e-01, -1.9095e-01, -2.0495e-02,  ...,  5.9457e-01,
           -1.7923e-02, -2.3116e-01],
          [-1.1156e-01, -2.5112e-01, -3.7146e-01,  ...,  3.9388e-01,
            2.8940e-01, -1.7420e-02],
          [-9.4251e-03, -1.6436e-02, -3.3763e-01,  ..., -1.1631e-02,
            1.7691e-01,  1.0555e-01]],

         [[-8.8509e-03, -6.3883e-03, -5.0388e-03,  ...,  4.3923e-03,
            1.0098e-01,  6.4890e-03],
          [-7.3702e-03, -2.3618e-03, -2.8961e-03,  ..., -2.6874e-01,
           -2.7887e-03,  1.6050e-01],
          [-5.0071e-03,  1.6799e-01,  2.7184e-01,  ..., -5.8726e-01,
           -3.1158e-01, -6.8676e-03],
          ...,
          [-1.5120e-01, -2.3115e-01, -1.0500e-01,  ...,  4.9077e-01,
           -8.2684e-03, -1.3619e-01],
          [-7.1860e-03, -1.8193e-01, -3.3559e-01,  ...,  2.7722e-01,
            1.6240e-01, -9.3599e-03],
          [-2.4150e-03, -7.1116e-03, -1.8751e-01,  ..., -1.6245e-04,
            1.1394e-01, -8.0694e-03]],

         [[-3.9283e-03, -3.3088e-03, -3.5065e-03,  ...,  3.9558e-03,
            1.9028e-03,  3.6099e-03],
          [-1.2535e-03,  1.3431e-03, -7.3769e-04,  ..., -2.8860e-03,
           -3.9271e-03, -4.3433e-03],
          [ 1.9957e-03,  6.6360e-03,  2.0233e-03,  ..., -1.9471e-01,
           -7.5721e-03, -8.9046e-03],
          ...,
          [-4.8502e-03, -9.2381e-02, -7.7613e-02,  ...,  1.2395e-01,
           -1.0181e-02, -9.2624e-03],
          [ 1.8123e-03, -3.5798e-03, -1.1415e-01,  ..., -5.0590e-03,
           -8.1797e-03, -9.5528e-03],
          [ 5.2121e-03, -2.0169e-04,  1.2455e-03,  ...,  1.7431e-03,
           -4.9568e-03, -8.3131e-03]]],


        [[[ 7.8349e-03,  1.3473e-02,  8.7682e-02,  ...,  1.4227e-01,
            1.4403e-02,  1.1514e-02],
          [ 5.7151e-03,  1.5254e-01, -1.6662e-01,  ..., -1.9651e-01,
            9.3270e-03,  9.1108e-03],
          [ 8.3750e-02, -6.4736e-02, -7.7611e-02,  ...,  1.6225e-01,
           -2.5000e-01,  2.0282e-01],
          ...,
          [-9.0023e-02,  1.0453e-01, -1.6298e-01,  ..., -1.1992e-01,
            3.8632e-03, -1.2307e-01],
          [ 1.1366e-02,  1.7510e-01, -2.6541e-01,  ...,  5.9963e-03,
           -2.4098e-01,  2.6635e-01],
          [ 9.7180e-03,  1.6214e-01, -2.6019e-01,  ..., -1.3749e-01,
           -6.4509e-02, -1.0232e-03]],

         [[ 1.8908e-02,  2.6516e-02, -8.7681e-02,  ...,  2.7610e-02,
            2.3307e-02,  1.2432e-01],
          [ 1.5978e-02,  1.7737e-02,  2.3648e-02,  ...,  2.1288e-02,
            1.9111e-02, -1.5350e-01],
          [ 1.8378e-02, -8.8506e-02,  3.1728e-01,  ..., -2.9368e-01,
            5.7689e-01, -1.4814e-01],
          ...,
          [-7.0136e-02,  2.9066e-02, -6.5593e-02,  ...,  3.0752e-03,
           -3.0135e-01,  1.3104e-01],
          [ 1.6101e-02,  2.9628e-01, -1.2312e-01,  ...,  1.9397e-01,
            2.5778e-01,  5.0888e-03],
          [ 1.4030e-02, -2.0212e-01,  3.2355e-01,  ...,  7.0312e-03,
            8.5274e-03, -1.1626e-01]],

         [[ 1.8303e-02,  2.4612e-02,  2.9882e-02,  ..., -9.8176e-02,
            2.3368e-02,  1.9999e-02],
          [ 1.5427e-02,  1.6103e-02,  1.4962e-01,  ...,  2.1504e-01,
            1.8854e-02,  1.6187e-01],
          [-5.3391e-02,  1.4294e-01, -2.2367e-01,  ...,  1.3511e-02,
           -1.8316e-01, -8.3739e-02],
          ...,
          [ 2.1427e-01, -1.5982e-01,  1.8060e-01,  ..., -9.3084e-02,
            4.8867e-01, -1.2281e-01],
          [ 9.6726e-03, -3.7949e-01,  3.9740e-01,  ..., -1.5790e-01,
           -1.5290e-01, -1.8989e-01],
          [ 7.3821e-03,  7.1633e-03,  2.5753e-03,  ..., -3.3214e-03,
            1.5466e-01, -6.4393e-03]]],


        [[[ 1.4643e-01,  5.8291e-03, -9.3539e-02,  ..., -1.0934e-01,
           -6.7021e-02,  1.4150e-02],
          [-1.4916e-01, -3.0628e-01, -3.8497e-01,  ..., -3.3362e-01,
           -2.2944e-01, -1.6075e-01],
          [-1.6706e-01, -2.8656e-01, -3.0256e-01,  ..., -2.0954e-01,
           -1.3465e-01,  9.8437e-03],
          ...,
          [ 5.5858e-03,  6.4734e-03,  1.8051e-01,  ...,  3.8381e-01,
            3.5352e-01,  2.8040e-01],
          [ 8.6895e-03,  1.1563e-01,  2.1870e-01,  ...,  3.3523e-01,
            2.3642e-01,  1.5989e-01],
          [ 9.8079e-02,  1.6960e-01,  2.1411e-01,  ...,  1.8952e-01,
            6.1990e-03,  1.1999e-02]],

         [[ 1.7169e-01,  3.0308e-03, -1.3982e-01,  ..., -1.6779e-01,
           -1.0978e-01,  7.9350e-03],
          [-1.9304e-01, -4.0593e-01, -5.3100e-01,  ..., -4.6119e-01,
           -3.2400e-01, -2.0214e-01],
          [-2.8363e-01, -4.3816e-01, -4.9162e-01,  ..., -3.1524e-01,
           -2.1846e-01, -8.9559e-02],
          ...,
          [ 4.0925e-03,  2.3280e-03,  2.0114e-01,  ...,  4.4946e-01,
            4.0134e-01,  3.4598e-01],
          [ 1.1379e-01,  1.8643e-01,  3.1593e-01,  ...,  4.7618e-01,
            3.8119e-01,  3.1284e-01],
          [ 1.8202e-01,  2.5852e-01,  3.0201e-01,  ...,  3.2789e-01,
            1.8261e-01,  1.0490e-01]],

         [[ 1.4723e-01, -1.3924e-02, -1.1114e-02,  ..., -3.3590e-03,
           -3.8765e-03, -9.1093e-04],
          [-1.7495e-02, -1.6472e-01, -2.6806e-01,  ..., -2.4088e-01,
           -1.9225e-01, -1.3964e-01],
          [-1.1816e-01, -2.2255e-01, -2.8439e-01,  ..., -1.7405e-01,
           -1.1457e-01, -9.2854e-03],
          ...,
          [-1.5639e-02, -1.7753e-02, -2.0895e-02,  ...,  2.2529e-01,
            1.7778e-01,  1.4106e-01],
          [-1.3705e-02, -1.7075e-02,  9.4111e-02,  ...,  2.2573e-01,
            1.6869e-01,  1.2757e-01],
          [ 1.0739e-01,  1.3187e-01,  1.1567e-01,  ...,  1.6808e-01,
            1.1086e-01,  8.2069e-02]]]], device='cuda:0', requires_grad=True)

learn = vision_learner(dls, resnet18, metrics=accuracy)
learn.unfreeze()

learn.fit_one_cycle(5)

epoch	train_loss	valid_loss	accuracy	time
0	0.711885	1.064277	0.843708	00:45
1	0.409735	0.217008	0.913396	00:03
2	0.265280	0.284833	0.898512	00:03
3	0.144334	0.158726	0.936401	00:03
4	0.082726	0.153889	0.939784	00:03

Let’s now try adding some sparsity in our model

learn = vision_learner(dls, resnet18, metrics=accuracy)
learn.unfreeze()

The SparsifyCallback requires a new argument compared to the Sparsifier. Indeed, we need to know the pruning schedule that we should follow during training in order to prune the parameters accordingly.

You can use any scheduling function already available in fastai or come up with your own ! For more information about the pruning schedules, take a look at the Schedules section.

sp_cb = SparsifyCallback(sparsity=50, granularity='weight', context='local', criteria=large_final, schedule=one_cycle)

learn.fit_one_cycle(5, cbs=sp_cb)

Pruning of weight until a sparsity of [50]%
Saving Weights at epoch 0
Sparsity at the end of epoch 0: [1.96]%
Sparsity at the end of epoch 1: [20.07]%
Sparsity at the end of epoch 2: [45.86]%
Sparsity at the end of epoch 3: [49.74]%
Sparsity at the end of epoch 4: [50.0]%
Final Sparsity: [50.0]%
Sparsity in Conv2d 2: 50.00%
Sparsity in Conv2d 8: 50.00%
Sparsity in Conv2d 11: 50.00%
Sparsity in Conv2d 14: 50.00%
Sparsity in Conv2d 17: 50.00%
Sparsity in Conv2d 21: 50.00%
Sparsity in Conv2d 24: 50.00%
Sparsity in Conv2d 27: 50.00%
Sparsity in Conv2d 30: 50.00%
Sparsity in Conv2d 33: 50.00%
Sparsity in Conv2d 37: 50.00%
Sparsity in Conv2d 40: 50.00%
Sparsity in Conv2d 43: 50.00%
Sparsity in Conv2d 46: 50.00%
Sparsity in Conv2d 49: 50.00%
Sparsity in Conv2d 53: 50.00%
Sparsity in Conv2d 56: 50.00%
Sparsity in Conv2d 59: 50.00%
Sparsity in Conv2d 62: 50.00%
Sparsity in Conv2d 65: 50.00%

epoch	train_loss	valid_loss	accuracy	time
0	0.711270	0.742762	0.775372	00:07
1	0.383374	0.307700	0.864005	00:07
2	0.219235	0.217708	0.905954	00:07
3	0.121921	0.213659	0.933018	00:07
4	0.067208	0.200506	0.930988	00:07

Surprisingly, our network that is composed of \(50 \%\) of zeroes performs reasonnably well when compared to our plain and dense network.

The SparsifyCallback also accepts a list of sparsities, corresponding to each layer of layer_type to be pruned. Below, we show how to prune only the intermediate layers of ResNet-18.

learn = vision_learner(dls, resnet18, metrics=accuracy)
learn.unfreeze()

sparsities = [0, 0, 0, 0, 0, 0, 50, 50, 50, 50, 50, 50, 50, 50, 0, 0, 0, 0, 0, 0]

sp_cb = SparsifyCallback(sparsity=sparsities, granularity='weight', context='local', criteria=large_final, schedule=cos)

learn.fit_one_cycle(5, cbs=sp_cb)

Pruning of weight until a sparsity of [0, 0, 0, 0, 0, 0, 50, 50, 50, 50, 50, 50, 50, 50, 0, 0, 0, 0, 0, 0]%
Saving Weights at epoch 0
Sparsity at the end of epoch 0: [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 4.77, 4.77, 4.77, 4.77, 4.77, 4.77, 4.77, 4.77, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]%
Sparsity at the end of epoch 1: [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 17.27, 17.27, 17.27, 17.27, 17.27, 17.27, 17.27, 17.27, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]%
Sparsity at the end of epoch 2: [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 32.73, 32.73, 32.73, 32.73, 32.73, 32.73, 32.73, 32.73, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]%
Sparsity at the end of epoch 3: [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 45.23, 45.23, 45.23, 45.23, 45.23, 45.23, 45.23, 45.23, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]%
Sparsity at the end of epoch 4: [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 50.0, 50.0, 50.0, 50.0, 50.0, 50.0, 50.0, 50.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]%
Final Sparsity: [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 50.0, 50.0, 50.0, 50.0, 50.0, 50.0, 50.0, 50.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]%
Sparsity in Conv2d 2: 0.00%
Sparsity in Conv2d 8: 0.00%
Sparsity in Conv2d 11: 0.00%
Sparsity in Conv2d 14: 0.00%
Sparsity in Conv2d 17: 0.00%
Sparsity in Conv2d 21: 0.00%
Sparsity in Conv2d 24: 50.00%
Sparsity in Conv2d 27: 50.00%
Sparsity in Conv2d 30: 50.00%
Sparsity in Conv2d 33: 50.00%
Sparsity in Conv2d 37: 50.00%
Sparsity in Conv2d 40: 50.00%
Sparsity in Conv2d 43: 50.00%
Sparsity in Conv2d 46: 50.00%
Sparsity in Conv2d 49: 0.00%
Sparsity in Conv2d 53: 0.00%
Sparsity in Conv2d 56: 0.00%
Sparsity in Conv2d 59: 0.00%
Sparsity in Conv2d 62: 0.00%
Sparsity in Conv2d 65: 0.00%

epoch	train_loss	valid_loss	accuracy	time
0	0.731650	0.570400	0.811908	00:07
1	0.396108	0.262083	0.895805	00:07
2	0.250992	0.210679	0.909337	00:07
3	0.132799	0.192091	0.925575	00:07
4	0.079732	0.159255	0.938430	00:07

On top of that, the SparsifyCallbackcan also take many optionnal arguments:

lth: whether training using the Lottery Ticket Hypothesis, i.e. reset the weights to their original value at each pruning step (more information in the Lottery Ticket Hypothesis section)
rewind_epoch: the epoch used as a reference for the Lottery Ticket Hypothesis with Rewinding (default to 0)
reset_end: whether you want to reset the weights to their original values after training (pruning masks are still applied)
save_tickets: whether to save intermediate winning tickets.
model: pass a model or a part of the model if you don’t want to apply pruning on the whole model trained.
round_to: if specified, the weights will be pruned to the closest multiple value of round_to.
layer_type: specify the type of layer that you want to apply pruning to (default to nn.Conv2d)`

For example, we correctly pruned the convolution layers of our model, but we could imagine pruning the Linear Layers of even only the BatchNorm ones !