In [1]:

    

from tfs.network import CustomNetwork
from tfs.dataset import Mnist


    

In [2]:

    

class MyNet(CustomNetwork):
    def setup(self):
        self.default_in_shape = [None,28,28,1]
        self.loss_input_layer_name = 'fc3'  # this is because we may not use the last layer to compute a loss.
        (self.nodes
            .fc(100,name='fc1')
            .fc(100,name='fc2')
            .fc(10, activation=None,name='fc3')
            .softmax(name='prob')
        )
        
net = MyNet()
net.build()
dataset = Mnist()
net.fit(dataset,batch_size=200,n_epoch=2)


    

    




step 10. loss 9.848982, score:0.557900
step 20. loss 5.812720, score:0.719800
step 30. loss 5.057542, score:0.778100
step 40. loss 4.946941, score:0.809300
step 50. loss 1.830094, score:0.823900
step 60. loss 2.381712, score:0.833700
step 70. loss 3.349250, score:0.829500
step 80. loss 3.084479, score:0.835300
step 90. loss 1.218812, score:0.856400
step 100. loss 2.914572, score:0.867000
step 110. loss 1.212699, score:0.865800
step 120. loss 1.842257, score:0.868400
step 130. loss 1.593982, score:0.859500
step 140. loss 1.379951, score:0.865100
step 150. loss 1.896303, score:0.864300
step 160. loss 1.409557, score:0.871400
step 170. loss 0.490144, score:0.884400
step 180. loss 0.882401, score:0.889500
step 190. loss 1.261182, score:0.888400
step 200. loss 0.895965, score:0.889900
step 210. loss 1.267546, score:0.891500
step 220. loss 1.222363, score:0.890000
step 230. loss 1.134108, score:0.885900
step 240. loss 0.932302, score:0.891800
step 250. loss 1.216363, score:0.896000
step 260. loss 0.928397, score:0.888200
step 270. loss 1.388842, score:0.893800
step 280. loss 0.751991, score:0.898700
step 290. loss 0.415645, score:0.903800
step 300. loss 0.511151, score:0.902900
step 310. loss 0.766532, score:0.899900
step 320. loss 0.810046, score:0.909600
step 330. loss 0.416960, score:0.907100
step 340. loss 0.735516, score:0.903900
step 350. loss 0.266978, score:0.911300
step 360. loss 0.734636, score:0.908300
step 370. loss 0.873375, score:0.908900
step 380. loss 0.387011, score:0.907200
step 390. loss 0.857715, score:0.910400
step 400. loss 0.492993, score:0.903500
step 410. loss 0.463242, score:0.911900
step 420. loss 0.709660, score:0.911200
step 430. loss 0.537762, score:0.908300
step 440. loss 0.377251, score:0.911100
step 450. loss 0.716996, score:0.914400
step 460. loss 0.443953, score:0.911800
step 470. loss 0.267376, score:0.916500
step 480. loss 0.467868, score:0.913500
step 490. loss 0.505337, score:0.912200
step 500. loss 0.616280, score:0.915600
step 510. loss 0.476829, score:0.914000
step 520. loss 0.406835, score:0.913600
step 530. loss 0.441625, score:0.920200
step 540. loss 0.345150, score:0.921700
step 550. loss 0.202693, score:0.916300
step 560. loss 0.275747, score:0.919400
step 570. loss 0.399429, score:0.918100
step 580. loss 0.296891, score:0.923000
step 590. loss 0.518799, score:0.918700
step 600. loss 0.469096, score:0.919800






    
Out[2]:





<__main__.MyNet at 0x114dfec50>

In [ ]: