Feature/#33

[ssaru]

Multipl-GPU에서의 학습은 아직 에러가 있지만, Single GPU를 먼저두고 진행하겠습니다.

데모코드 확인은 아래와 같은 명령어를 이용하여 확인할 수 있습니다. 프로젝트 폴더 내에서

$ python3 models/binarized_mlp.py
>>>
----------------------------------------------------------------
        Layer (type)               Output Shape         Param #
================================================================
   BinarizedLinear-1                  [-1, 512]         401,408
       BatchNorm1d-2                  [-1, 512]           1,024
   BinarizedLinear-3                  [-1, 512]         262,144
       BatchNorm1d-4                  [-1, 512]           1,024
           Dropout-5                  [-1, 512]               0
   BinarizedLinear-6                   [-1, 10]           5,120
================================================================
Total params: 670,720
Trainable params: 670,720
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 0.00
Forward/backward pass size (MB): 0.02
Params size (MB): 2.56
Estimated Total Size (MB): 2.58
----------------------------------------------------------------
      Name             Type Params
0      fc1  BinarizedLinear  401 K
1  dropout          Dropout    0
2   batch1      BatchNorm1d    1 K
3      fc2  BinarizedLinear  262 K
4   batch2      BatchNorm1d    1 K
5      fc3  BinarizedLinear    5 K
1922it [00:33, 57.70it/s, batch_nb=46, epoch=0, loss=31.976, v_nb=2]                                                                                        Name             Type Params
0      fc1  BinarizedLinear  401 K
1  dropout          Dropout    0
2   batch1      BatchNorm1d    1 K
3      fc2  BinarizedLinear  262 K
4   batch2      BatchNorm1d    1 K
5      fc3  BinarizedLinear    5 K
model and trainer restored from checkpoint: /Users/martin/Documents/dev/Seoultech/BNN/BinaryConnect/lightning_logs/version_2/checkpoints/_ckpt_epoch_1.ckpt
257it [00:04, 57.32it/s]

[chromatices]

확인했습니다

[ssaru]

approve 눌러주세요!