y-kawagu / dcase2020_task2_baseline Goto Github PK

============== MODEL TRAINING ==============

      Layer (type)                 Output Shape              Param #   

     input_1 (InputLayer)         (None, 128, 5, 1)         0         

    conv2d_1 (Conv2D)            (None, 128, 5, 32)        320       

    max_pooling2d_1 (MaxPooling2 (None, 64, 3, 32)         0         

    conv2d_2 (Conv2D)            (None, 64, 3, 16)         4624      

    max_pooling2d_2 (MaxPooling2 (None, 32, 2, 16)         0         
   
    conv2d_3 (Conv2D)            (None, 32, 2, 8)          1160      

    max_pooling2d_3 (MaxPooling2 (None, 16, 1, 8)          0         

    conv2d_4 (Conv2D)            (None, 16, 1, 8)          584       

    up_sampling2d_1 (UpSampling2 (None, 32, 2, 8)          0         

    conv2d_5 (Conv2D)            (None, 32, 2, 16)         1168      

    up_sampling2d_2 (UpSampling2 (None, 64, 4, 16)         0         

     conv2d_6 (Conv2D)            (None, 64, 4, 32)         4640      

     up_sampling2d_3 (UpSampling2 (None, 128, 8, 32)        0         

     conv2d_7 (Conv2D)            (None, 128, 8, 1)         33        

Total params: 12,529
Trainable params: 12,529
Non-trainable params: 0

Traceback (most recent call last):
File "00_train.py", line 211, in
verbose=param["fit"]["verbose"])
File "/usr/local/lib/python3.6/dist-packages/keras/engine/training.py", line 1630, in fit
batch_size=batch_size)
File "/usr/local/lib/python3.6/dist-packages/keras/engine/training.py", line 1480, in _standardize_user_data
exception_prefix='target')
File "/usr/local/lib/python3.6/dist-packages/keras/engine/training.py", line 123, in _standardize_input_data
str(data_shape))
ValueError: Error when checking target: expected conv2d_7 to have shape (128, 8, 1) but got array with shape (128, 5, 1)

I tried using the convolution2D and LSTM autoencoder so made this change to split the input layer

model = keras_model.get_model(param["feature"]["n_mels"] ,param["feature"]["frames"]) so I will have input as (128,5,1) 1-is channel

but I get this error
ValueError: Error when checking input: expected input_1 to have 4 dimensions, but got array with shape (1360000, 640)

Hi, Yohei, it is a fantastic work. Can I ask some questions about the requirements for training the model.

1. What is the least requirements for a pc? My pc has a memory of 8G, and 1050Ti (4G GPU memory). Even though I changed the batch_size to 1, it still ran out of memory. From my experience, my pc could train the model with only 267k parameters.
2. My pc can not run the model according to the 'requirements.txt'. I did the following changes then it works. I have no idea whether it is common or only for me. I hope this will help.
   Keras-Preprocessing==1.1.0
   gast==0.3.3 (It always reminds me that gast==0.3.3 is imcompatible with tensorflow==1.15.)