IndexError: list index out of range #4

sauravtii · 2022-09-04T18:41:30Z

I have trained and saved a model and want to convert it to SNN by using SpKeras, but after running the code to convert I am facing the following issue:

---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
Input In [10], in <cell line: 7>()
      2 from spkeras.spkeras.models import cnn_to_snn
      4 #Current normalisation using cnn_to_snn
      5 ##Default: signed_bit=0, amp_factor=100, method=1, epsilon = 0.001
----> 7 snn_model = cnn_to_snn(signed_bit=0)(cnn_model,x_train)

File ~/Desktop/Toy network/spkeras/spkeras/models.py:29, in cnn_to_snn.__call__(self, mdl, x_train)
     27 self.use_bias = use_bias        
     28 self.get_config()
---> 29 self.model = self.convert(mdl,x_train,                    
     30                           thresholding = self.thresholding,
     31                           scaling_factor = self.scaling_factor,
     32                           method = self.method,
     33                           timesteps=self.timesteps)
     35 return self

File ~/Desktop/Toy network/spkeras/spkeras/models.py:101, in cnn_to_snn.convert(self, mdl, x_train, thresholding, scaling_factor, method, timesteps)
     98     _weights[0] = _weights[0].astype(int)   
     99     _weights[0] = _weights[0]/2**bit
--> 101 _bias = kappa*_weights[1]/lmax[num+1]
    102 _bias = _bias/norm
    103 bias.append(_bias.tolist())    

IndexError: list index out of range

My model:

Model: "model"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 input_1 (InputLayer)        [(None, 32, 32, 3)]       0         
                                                                 
 conv2d (Conv2D)             (None, 32, 32, 4)         16        
                                                                 
 conv2d_1 (Conv2D)           (None, 16, 16, 64)        2368      
                                                                 
 conv2d_2 (Conv2D)           (None, 16, 16, 72)        41544     
                                                                 
 conv2d_3 (Conv2D)           (None, 8, 8, 256)         166144    
                                                                 
 conv2d_4 (Conv2D)           (None, 8, 8, 256)         65792     
                                                                 
 conv2d_5 (Conv2D)           (None, 8, 8, 64)          16448     
                                                                 
 flatten (Flatten)           (None, 4096)              0         
                                                                 
 dropout (Dropout)           (None, 4096)              0         
                                                                 
 dense (Dense)               (None, 100)               409700    
                                                                 
 dense_1 (Dense)             (None, 10)                1010      
                                                                 
=================================================================
Total params: 703,022
Trainable params: 703,022
Non-trainable params: 0

Any solutions ?

I also tried solving it by referring to this, but I wasn't able to.

The text was updated successfully, but these errors were encountered:

Dengyu-Wu · 2022-09-05T12:05:26Z

Try adding the Activation Layer after Conv2D and Dense, such as,

inputs = Input((32, 32, 3))
x = Conv2D()(inputs)
x = Activation("relu")(x)

The problem is due to that SpKeras counts Activation layer for conversion.

sauravtii · 2022-09-06T06:52:44Z

I did that, but still facing the same issue.

My code:

input_shape = (32, 32, 3)
input_layer = Input(input_shape)

layer = Conv2D(filters=4,
               kernel_size=(1, 1),
               strides=(1, 1),
               padding="same", use_bias=False)(input_layer)

layer = tf.keras.activations.relu(layer, alpha=0.0, max_value=None, threshold=0.0)


layer = Conv2D(filters=64,
               kernel_size=(3, 3),
               strides=(2, 2),
               padding="same", use_bias=False)(layer)


layer = Conv2D(filters=72,
               kernel_size=(3, 3),
               strides=(1, 1),
               padding="same", use_bias=False)(layer)

layer = tf.keras.activations.relu(layer, alpha=0.0, max_value=None, threshold=0.0)


layer = Conv2D(filters=256,
               kernel_size=(3, 3),
               strides=(2, 2),
               padding="same", use_bias=False)(layer)


layer = Conv2D(filters=256,
               kernel_size=(1, 1),
               strides=(1, 1),
               padding="same", use_bias=False)(layer)

layer = tf.keras.activations.relu(layer, alpha=0.0, max_value=None, threshold=0.0)


layer = Conv2D(filters=64,
               kernel_size=(1, 1),
               strides=(1, 1),
               padding="same", use_bias=False)(layer)


layer = Flatten()(layer)

layer = Dropout(0.5)(layer)

layer = Dense(units=100)(layer)

layer = tf.keras.activations.relu(layer, alpha=0.0, max_value=None, threshold=0.0)

layer = Dense(units=10)(layer)

layer = tf.keras.activations.softmax(layer, axis=-1)

model = Model(input_layer, layer)

My model summary:

Model: "model"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 input_1 (InputLayer)        [(None, 32, 32, 3)]       0         
                                                                 
 conv2d (Conv2D)             (None, 32, 32, 4)         12        
                                                                 
 tf.nn.relu (TFOpLambda)     (None, 32, 32, 4)         0         
                                                                 
 conv2d_1 (Conv2D)           (None, 16, 16, 64)        2304      
                                                                 
 conv2d_2 (Conv2D)           (None, 16, 16, 72)        41472     
                                                                 
 tf.nn.relu_1 (TFOpLambda)   (None, 16, 16, 72)        0         
                                                                 
 conv2d_3 (Conv2D)           (None, 8, 8, 256)         165888    
                                                                 
 conv2d_4 (Conv2D)           (None, 8, 8, 256)         65536     
                                                                 
 tf.nn.relu_2 (TFOpLambda)   (None, 8, 8, 256)         0         
                                                                 
 conv2d_5 (Conv2D)           (None, 8, 8, 64)          16384     
                                                                 
 flatten (Flatten)           (None, 4096)              0         
                                                                 
 dropout (Dropout)           (None, 4096)              0         
                                                                 
 dense (Dense)               (None, 100)               409700    
                                                                 
 tf.nn.relu_3 (TFOpLambda)   (None, 100)               0         
                                                                 
 dense_1 (Dense)             (None, 10)                1010      
                                                                 
 tf.nn.softmax (TFOpLambda)  (None, 10)                0         
                                                                 
=================================================================
Total params: 702,306
Trainable params: 702,306
Non-trainable params: 0
_____________________________

I also tried adding activation after each Conv2D layer, but faced the same issue.

My code:

input_shape = (32, 32, 3)
input_layer = Input(input_shape)

layer = Conv2D(filters=4,
               kernel_size=(1, 1),
               strides=(1, 1),
               padding="same", use_bias=False)(input_layer)

layer = tf.keras.activations.relu(layer, alpha=0.0, max_value=None, threshold=0.0)


layer = Conv2D(filters=64,
               kernel_size=(3, 3),
               strides=(2, 2),
               padding="same", use_bias=False)(layer)

layer = tf.keras.activations.relu(layer, alpha=0.0, max_value=None, threshold=0.0)


layer = Conv2D(filters=72,
               kernel_size=(3, 3),
               strides=(1, 1),
               padding="same", use_bias=False)(layer)

layer = tf.keras.activations.relu(layer, alpha=0.0, max_value=None, threshold=0.0)


layer = Conv2D(filters=256,
               kernel_size=(3, 3),
               strides=(2, 2),
               padding="same", use_bias=False)(layer)

layer = tf.keras.activations.relu(layer, alpha=0.0, max_value=None, threshold=0.0)

layer = Conv2D(filters=256,
               kernel_size=(1, 1),
               strides=(1, 1),
               padding="same", use_bias=False)(layer)

layer = tf.keras.activations.relu(layer, alpha=0.0, max_value=None, threshold=0.0)


layer = Conv2D(filters=64,
               kernel_size=(1, 1),
               strides=(1, 1),
               padding="same", use_bias=False)(layer)

layer = tf.keras.activations.relu(layer, alpha=0.0, max_value=None, threshold=0.0)

layer = Flatten()(layer)

layer = Dropout(0.5)(layer)

layer = Dense(units=100)(layer)

layer = tf.keras.activations.relu(layer, alpha=0.0, max_value=None, threshold=0.0)

layer = Dense(units=10)(layer)

layer = tf.keras.activations.softmax(layer, axis=-1)

model = Model(input_layer, layer)

Dengyu-Wu · 2022-09-06T08:58:17Z

Can you change
layer = tf.keras.activations.relu(layer, alpha=0.0, max_value=None, threshold=0.0)
into
layer=Activation('relu')(layer)
and try again.

sauravtii · 2022-09-06T11:41:38Z

It worked, thanks!
But after evaluating, the accuracy is not good.

Can you help with this ?

Changing model timesteps...
New model generated!
{'timesteps': 256, 'thresholding': 0.5, 'amp_factor': 100, 'signed_bit': 0, 'spike_ext': 0, 'epsilon': 0.001, 'use_bias': False, 'scaling_factor': 1, 'noneloss': False, 'method': 1}
313/313 [==============================] - 3s 9ms/step - loss: 163.4400 - accuracy: 0.1000

I am running it for 25 epochs and getting around 83 % of train accuracy.

Dengyu-Wu · 2022-09-06T11:59:38Z

Have you changed
layer = tf.keras.activations.softmax(layer, axis=-1)
into
layer=Activation('softmax')(layer)
as well

sauravtii · 2022-09-06T12:25:47Z

Yes, below is my code for the architecture:

input_shape = (32, 32, 3)
input_layer = Input(input_shape)

layer = Conv2D(filters=4,
               kernel_size=(1, 1),
               strides=(1, 1),
               padding="same", use_bias=False)(input_layer)

layer=Activation('relu')(layer)

layer = Conv2D(filters=64,
               kernel_size=(3, 3),
               strides=(2, 2),
               padding="same", use_bias=False)(layer)

layer=Activation('relu')(layer)

layer = Conv2D(filters=72,
               kernel_size=(3, 3),
               strides=(1, 1),
               padding="same", use_bias=False)(layer)

layer=Activation('relu')(layer)

layer = Conv2D(filters=256,
               kernel_size=(3, 3),
               strides=(2, 2),
               padding="same", use_bias=False)(layer)

layer=Activation('relu')(layer)

layer = Conv2D(filters=256,
               kernel_size=(1, 1),
               strides=(1, 1),
               padding="same", use_bias=False)(layer)

layer=Activation('relu')(layer)

layer = Conv2D(filters=64,
               kernel_size=(1, 1),
               strides=(1, 1),
               padding="same", use_bias=False)(layer)

layer=Activation('relu')(layer)

layer = Flatten()(layer)

layer = Dropout(0.05)(layer)

layer = Dense(units=100)(layer)

layer=Activation('relu')(layer)

layer = Dense(units=10)(layer)

layer=Activation('softmax')(layer)

model = Model(input_layer, layer)

Dengyu-Wu · 2022-09-06T13:04:48Z

Did you normalise your input into [0,1]?

sauravtii · 2022-09-07T16:44:59Z

Hi, I did normalize the input into [0,1] but due to some reasons it didn't get normalized. But now after normalizing it properly I am getting a good accuracy (65.47 %) as compared to before but is not that good. Do you have any suggestions in order to improve my SNN accuracy ?

Dengyu-Wu · 2022-09-07T18:24:09Z

You can refer our paper to include BatchNormalisation Layer (after convolutional and dense layer) and clipping methods to improve conversion efficiency.

sauravtii · 2022-09-07T18:28:12Z

Okay, I will read that. Thank you for the suggestion and your help!

Dengyu-Wu closed this as completed Sep 7, 2022

sauravtii mentioned this issue Sep 30, 2022

Which neuromorphic hardware does SpKeras simulate ? #5

Closed

sauravtii mentioned this issue Nov 8, 2022

List index out of range #7

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

IndexError: list index out of range #4

IndexError: list index out of range #4

sauravtii commented Sep 4, 2022 •

edited

Loading

Dengyu-Wu commented Sep 5, 2022

sauravtii commented Sep 6, 2022 •

edited

Loading

Dengyu-Wu commented Sep 6, 2022

sauravtii commented Sep 6, 2022 •

edited

Loading

Dengyu-Wu commented Sep 6, 2022

sauravtii commented Sep 6, 2022

Dengyu-Wu commented Sep 6, 2022

sauravtii commented Sep 7, 2022

Dengyu-Wu commented Sep 7, 2022 •

edited

Loading

sauravtii commented Sep 7, 2022

IndexError: list index out of range #4

IndexError: list index out of range #4

Comments

sauravtii commented Sep 4, 2022 • edited Loading

Dengyu-Wu commented Sep 5, 2022

sauravtii commented Sep 6, 2022 • edited Loading

Dengyu-Wu commented Sep 6, 2022

sauravtii commented Sep 6, 2022 • edited Loading

Dengyu-Wu commented Sep 6, 2022

sauravtii commented Sep 6, 2022

Dengyu-Wu commented Sep 6, 2022

sauravtii commented Sep 7, 2022

Dengyu-Wu commented Sep 7, 2022 • edited Loading

sauravtii commented Sep 7, 2022

sauravtii commented Sep 4, 2022 •

edited

Loading

sauravtii commented Sep 6, 2022 •

edited

Loading

sauravtii commented Sep 6, 2022 •

edited

Loading

Dengyu-Wu commented Sep 7, 2022 •

edited

Loading