Not found: No algorithm worked!
Mon, Oct 12, 2020
This error occurs when the Conv2D layer input shape doesn’t match what was specified in the actual data. Error text:
I noticed the “Not found: No algorithm worked!” error after changing the order of the data I was passing into the model. From
shape=(depth, rows, columns) to shape=(rows,columns,depth).
Unfortunately I missed the change in my parse function, so I was passing in the wrong shape to model.fit().
def parse_element_function:
example = tf.io.parse_single_example(example_proto, feature_description)
raw_image = example['raw_image']
# This line put the indexes in the wrong order!
image = tf.reshape(raw_image,tf.stack([input_depth,input_rows,input_columns]))
return (image,example['label'],example['filename'])
preprocessed_dataset = dataset.map(add_nose)
The interesting thing is that the input layer is expecting the correct shape, but didn’t give an error when the shapes didn’t match.
inputs = keras.layers.Input(shape=(input_rows,input_columns,input_depth))
x = inputs
x = keras.layers.Conv2D(8,(3,3),padding='same',activation='relu')(x)
...
This is the error text I was getting from Tensorflow:
Traceback (most recent call last):
...
tensorflow.python.framework.errors_impl.NotFoundError: 2 root error(s) found.
(0) Not found: No algorithm worked!
[[{{node conv2d_8/Conv2D}}]]
[[metrics_2/acc/Identity_2/_789]]
(1) Not found: No algorithm worked!
[[{{node conv2d_8/Conv2D}}]]
0 successful operations.
0 derived errors ignored.
Tried running in CPU only mode by disabling GPU’s:
(venv) host$ env CUDA_VISIBLE_DEVICES="" python train_model.py
And got a different error.
Traceback (most recent call last):
...
tensorflow.python.framework.errors_impl.UnimplementedError: Fused conv
implementation does not support grouped convolutions for now.
[[{{node conv2d_8/Relu}}]]