support keras?

[idata]

since now movidius support TF, does mvNCCompile also support Keras with tensorflow backend?

[idata]

I would also like to know this answer. It must be possible, but is there coding or configuration required?

[idata]

I am still working on it, but the input and output I think should be implemented by raw tensorflow, and you could stack your layer using keras.

[idata]

as I said above, I use raw tensorflow to implement Input and Output, between them it is built by Keras. Then I trained my model in the laptop. Everything goes well, and the model was saved as .meta file, but when I use mvNCCompile to compile my model, error occurs:

InvalidArgumentError (see above for traceback): You must feed a value for placeholder tensor 'y_true' with dtype float and shape [?,2]

[[Node: y_true = Placeholder[dtype=DT_FLOAT, shape=[?,2], _device="/job:localhost/replica:0/task:0/device:CPU:0"]()]]img_input = tf.placeholder(tf.float32, shape=(None, 224, 224, 3), name="x") y_true = tf.placeholder(tf.float32, shape=(None, 2), name="y_true") # Block 1 x = Conv2D(64, (3, 3), activation='relu', padding='same', name='block1_conv1')(img_input) x = Conv2D(64, (3, 3), activation='relu', padding='same', name='block1_conv2')(x) x = MaxPooling2D((2, 2), strides=(2, 2), name='block1_pool')(x) # Block 2 x = Conv2D(128, (3, 3), activation='relu', padding='same', name='block2_conv1')(x) x = Conv2D(128, (3, 3), activation='relu', padding='same', name='block2_conv2')(x) x = MaxPooling2D((2, 2), strides=(2, 2), name='block2_pool')(x) # Block 3 x = Conv2D(256, (3, 3), activation='relu', padding='same', name='block3_conv1')(x) x = Conv2D(256, (3, 3), activation='relu', padding='same', name='block3_conv2')(x) x = Conv2D(256, (3, 3), activation='relu', padding='same', name='block3_conv3')(x) x = MaxPooling2D((2, 2), strides=(2, 2), name='block3_pool')(x) # Block 4 x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block4_conv1')(x) x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block4_conv2')(x) x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block4_conv3')(x) x = MaxPooling2D((2, 2), strides=(2, 2), name='block4_pool')(x) # Block 5 x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block5_conv1')(x) x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block5_conv2')(x) x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block5_conv3')(x) x = MaxPooling2D((2, 2), strides=(2, 2), name='block5_pool')(x) #dense layers are removed predictions = Dense(2, activation="softmax")(x) loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y_true, logits=predictions)) train_step = tf.train.AdamOptimizer(learning_rate=1e-3).minimize(loss) correct_pred = tf.equal(tf.argmax(predictions, 1), tf.argmax(y_true, 1)) acc_value = tf.reduce_mean(tf.cast(correct_pred, tf.float32))

And I think I do feed the value to the y_true

for j in range(n_batch): start = j * BATCH_SIZE batch_train = train[start:start+BATCH_SIZE] X_batch_train = np.array([i[0] for i in batch_train]) y_batch_train = np.array([i[1] for i in batch_train]) feed_dict_train = {img_input: X_batch_train, y_true: y_batch_train} _, acc = sess.run([train_step, acc_value], feed_dict=feed_dict_train) if j % 10 == 0: print("training accuracy:{}".format(acc))

Is it the compile command problem or my code itself?