This message was deleted.
# ask-for-helps
Slackbot
07/07/2023, 10:19 PMThis message was deleted.
c
Chaoyu
07/10/2023, 11:22 PMHi @Paul Roopson Pradeep, could you also share your Keras model definition code, especially the
predictChaoyu
07/10/2023, 11:23 PMOr sample code of how do you use the predict method without the runner?
p
Paul Roopson Pradeep
07/13/2023, 2:35 AMsorry for the late response:
Here is the model definition:
Copy code
InputA = Input(shape=(108,))
InputB = Input(shape=(6,))
x = Dense(16,activation="relu")(InputA)
x_model = Model(inputs=InputA, outputs=x)
y = Embedding(input_dim=len(vocab), output_dim=16, input_length=6)(InputB)
y = Flatten()(y)
y_model = Model(inputs=InputB, outputs=y)
combined = concatenate([x_model.output, y_model.output])
z = Dense(16, activation='relu')(combined)
z = Dense(1, activation='linear')(z)
model = Model(inputs=[x_model.input, y_model.input], outputs=z)
model.compile(loss='mse', optimizer='adam', metrics=['mae'])Paul Roopson Pradeep
07/13/2023, 2:37 AMand here is the predict call:
pred = model.predict([x_test,a_x_test])x_testa_x_testx_testa_x_testPaul Roopson Pradeep
07/13/2023, 2:39 AMThis works fine. when i tried to package is as a bentoml model and run the prediction with the same test inputs, it doesnt work
c
Chaoyu
07/13/2023, 2:49 AMdid you enable batching or specify the batch dim when saving the model with BentoML?
p
Paul Roopson Pradeep
07/13/2023, 3:06 AMHere is how i am saving the model:
bentoml.keras.save_model("embedding_price_model", model)Paul Roopson Pradeep
07/13/2023, 3:07 AMis there anything else i should be passing in?
c
Chaoyu
07/13/2023, 4:12 AMAre you using Tensorflow 1.x or 2.0+?
Chaoyu
07/13/2023, 4:12 AMcc @Jiang @larme (shenyang)
p
Paul Roopson Pradeep
07/13/2023, 6:27 AMtensorflow 2
Paul Roopson Pradeep
07/13/2023, 6:27 AMspecifically 2.13.0-rc1
j
Jiang
07/13/2023, 7:54 AM@Paul Roopson Pradeep Hi. Just try
price_predictions = runner.predict.run(input_a, input_b)p
Paul Roopson Pradeep
07/13/2023, 6:40 PMah cool let me try this