CODEWITHSUNDEEP
pythonpython-3.xtensorflowtensorflow-litetensorflow-datasets
Pruthvi B.
Pruthvi B.

Reputation: 11

'EndVector() takes 1 positional argument but 2 were given' while trying to quantize a tensorflow model

I was trying to quantize a TF model into a TFLite model to deploy it on my ESP32 by calling the dataset through tf.keras.preprocessing.image_dataset_from_directory() and used images_batch and labels_batch to iterate in the representative dataset() function. But, I am getting the error: 'EndVector() takes 1 positional argument but 2 were given' and unable to rectify it. Can someone please help me out?

TFlite convertor code:

converter = tf.lite.TFLiteConverter.from_saved_model(MODEL_DIR)


def representative_dataset():
    for image_batch, labels_batch in train_ds:
        yield [image_batch]

converter.optimizations = [tf.lite.Optimize.DEFAULT]

converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS_INT8]
converter.inference_input_type = tf.int8
converter.inference_output_type = tf.int8
converter.representative_dataset = representative_dataset
model_tflite = converter.convert()

open('modelwithquant.tflite', "wb").write(model_tflite)

Upvotes: 0

Views: 1817

Answers (2)

Ramon Hollands
Ramon Hollands

Reputation: 21

Downgrading flatbuffer seems to help:

pip install --upgrade flatbuffers==1.12

Upvotes: 2

daverim
daverim

Reputation: 111

Assuming train_ds is tensorflow dataset:

Try changing this to:

def representative_dataset():
    for image in train_ds.batch(1).take(100):
        yield [image]

Representative dataset should generally yield a single representative input.

Upvotes: 0

Related Questions