Hi, there. I’m just using my own AMB82-mini Board. To apply my image classification model, I just made my own .h5 file and attempted to convert it to .nb file via online conversion tool.(https://www.amebaiot.com/en/amebapro2-ai-convert-model/) I tried to convert it, but nothing happened. Please help me… Even though I checked my email spam folder, there was nothing sent from the online conversion tool.
Im trying to do the same thing here and no .nb models being sent my way. Earlier about 3-4 hours ago, it was working just fine too. I think their servers might just be having some issue and we will just have to wait it out. Judging from past threads they have had such issues before and soooo the only solution is to just wait it out.
Would you mind to convert your keras model to onnx first before starting the conversion process? The conversion tool might not support sequential operation, thus import failure. Please refer to the script below for keras → onnx conversion, kindly use pip3 install tf2onnx==1.10.1
import tensorflow as tf
import tf2onnx
# load model
model = tf.keras.models.load_model('keras_model.h5')
# check model structure
model.summary()
# convert to ONNX
input_shape = model.input_shape
if isinstance(input_shape, list):
input_shape = input_shape[0]
spec = (tf.TensorSpec(input_shape, tf.float32, name="input"),)
model_proto, _ = tf2onnx.convert.from_keras(
model,
input_signature=spec,
output_path="model.onnx",
opset=11
)
print("converted model: model.onnx")