Conversion of mobilevit to Keras 3 with tf.keras backwards compatibility #18827
Conversation
|
Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA). View this failed invocation of the CLA check for more information. For the most up to date status, view the checks section at the bottom of the pull request. |
|
|
||
| def conv_block(x, filters=16, kernel_size=3, strides=2): | ||
| conv_layer = layers.Conv2D( | ||
| filters, kernel_size, strides=strides, activation=tf.nn.swish, padding="same" |
| m = layers.Conv2D(output_channels, 1, padding="same", use_bias=False)(m) | ||
| m = layers.BatchNormalization()(m) | ||
|
|
||
| if tf.math.equal(x.shape[-1], output_channels) and strides == 1: |
There was a problem hiding this comment.
Use if x.shape[-1] == output_channels
|
|
||
| import tensorflow as tf | ||
|
|
||
| # For versions <TF2.13 change the above import to: |
| models with the same or higher complexity ([MobileNetV3](https://arxiv.org/abs/1905.02244), | ||
| for example), while being efficient on mobile devices. | ||
|
|
||
| Note: This example should be run with Tensorflow 2.13 and higher. |
| inference with TFLite models, check out | ||
| [this official resource](https://www.tensorflow.org/lite/performance/post_training_quantization). | ||
|
|
||
| You can use the trained model hosted on [Hugging Face Hub](https://huggingface.co/keras-io/mobile-vit-xxs) |
There was a problem hiding this comment.
Remove these 2 outdated lines
| tf.lite.OpsSet.SELECT_TF_OPS, # Enable TensorFlow ops. | ||
| ] | ||
| tflite_model = converter.convert() | ||
| open("mobilevit_xxs.tflite", "wb").write(tflite_model) |
There was a problem hiding this comment.
Did you check that the TFLite model works?
There was a problem hiding this comment.
I didn't perform the inference using TFLite model, but I verified that this line is saving TFLite model.
There was a problem hiding this comment.
Can you try using the TFLite interpreter to check that it works as expected? https://www.tensorflow.org/lite/guide/inference#load_and_run_a_model_in_python
|
@fchollet Completed all the changes asked by you. Also, I checked converted TFLite model is working as expected. |
google-ml-butler
bot
added
kokoro:force-run
ready to pull
google-ml-butler
bot
removed
ready to pull
|
I'm not able to render the example as a notebook, the kernel dies at the line that writes the serialized model (even with 60GB of RAM). Are you able to run |
I couldn't find |
|
@fchollet Since I have made changes to work with tensorflow backend and it's merged to be available in |
Conversion of mobilevit.py to Keras 3 with tensorflow backend as per stage 1 of Issue: Keras.io examples conversion gameplan #18468.
Provided solutions of the all problems raised by me in Issue #18613 .