-1

My understanding is that tflite was developed to avoid installing the full package of Tensorflow in embedded system such Raspberry Pi, saving processing and power resources.

Following the directions from https://pypi.org/project/tflite/, the documentation is linked to https://ai.google.dev/edge/litert/models/convert_tf, but the examples shows the usage of Tensorflow full package.

Additionally, Tensorflow API shows tf.lite (https://www.tensorflow.org/api_docs/python/tf/lite), so what would be the point of installing tflite if the whole Tensorflow package needs to be installed?

What am I missing or misunderstanding? To use Tensorflow Lite, the full Tensorflow package needs to be installed anyway? If so, what resources would be host saved?

Improve this question

1 Answer 1

Reset to default
1

the expected flow for your use case is to use TensorFlow to train your model on hardware with sufficient processing power and then, once the model is trained, convert it to TFLite and install an interpreter (tflite-runtime) on your Raspberry Pi

TensorFlow contains the ability to convert .tf files into .tflite by using the tflite package you have linked.

Files which are then understood by tflite-runtime

see:

Google

PyPi

GitHub

Improve this answer
Sign up to request clarification or add additional context in comments.

2 Comments

Thanks! it gives me a better understanding of the whole process. The doubt I have then is from Tensorflow, what's the purpose of tf.lite?
tf.lite would be what you use on your raspberry pi to do stuff with the .tflite file. you cant just put a .tflite file on the Pi & expect it to perform action. you need an interpreter to understand the .tflite file & a language package to do something based on the memory

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.