When you have your TensorFlow Model and want to use it on a mobile device, you'll need to convert it to the TFLite format.
This process can be done in two ways:
- Using the Python API
- Using a command line tool
Let's look into some more details….
1/6🧵
This process can be done in two ways:
- Using the Python API
- Using a command line tool
Let's look into some more details….
1/6🧵
Why do we need to convert?
The TFLite is an optimized format (Flatbuffer) for faster loading
To keep the framework lite and fast, all the Operations are optimized for mobile execution but not all TF operations are available
The available ops: tensorflow.org/lite/guide/ops…
2/6🧵
The TFLite is an optimized format (Flatbuffer) for faster loading
To keep the framework lite and fast, all the Operations are optimized for mobile execution but not all TF operations are available
The available ops: tensorflow.org/lite/guide/ops…
2/6🧵
How to convert the model?
The Python API to convert a model is straightforward and it's the recommended method of conversion
You can, during the conversion, apply some optimizations, like post training quantization to reduce your model size and latency.
3/6🧵
The Python API to convert a model is straightforward and it's the recommended method of conversion
You can, during the conversion, apply some optimizations, like post training quantization to reduce your model size and latency.
3/6🧵
The command line version is similar but you'll need to install TF 2 from pip in your machine.
The idea is similar to the Python API seen before.
4/6🧵
The idea is similar to the Python API seen before.
4/6🧵
If the available operations by default on TFLite are not enough, you can enable selected ops to be added to your .tflite converted model.
Drawback:
- You'll need additional dependencies on your mobile app, making it bigger
- Ops may not be optimized
5/6🧵
Drawback:
- You'll need additional dependencies on your mobile app, making it bigger
- Ops may not be optimized
5/6🧵
This is the basics of converting a model to TFLite.
Several times you don't really need to do any of this. The model publisher might have done that for you already.
On #TFHub (tfhub.dev) you already have multiple converted models for you to try.
6/6🧵
Several times you don't really need to do any of this. The model publisher might have done that for you already.
On #TFHub (tfhub.dev) you already have multiple converted models for you to try.
6/6🧵
• • •
Missing some Tweet in this thread? You can try to
force a refresh
