Skip to content

GitLab

Commit bef8f110 authored by Liangzhe Yuan's avatar Liangzhe Yuan Committed by A. Unique TensorFlower
Browse files

Internal change 

PiperOrigin-RevId: 416144744
parent 2114ee0b
Show whitespace changes
Inline Side-by-side
Showing with 6 additions and 0 deletions
official/vision/beta/projects/movinet/README.md
View file @ bef8f110
...@@ -300,6 +300,12 @@ non_streaming_prediction = tf.argmax(non_streaming_output, -1) ...@@ -300,6 +300,12 @@ non_streaming_prediction = tf.argmax(non_streaming_output, -1)
This section outlines an example on how to export a model to run on mobile This section outlines an example on how to export a model to run on mobile
devices with [TF Lite](https://www.tensorflow.org/lite). devices with [TF Lite](https://www.tensorflow.org/lite).
[Optional] For streaming models, they are typically trained with
`conv_type = 3d_2plus1d` for better training throughpouts. In order to achieve
better inference performance on CPU, we need to convert the `3d_2plus1d`
checkpoint to make it compatible with the `2plus1d` graph.
You could achieve this by running `tools/convert_3d_2plus1d.py`.
First, convert to [TF SavedModel](https://www.tensorflow.org/guide/saved_model) First, convert to [TF SavedModel](https://www.tensorflow.org/guide/saved_model)
by running `export_saved_model.py`. For example, for `MoViNet-A0-Stream`, run: by running `export_saved_model.py`. For example, for `MoViNet-A0-Stream`, run:
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment