Change EfficientNetB0 to EfficientNetB1 in the final line #259
Conversation
|
@fchollet just did. |
| @@ -1,733 +1,766 @@ | |||
| { | |||
There was a problem hiding this comment.
I'm surprising to see a complete rewrite of the notebook file -- it should just be a 2 line change?
There was a problem hiding this comment.
Same here :o
Could it be because I edited it in Colab and committed directly from there?
There was a problem hiding this comment.
Most likely. You can either apply the edit to the ipynb by hand or regenerate the ipynb with the tutobook.py function py_to_nb (that way you don't have to execute the notebook, which takes time)
There was a problem hiding this comment.
Okay. Then I will work on the JSON of the ipynb directly so that the state does not get changed.
There was a problem hiding this comment.
@fchollet fixed it up.
Operating directly on the JSON of the notebook seems like an easier choice for small changes like this.
…#259) * Bump B0 to B1 in the final line * Update image_classification_efficientnet_fine_tuning.md * Bumped to B1 * Updated the lines directly on ipynb
* Show the encoded text in the preprocessing guide (#261) * Show the encoded text in the preprocessing guide * python autogen.py generate_tf_guides preprocessing_layers.py * python autogen.py add_guide preprocessing_layers.py * Change EfficientNetB0 to EfficientNetB1 in the final line (#259) * Bump B0 to B1 in the final line * Update image_classification_efficientnet_fine_tuning.md * Bumped to B1 * Updated the lines directly on ipynb * Add SaveOptions to tutorial. * add release date * Add save_traces description to SavedModel guide Co-authored-by: Mark Daoust <markdaoust@google.com> Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>
With reference to the EfficientNet tutorial the last line tries to load B1 weights into B0 which results in incompatibility problems. Additionally, I think it would be useful to just provide a link to the
efficientnet_weight_update_util.pysince it does not come with the standard installation.This PR address the above pointers.