François Chollet comments

Results 342 comments of


                                            François Chollet

[Contributors Wanted] A real-world model benchmark

@[fsx950223](https://github.com/fsx950223) generally speaking we should implement models from scratch rather than using pretrained models and fine-tuning them (e.g. no TF-Hub, just Keras layers). Also it's not very important to use...

[Contributors Wanted] A real-world model benchmark

> Could I use tf model official apis? For Bert specifically this seems ok. You can generally used any reusable building blocks from an official Keras/TF lib (like KerasCV, KerasNLP,...

More Tests for customizable reduction strategy in model

Thank you for the PR. @rchao, can you please take a look?

Add textual inversion guide for StableDiffusion()

> load the VAEEncoder with weights ( could use some help porting the weights if @fchollet or @divamgupta would be willing to share their porting process) The way I do...

Fix masks in English-to-Spanish translation example

Thank you for the PR! As of TF 2.10, the MHA layer is able to leverage the `_keras_mask` attributes of input tensors (usually set by the Embedding layer) and to...

Fix masks in English-to-Spanish translation example

> It seems because + is not preserving the mask. I replaced it now with an Add layer. Is this the best way to handle this? Correct. We need a...

Extract Nested Layers from Encapsulated Model

I think the error message could be better, but the behavior here is logical. When you create the `EfficientNet` model, you create a first graph of layers linking `pretrained_model.inputs` to...

Instantiating a `tf.keras.applications` Model Nullifies Augmentation Layers

The expected behavior is that the default `training` value received by the layers should be `True` since it defaults to `True` in the layer signatures and isn't set anywhere else....

Instantiating a `tf.keras.applications` Model Nullifies Augmentation Layers

Yes, let's keep it open until resolved.

Add layer trainable information in model summary and plot model

Thanks for the suggestion. We think this could be a useful feature, so please feel free to open a PR to add an optional `trainable` column to the summary (let's...