Wrappers for freeze() and unfreeze()

[Cdk29]

Wrappers for the freeze() and unfreeze() methods from fastai.

Creating function allow to pipe the model into them. Also, I took the freedom to add a message, to make the functions more ergonomics and prevent doubts from the user on the success of the operation.

The function has been tested here : https://github.com/Cdk29/Cassava/blob/main/dev_freeze_unfreeze_fastai.md

[turgut090]

@Cdk29 Btw, you could add here a ggplot method for visualization instead of matplotlob. (additional argument to function, e.g. method="ggplot2") Because learn %>% lr_find() returns a dataframe.

https://github.com/henry090/fastai/blob/0cffb51bef19723a9e3533e40a5a770f3b5f2d9a/R/test.R#L281-L297

[Cdk29]

@henry090 yes, and this would allow to pipe directly like this :

learn %>% lr_find() %>% plot_lr_find()

I will do this in an other PR ^^

[turgut090]

I see there could be some parameters in learn$freeze(...). Could you add 3 dots after the object argument? Pass these dots to args, then, please. The same for freeze/unfreeze.
@Cdk29 could you make the necessary changes, so I merge this.

[Cdk29]

@henry090 I can yes. But the original fastai code cannot have additional argument so I am not sure it is a good idea : https://github.com/fastai/fastai/blob/master/fastai/learner.py#L521

[turgut090]

@Cdk29 It will not throw an error. It just allows passing something in case of future modifications. So, this is safe.

[Cdk29]

@henry090 done !

[turgut090]

@Cdk29 you forgot to add @param to the documentation part. And please change @return Nothing to None.

[Cdk29]

@henry090 Indeed, sorry for that. I have done the requested changes.

[turgut090]

Thanks very much!