Fastai v2 chat

rmkn85 · October 17, 2019, 12:01am

Why does drop_last equal shuffles?

fastai/fastai_dev/blob/c391844103e3e0bff0cfb417a8340deea74ffecd/dev/local/data/core.py#L144


    def n_subsets(self): return len(self.splits)
    def _new(self, items, **kwargs): return super()._new(items, splits=self.splits, **kwargs)
    def subset(self): raise NotImplemented


    def databunch(self, bs=16, val_bs=None, shuffle_train=True, n=None, dl_type=None, dl_kwargs=None, **kwargs):
        if dl_kwargs is None: dl_kwargs = [{}] * self.n_subsets
        ns = self.n_subsets-1
        bss = [bs] + [2*bs]*ns if val_bs is None else [bs] + [val_bs]*ns
        shuffles = [shuffle_train] + [False]*ns
        if dl_type is None: dl_type = self._dl_type
        dls = [dl_type(self.subset(i), bs=b, shuffle=s, drop_last=s, n=n if i==0 else None, **kwargs, **dk)
               for i,(b,s,dk) in enumerate(zip(bss,shuffles,dl_kwargs))]
        return DataBunch(*dls)


FilteredBase.train,FilteredBase.valid = add_props(lambda i,x: x.subset(i), 2)


#Cell
class TfmdList(FilteredBase, L):
    "A `Pipeline` of `tfms` applied to a collection of `items`"
    _default='tfms'
    def __init__(self, items, tfms, use_list=None, do_setup=True, as_item=True, split_idx=None, train_setup=True, splits=None):

jeremy · October 17, 2019, 1:06am

That’s right. I just added this thingie the other day

jeremy · October 17, 2019, 1:07am

Generally you just want your training set to have drop_last.

rmkn85 · October 17, 2019, 6:06am

So it’s implicit there that training set has shuffle=True and test set has shuffle=False?

MadeUpMasters · October 17, 2019, 2:39pm

Are there any end-to-end training examples where we can see models (especially vision) training using an example of the high-level API? I checked around the nbs but, besides a brief example in 21_tutorial_imagenette, didn’t see any.

I just got one working for audio. 97% accuracy on speaker recognition using basic fit and xresnet18, even though the rest of our audio stuff is still pretty buggy.

muellerzr · October 17, 2019, 3:38pm

IIRC Jeremy said the high level isn’t quite done yet. Just medium and low level

muellerzr · October 19, 2019, 8:23pm

Do we have an example of inference on a test set? (Or how to add that to the databunch?)

sgugger · October 19, 2019, 9:36pm

At the end of the imagenette tutorial, there is an example of adding a test set (there is a test_dl function for that). Note that this part is still under development and isn’t thoroughly tested yet.

muellerzr · October 21, 2019, 12:44pm

Do we know the equivalent for a pip install for mkl? I cannot import anything from fastai2.basics due to an import error “No module named ‘mkl’”. Which seems to be only installable via conda (but I cannot do that as I’m using Google Colab). Thoughts?

Edit: for those who do not know (like me just now!), pip is for python packages, mkl is not python

Edit x2: things I have tried:

I tried following this to install conda, and when I double checked and did a !conda install mkl-service (and did regular mkl too) it says it’s already installed.

fanyi · October 21, 2019, 1:54pm

same thing happen to me. I can only turn around by remove the “MKL” in local/core/imports.py for the time being. Hope anyone can help to solve the problem.

jeremy · October 21, 2019, 2:25pm

I’ll remove that import soon. In the meantime just remove it as @fanyi suggested.

muellerzr · October 21, 2019, 2:29pm

Will do, thanks Jeremy!

To the high-level API question earlier, is your finalized version of what you want that to look like is a PipeLine + .databunch()? Or is there any intentions of doing it in one step (I don’t see how to do that exactly so just pondering your thoughts/intentions, like how we could ImageDataBunch in one go)

jeremy · October 21, 2019, 2:32pm

Sorry @muellerzr I don’t understand your question. Could you provide more context and detail please? Perhaps also some examples?

muellerzr · October 21, 2019, 2:39pm

Realized I answered my own question there, my apologies!

Although a separate question on test sets, Sylvain mentioned test sets were at the end of the Imagenette turtorial (nb 22) but I did not see test set in there.

Do you know which notebook I should look in to see how to do this?

fanyi · October 21, 2019, 2:48pm

in 38_tutorial_ulmfit.ipynb, when I run:
learn = language_model_learner(dbunch, AWD_LSTM, vocab, metrics=[accuracy, Perplexity()], path=path, opt_func = partial(Adam, wd=0.1)).to_fp16()

I got an error “not a gzip file”, any suggest from your experts.

sgugger · October 21, 2019, 3:15pm

Ah yes, the example has been removed. Though the function is documented with an example.

muellerzr · October 21, 2019, 3:47pm

How do we use the new test_dl? Do we replace our Learner's data? Or create a new one with the three? And how do we get predictions off of it? (A bit lost on that end of it)

sgugger · October 21, 2019, 4:21pm

Learner.get_preds takes a dataloader if you pass it with dl=...

muellerzr · October 21, 2019, 4:35pm

Awesome! Thanks

MadeUpMasters · October 22, 2019, 4:13pm

Any interest in a test_warns function for 00_test? Python warnings library has a nice context manager for testing that we could wrap. I’d be happy to do it, or let me know if there’s a better way.

#export
def test_warns(f, args, kwargs, show=False):
    with warnings.catch_warnings(record=True) as w:
        test_eq(len(w), 0)
        f(*args, **kwargs)
        test_ne(len(w), 0)
        if show: 
            for e in w: print(f"{e.category}: {e.message}")

It’d also be pretty easy to add a regex or msg to make sure a certain warning is given similar to how test_stdout works, not sure if it’s worth the trouble though.