Erni

But the implementation of fastai does not do it that way. Please try to run the following code: ``` from fastai.text.all import * path = untar_data(URLs.IMDB_SAMPLE) imdb = pd.read_csv(path/'texts.csv') imdb_lm...

Well the expected behaviour is that you get the statistics of the activations for each of the layers. So instead of a `None` I would expect to find something of...

Ok Let's try the following. First: I modify the [AWD_LSTM](https://github.com/fastai/fastai/blob/master/fastai/text/models/awdlstm.py) class in the following way: ``` class AWD_LSTM_M(Module): "AWD-LSTM inspired by https://arxiv.org/abs/1708.02182" initrange=0.1 def __init__(self, vocab_sz, emb_sz, n_hid, n_layers, pad_token=1,...

I think basically all the `None` should have some statistics right? I do not know how the people who coded that expected it to work but in the code there...

Hey @sutt ! I saw your pull request. I wanted to ask you what about the Embedding layer? I have spotted the problem: basically [EmbeddingDropout](https://github.com/fastai/fastai/blob/master/fastai/text/models/awdlstm.py#L81) is not captured by `flatten_model`because...

Found a solution! Just wrote a new `EmbeddingDropout` class, and modified a couple of line from the original code of `AWD_LSTM`. Now the hooks work good and besides there is...

Hey @sutt ! encoder_dp is just used if `from_embeds=False`, which it is by default. In case `from_embeds=True` basically we are skipping the `self.encoder` layer from AWD_LSTM because we are using...

Ok, I have created a notebook where I tried to explain as good as possible why I am doing this PR and what are the possible solutions that I see...

I modified the PR to pass the previous test, but now is failing in another test: ``` 1 enc = nn.Embedding(10, 7, padding_idx=1) ----> 2 enc_dp = EmbeddingDropout(enc, 0.5) 3...

Please before accepting the modifications of this PR check the Notebook I created explaining how to make full sense of this PR. Right now everything works, but the code can...