AWD-LSTM

    AWD LSTM from

    On top of the pytorch or the fastai layers, the language models use some custom layers specific to NLP.

    dropout_mask[source]

    Return a dropout mask of the same type as x, size sz, with probability p to cancel an element.

    1. t = dropout_mask(torch.randn(3,4), [4,3], 0.25)
    2. test_eq(t.shape, [4,3])
    3. assert ((t == 4/3) + (t==0)).all()

    RNNDropout(p=0.5) ::

    Dropout with probability p that is consistent on the seq_len dimension.

    It also supports doing dropout over a sequence of images where time dimesion is the 1st axis, 10 images of 3 channels and 32 by 32.

    1. _ = dp(torch.rand(4,10,3,32,32))

    A module that wraps another layer in which some weights will be replaced by 0 during training.

    Apply dropout with probability embed_p to an embedding layer emb.

    1. enc = nn.Embedding(10, 7, padding_idx=1)
    2. enc_dp = EmbeddingDropout(enc, 0.5)
    3. tst_inp = torch.randint(0,10,(8,))
    4. tst_out = enc_dp(tst_inp)
    5. for i in range(8):

    AWD_LSTM(vocab_sz, emb_sz, n_hid, n_layers, pad_token=1, hidden_p=0.2, input_p=0.6, embed_p=0.1, weight_p=0.5, bidir=False) :: Module

    AWD-LSTM inspired by

    This is the core of an AWD-LSTM model, with embeddings from vocab_sz and emb_sz, n_layers LSTMs potentially bidir stacked, the first one going from emb_sz to n_hid, the last one from n_hid to emb_sz and all the inner ones from n_hid to . pad_token is passed to the PyTorch embedding layer. The dropouts are applied as such:

    • the embeddings are wrapped in EmbeddingDropout of probability embed_p;
    • the result of this embedding layer goes through an of probability input_p;
    • each LSTM has WeightDropout applied with probability weight_p;
    • between two of the inner LSTM, an is applied with probability hidden_p.

    THe module returns two lists: the raw outputs (without being applied the dropout of hidden_p) of each inner LSTM and the list of outputs with dropout. Since there is no dropout applied on the last output, those two lists have the same last element, which is the output that should be fed to a decoder (in the case of a language model).

    awd_lstm_lm_split

    Split a RNN model in groups for differential learning rates.

    awd_lstm_clas_split

    Split a RNN model in groups for differential learning rates.

    QRNN

    AWD_QRNN(vocab_sz, emb_sz, n_hid, n_layers, pad_token=1, hidden_p=0.2, input_p=0.6, embed_p=0.1, weight_p=0.5, bidir=False) ::

    Same as an AWD-LSTM, but using QRNNs instead of LSTMs

    1. # cpp
    2. model = AWD_QRNN(vocab_sz=10, emb_sz=20, n_hid=16, n_layers=2, bidir=False)
    3. x = torch.randint(0, 10, (7,5))
    5. test_eq(y.shape, (7, 5, 20))

    ©2021 fast.ai. All rights reserved.
    Site last generated: Mar 31, 2021