These are RNN, LSTM and GRU PyTorch implementations created by Ignacio Oguiza - oguiza@gmail.com based on:

class RNN[source]

RNN(c_in, c_out, hidden_size=100, n_layers=1, bias=True, rnn_dropout=0, bidirectional=False, fc_dropout=0.0) :: _RNN_Base

Same as nn.Module, but no need for subclasses to call super().__init__

class LSTM[source]

LSTM(c_in, c_out, hidden_size=100, n_layers=1, bias=True, rnn_dropout=0, bidirectional=False, fc_dropout=0.0) :: _RNN_Base

Same as nn.Module, but no need for subclasses to call super().__init__

class GRU[source]

GRU(c_in, c_out, hidden_size=100, n_layers=1, bias=True, rnn_dropout=0, bidirectional=False, fc_dropout=0.0) :: _RNN_Base

Same as nn.Module, but no need for subclasses to call super().__init__

bs = 16
c_in = 3
seq_len = 12
c_out = 2
xb = torch.rand(bs, c_in, seq_len)
test_eq(RNN(c_in, c_out, hidden_size=100, n_layers=2, bias=True, rnn_dropout=0.2, bidirectional=True, fc_dropout=0.5)(xb).shape, [bs, c_out])
test_eq(RNN(c_in, c_out)(xb).shape, [bs, c_out])
test_eq(RNN(c_in, c_out, hidden_size=100, n_layers=2, bias=True, rnn_dropout=0.2, bidirectional=True, fc_dropout=0.5)(xb).shape, [bs, c_out])
test_eq(LSTM(c_in, c_out)(xb).shape, [bs, c_out])
test_eq(GRU(c_in, c_out)(xb).shape, [bs, c_out])

from tsai.data.all import *
dsid = 'NATOPS' 
bs = 16
X, y, splits = get_UCR_data(dsid, return_split=False)
tfms  = [None, [Categorize()]]
dsets = TSDatasets(X, y, tfms=tfms, splits=splits, inplace=True)
dls   = TSDataLoaders.from_dsets(dsets.train, dsets.valid, bs=bs, num_workers=0, shuffle=False)
model = LSTM(dls.vars, dls.c)
learn = Learner(dls, model,  metrics=accuracy)
learn.fit_one_cycle(1, 3e-3)
epoch train_loss valid_loss accuracy time
0 1.795245 1.774184 0.238889 00:02 
m = RNN(c_in, c_out, hidden_size=100,n_layers=2,bidirectional=True,rnn_dropout=.5,fc_dropout=.5)
print(m)
print(total_params(m))
m(xb).shape

RNN(
  (rnn): RNN(3, 100, num_layers=2, batch_first=True, dropout=0.5, bidirectional=True)
  (dropout): Dropout(p=0.5, inplace=False)
  (fc): Linear(in_features=200, out_features=2, bias=True)
)
(81802, True)

torch.Size([16, 2])
m = LSTM(c_in, c_out, hidden_size=100,n_layers=2,bidirectional=True,rnn_dropout=.5,fc_dropout=.5)
print(m)
print(total_params(m))
m(xb).shape

LSTM(
  (rnn): LSTM(3, 100, num_layers=2, batch_first=True, dropout=0.5, bidirectional=True)
  (dropout): Dropout(p=0.5, inplace=False)
  (fc): Linear(in_features=200, out_features=2, bias=True)
)
(326002, True)

torch.Size([16, 2])
m = GRU(c_in, c_out, hidden_size=100,n_layers=2,bidirectional=True,rnn_dropout=.5,fc_dropout=.5)
print(m)
print(total_params(m))
m(xb).shape

GRU(
  (rnn): GRU(3, 100, num_layers=2, batch_first=True, dropout=0.5, bidirectional=True)
  (dropout): Dropout(p=0.5, inplace=False)
  (fc): Linear(in_features=200, out_features=2, bias=True)
)
(244602, True)

torch.Size([16, 2])