lstms.pth

lstms.pth

Implementation of LSTM variants, in PyTorch.

For now, they only support a sequence size of 1, and meant for RL use-cases. Besides that, they are a stripped-down version of PyTorch’s RNN layers. (no bidirectional, no num_layers, no batch_first)

Base Modules:

• SlowLSTM: a (mostly useless) pedagogic example.
• LayerNorm: Layer Normalization as in Ba & al.: Layer Normalization.

Dropout Modules:

• LSTM: the original.
• GalLSTM: using dropout as in Gal & Ghahramami: A Theoretically Grounded Application of Dropout in RNNs.
• MoonLSTM: using dropout as in Moon & al: RNNDrop: A Novel Dropout for RNNs in ASR.
• SemeniutaLSTM: using dropout as in Semeniuta & al: Recurrent Dropout without Memory Loss.

Normalization + Dropout Modules:

• LayerNormLSTM: Dropout + Layer Normalization.
• LayerNormGalLSTM: Gal Dropout + Layer Normalization.
• LayerNormMoonLSTM: Moon Dropout + Layer Normalization.
• LayerNormSemeniutaLSTM: Semeniuta Dropout + Layer Normalization.

Container Modules:

• MultiLayerLSTM: helper class to build multiple layers LSTMs.

Convention: If applicable, the activations are computed first, and then the nodes are droped. (dropout on the output, not the input, just like PyTorch)

Install

pip install -e .

Usage

You can find a good example of how to use the layers in test/test_speed.py.

All Dropout models share the same signature:

LSTM(self, input_size, hidden_size, bias=True, dropout=0.0, dropout_method='pytorch')

All Normalization + Dropout models share the same signature:

LayerNormLSTM(self, input_size, hidden_size, bias=True, dropout=0.0, 
             dropout_method='pytorch', ln_preact=True, learnable=True):

And all models use the same out, hidden = model.forward(x, hidden)signature as the official PyTorch LSTM layers. They also all provide a model.sample_mask() method, which needs to be called in order to sample a new Dropout mask. (e.g, when processing a new sequence)

Note: LayerNorm is not an LSTM layer, and thus uses out = model.forward(x).

Containers

This package provides a helper class, MultiLayerLSTM, which can be use to stack multiple LSTMs together.

lstm = MultiLayerLSTM(input_size=256, layer_type=LayerNormSemeniutaLSTM,
                      layer_sizes=(64, 64, 16), dropout=0.7, ln_preact=False)
hiddens = lstm.create_hiddens(bsz=batch_size)
x = Variable(th.rand(1, 1, 256))
for _ in range(10):
    out, hiddens = lstm(x, hiddens)

Note that hiddens doesn’t match the PyTorch specification. It is the list of (h_i, c_i) for each LSTM layer. Instead, the LSTM layers in PyTorch return a single tuple of (h_n, c_n), where h_n and c_n have sizes (num_layers * num_directions, batch, hidden_size).

Capacity Benchmarks

Warning: This is an artificial memory benchmark, not necessarily representative of each method’s capacity.

Note: nn.LSTM and SlowLSTM do not have dropout in these experiments.

Info: dropout = 0.9 , SEQ_LEN = 10 , dataset size = 100 layer size = 256

model	error
nn.LSTM	3.515
SlowLSTM	4.171
LSTM	4.160
GalLSTM	4.456
MoonLSTM	4.442
SemeniutaLSTM	3.762
GalLSTM	4.456
MoonLSTM	4.442
SemeniutaLSTM	3.762

Speed Benchmarks

Available by running make speed.

Warning: Inference timings only, and on a single sequence of length 1000 with dropout = 0.5 .

SlowLSTM Benchmark

size	nn.LSTM	SlowLSTM	Speedup
128	0.628	0.666	0.943
256	0.676	0.759	0.890
512	0.709	1.026	0.690
1024	2.364	2.867	0.824
2048	6.161	8.261	0.746

LSTM Benchmark

size	nn.LSTM	LSTM	Speedup
128	0.568	0.387	1.466
256	0.668	0.419	1.594
512	0.803	0.769	1.045
1024	2.966	2.002	1.482
2048	6.291	6.393	0.984

GalLSTM Benchmark

size	nn.LSTM	GalLSTM	Speedup
128	0.557	0.488	1.142
256	0.683	0.446	1.530
512	0.966	0.548	1.763
1024	2.524	2.587	0.975
2048	6.618	6.099	1.085

MoonLSTM Benchmark

size	nn.LSTM	MoonLSTM	Speedup
128	0.667	0.445	1.499
256	0.818	0.535	1.530
512	0.908	0.695	1.306
1024	2.517	2.553	0.986
2048	6.475	6.779	0.955

SemeniutaLSTM Benchmark

size	nn.LSTM	SemeniutaLSTM	Speedup
128	0.692	0.513	1.348
256	0.685	0.697	0.983
512	0.717	0.701	1.022
1024	2.639	2.751	0.959
2048	7.294	6.122	1.191

LayerNormLSTM Benchmark

size	nn.LSTM	LayerNormLSTM	Speedup
128	0.646	1.656	0.390
256	0.583	1.800	0.324
512	0.770	1.989	0.387
1024	2.623	3.844	0.682
2048	6.573	9.592	0.685

LayerNormGalLSTM Benchmark

size	nn.LSTM	LayerNormGalLSTM	Speedup
128	0.566	0.486	1.163
256	0.592	0.350	1.693
512	0.920	0.606	1.517
1024	2.508	2.427	1.034
2048	7.356	10.268	0.716

LayerNormMoonLSTM Benchmark

size	nn.LSTM	LayerNormMoonLSTM	Speedup
128	0.507	0.389	1.305
256	0.685	0.511	1.342
512	0.762	0.685	1.111
1024	2.661	2.261	1.177
2048	8.904	9.710	0.917

LayerNormSemeniutaLSTM Benchmark

size	nn.LSTM	LayerNormSemeniutaLSTM	Speedup
128	0.492	0.388	1.267
256	0.583	0.360	1.616
512	0.760	0.578	1.316
1024	2.586	2.328	1.111
2048	6.970	10.725	0.650