def _lstm_test(self, layers, bidirectional, initial_state, packed_sequence,
dropout):
model = LstmFlatteningResult(RNN_INPUT_SIZE,
RNN_HIDDEN_SIZE,
layers,
bidirectional=bidirectional,
dropout=dropout)
if packed_sequence == 1:
model = RnnModelWithPackedSequence(model, False)
if packed_sequence == 2:
model = RnnModelWithPackedSequence(model, True)
def make_input(batch_size):
seq_lengths = np.random.randint(1,
RNN_SEQUENCE_LENGTH + 1,
size=batch_size)
seq_lengths = list(reversed(sorted(map(int, seq_lengths))))
inputs = [
Variable(torch.randn(l, RNN_INPUT_SIZE)) for l in seq_lengths
]
inputs = rnn_utils.pad_sequence(inputs)
if packed_sequence == 2:
inputs = inputs.transpose(0, 1)
inputs = [inputs]
directions = 2 if bidirectional else 1
if initial_state:
h0 = Variable(
torch.randn(directions * layers, batch_size,
RNN_HIDDEN_SIZE))
c0 = Variable(
torch.randn(directions * layers, batch_size,
RNN_HIDDEN_SIZE))
inputs.append((h0, c0))
if packed_sequence != 0:
inputs.append(Variable(torch.IntTensor(seq_lengths)))
if len(inputs) == 1:
input = inputs[0]
else:
input = tuple(inputs)
return input
input = make_input(RNN_BATCH_SIZE)
self.run_model_test(model,
train=False,
batch_size=RNN_BATCH_SIZE,
input=input,
use_gpu=False)
# test that the model still runs with a different batch size
onnxir, _ = do_export(model, input)
other_input = make_input(RNN_BATCH_SIZE + 1)
_ = run_embed_params(onnxir, model, other_input, use_gpu=False)
def _lstm_test(self, layers, bidirectional, initial_state,
packed_sequence, dropout):
model = LstmFlatteningResult(
RNN_INPUT_SIZE, RNN_HIDDEN_SIZE, layers,
bidirectional=bidirectional, dropout=dropout)
if packed_sequence:
model = RnnModelWithPackedSequence(model)
seq_lengths = np.random.randint(1, RNN_SEQUENCE_LENGTH + 1, size=RNN_BATCH_SIZE)
seq_lengths = list(reversed(sorted(map(int, seq_lengths))))
inputs = [ Variable(torch.randn(l, RNN_INPUT_SIZE)) for l in seq_lengths ]
inputs = [rnn_utils.pad_sequence(inputs)]
directions = 2 if bidirectional else 1
if initial_state:
h0 = Variable(torch.randn(directions * layers, RNN_BATCH_SIZE, RNN_HIDDEN_SIZE))
c0 = Variable(torch.randn(directions * layers, RNN_BATCH_SIZE, RNN_HIDDEN_SIZE))
inputs.append((h0, c0))
if packed_sequence:
inputs.append(Variable(torch.IntTensor(seq_lengths)))
if len(inputs) == 1:
input = inputs[0]
else:
input = tuple(inputs)
self.run_model_test(model, train=False, batch_size=RNN_BATCH_SIZE, input=input, use_gpu=False)
def _gru_test(self, layers, bidirectional, initial_state, packed_sequence,
dropout):
batch_first = True if packed_sequence == 2 else False
model = torch.nn.GRU(RNN_INPUT_SIZE,
RNN_HIDDEN_SIZE,
layers,
bidirectional=bidirectional,
dropout=dropout,
batch_first=batch_first)
if packed_sequence == 1:
model = RnnModelWithPackedSequence(model, False)
if packed_sequence == 2:
model = RnnModelWithPackedSequence(model, True)
def make_input(batch_size):
seq_lengths = np.random.randint(1,
RNN_SEQUENCE_LENGTH + 1,
size=batch_size)
seq_lengths = list(reversed(sorted(map(int, seq_lengths))))
inputs = [torch.randn(l, RNN_INPUT_SIZE) for l in seq_lengths]
inputs = rnn_utils.pad_sequence(inputs, batch_first=batch_first)
inputs = [inputs]
directions = 2 if bidirectional else 1
if initial_state:
h0 = torch.randn(directions * layers, batch_size,
RNN_HIDDEN_SIZE)
inputs.append(h0)
if packed_sequence != 0:
inputs.append(torch.IntTensor(seq_lengths))
if len(inputs) == 1:
input = inputs[0]
else:
input = tuple(inputs)
return input
input = make_input(RNN_BATCH_SIZE)
self.run_test(
model,
input,
batch_size=RNN_BATCH_SIZE,
)
# test that the model still runs with a different batch size
other_input = make_input(RNN_BATCH_SIZE + 1)
self.run_test(model, other_input, batch_size=RNN_BATCH_SIZE + 1)
def _elman_rnn_test(self, layers, nonlinearity, bidirectional,
initial_state, packed_sequence, dropout):
model = nn.RNN(RNN_INPUT_SIZE,
RNN_HIDDEN_SIZE,
layers,
nonlinearity=nonlinearity,
bidirectional=bidirectional,
dropout=dropout)
if packed_sequence == 1:
model = RnnModelWithPackedSequence(model, False)
if packed_sequence == 2:
model = RnnModelWithPackedSequence(model, True)
seq_lengths = np.random.randint(1,
RNN_SEQUENCE_LENGTH + 1,
size=RNN_BATCH_SIZE)
seq_lengths = list(reversed(sorted(map(int, seq_lengths))))
inputs = [
Variable(torch.randn(l, RNN_INPUT_SIZE)) for l in seq_lengths
]
inputs = rnn_utils.pad_sequence(inputs)
if packed_sequence == 2:
inputs = inputs.transpose(0, 1)
inputs = [inputs]
directions = 2 if bidirectional else 1
if initial_state:
h0 = Variable(
torch.randn(directions * layers, RNN_BATCH_SIZE,
RNN_HIDDEN_SIZE))
inputs.append(h0)
if packed_sequence != 0:
inputs.append(Variable(torch.IntTensor(seq_lengths)))
if len(inputs) == 1:
input = inputs[0]
else:
input = tuple(inputs)
self.run_model_test(model,
train=False,
batch_size=RNN_BATCH_SIZE,
input=input,
use_gpu=False)