czzhangheng
/
DCRNN
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fixed bugs with refactoring

This commit is contained in:
Chintan Shah 2019-10-04 16:05:52 -04:00
parent 2b8d5e6b31
commit 20c6aa5862
2 changed files with 8 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
model/pytorch/dcrnn_model.py
View File

@ -67,7 +67,6 @@ class DecoderModel(nn.Module, Seq2SeqAttrs):
nn.Module.__init__(self) nn.Module.__init__(self)
Seq2SeqAttrs.__init__(self, adj_mx, **model_kwargs) Seq2SeqAttrs.__init__(self, adj_mx, **model_kwargs)
self.output_dim = int(model_kwargs.get('output_dim', 1)) self.output_dim = int(model_kwargs.get('output_dim', 1))
self.use_curriculum_learning = bool(model_kwargs.get('use_curriculum_learning', False))
self.horizon = int(model_kwargs.get('horizon', 1)) # for the decoder self.horizon = int(model_kwargs.get('horizon', 1)) # for the decoder
self.projection_layer = nn.Linear(self.hidden_state_size, self.num_nodes * self.output_dim) self.projection_layer = nn.Linear(self.hidden_state_size, self.num_nodes * self.output_dim)
self.dcgru_layers = nn.ModuleList([nn.GRUCell(input_size=self.num_nodes * self.output_dim, self.dcgru_layers = nn.ModuleList([nn.GRUCell(input_size=self.num_nodes * self.output_dim,
@ -105,8 +104,14 @@ class DCRNNModel(nn.Module, Seq2SeqAttrs):
Seq2SeqAttrs.__init__(self, adj_mx, **model_kwargs) Seq2SeqAttrs.__init__(self, adj_mx, **model_kwargs)
self.encoder_model = EncoderModel(adj_mx, **model_kwargs) self.encoder_model = EncoderModel(adj_mx, **model_kwargs)
self.decoder_model = DecoderModel(adj_mx, **model_kwargs) self.decoder_model = DecoderModel(adj_mx, **model_kwargs)
self.cl_decay_steps = int(model_kwargs.get('cl_decay_steps', 1000))
self.use_curriculum_learning = bool(model_kwargs.get('use_curriculum_learning', False))
self._logger = logger self._logger = logger
def _compute_sampling_threshold(self, batches_seen):
return self.cl_decay_steps / (
self.cl_decay_steps + np.exp(batches_seen / self.cl_decay_steps))
def encoder(self, inputs): def encoder(self, inputs):
""" """
encoder forward pass on t time steps encoder forward pass on t time steps
@ -128,7 +133,7 @@ class DCRNNModel(nn.Module, Seq2SeqAttrs):
:return: output: (self.horizon, batch_size, self.num_nodes * self.output_dim) :return: output: (self.horizon, batch_size, self.num_nodes * self.output_dim)
""" """
batch_size = encoder_hidden_state.size(1) batch_size = encoder_hidden_state.size(1)
go_symbol = torch.zeros((batch_size, self.num_nodes * self.output_dim)) go_symbol = torch.zeros((batch_size, self.num_nodes * self.decoder_model.output_dim))
decoder_hidden_state = encoder_hidden_state decoder_hidden_state = encoder_hidden_state
decoder_input = go_symbol decoder_input = go_symbol

7
model/pytorch/dcrnn_supervisor.py
View File

@ -32,7 +32,6 @@ class DCRNNSupervisor:
self.input_dim = int(self._model_kwargs.get('input_dim', 1)) self.input_dim = int(self._model_kwargs.get('input_dim', 1))
self.seq_len = int(self._model_kwargs.get('seq_len')) # for the encoder self.seq_len = int(self._model_kwargs.get('seq_len')) # for the encoder
self.output_dim = int(self._model_kwargs.get('output_dim', 1)) self.output_dim = int(self._model_kwargs.get('output_dim', 1))
self.cl_decay_steps = int(self._model_kwargs.get('cl_decay_steps', 1000))
self.use_curriculum_learning = bool( self.use_curriculum_learning = bool(
self._model_kwargs.get('use_curriculum_learning', False)) self._model_kwargs.get('use_curriculum_learning', False))
self.horizon = int(self._model_kwargs.get('horizon', 1)) # for the decoder self.horizon = int(self._model_kwargs.get('horizon', 1)) # for the decoder
@ -94,7 +93,7 @@ class DCRNNSupervisor:
x, y = self._get_x_y(x, y) x, y = self._get_x_y(x, y)
x, y = self._get_x_y_in_correct_dims(x, y) x, y = self._get_x_y_in_correct_dims(x, y)
output = self.dcrnn_model(x, y) output = self.dcrnn_model(x, y, batches_seen)
loss = self._compute_loss(y, output, criterion) loss = self._compute_loss(y, output, criterion)
self._logger.info(loss.item()) self._logger.info(loss.item())
losses.append(loss.item()) losses.append(loss.item())
@ -143,10 +142,6 @@ class DCRNNSupervisor:
self.num_nodes * self.output_dim) self.num_nodes * self.output_dim)
return x, y return x, y
def _compute_sampling_threshold(self, batches_seen):
return self.cl_decay_steps / (
self.cl_decay_steps + np.exp(batches_seen / self.cl_decay_steps))
def _compute_loss(self, y_true, y_predicted, criterion): def _compute_loss(self, y_true, y_predicted, criterion):
loss = 0 loss = 0
for t in range(self.horizon): for t in range(self.horizon):