Skip to content
This repository was archived by the owner on Feb 24, 2025. It is now read-only.

updating files and model for tf2 #16

Open
wants to merge 1 commit into
base: master
Choose a base branch
from
Open

updating files and model for tf2 #16

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions addressnet/dataset.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -463,10 +463,10 @@ def dataset(filenames: [str], batch_size: int = 10, shuffle_buffer: int = 1000,
def input_fn() -> tf.data.Dataset:
ds = tf.data.TFRecordDataset(filenames, compression_type="GZIP")
ds = ds.shuffle(buffer_size=shuffle_buffer)
ds = ds.map(lambda record: tf.parse_single_example(record, features=_features), num_parallel_calls=8)
ds = ds.map(lambda record: tf.io.parse_single_example(serialized=record, features=_features), num_parallel_calls=8)
ds = ds.map(
lambda record: tf.py_func(synthesise_address, [record[k] for k in _features.keys()],
[tf.int64, tf.int64, tf.bool],
lambda record: tf.compat.v1.py_func(synthesise_address, [record[k] for k in _features.keys()],
[tf.int32, tf.int64, tf.bool],
stateful=False),
num_parallel_calls=num_parallel_calls
)
Expand Down
2 changes: 1 addition & 1 deletion addressnet/lookups.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -70,7 +70,7 @@
'LANE': 'LANE', 'LANEWAY': 'LNWY', 'LINE': 'LINE', 'LINK': 'LINK', 'LOOKOUT': 'LKT',
'LOOP': 'LOOP', 'MALL': 'MALL', 'MEANDER': 'MNDR', 'MEWS': 'MEWS', 'MOTORWAY': 'MTWY',
'NOOK': 'NOOK', 'OUTLOOK': 'OTLK', 'PARADE': 'PDE', 'PARKWAY': 'PWY', 'PASS': 'PASS',
'PASSAGE': 'PSGE', 'PATH': 'PATH', 'PATHWAY': 'PWAY', 'PIAZZA': 'PIAZ', 'PLAZA': 'PLZA',
'PASSAGE': 'PSGE', 'PATH': 'PATH', 'PATHWAY': 'PWAY', 'PIAZZA': 'PIAZ', 'PLACE': 'PL', 'PLAZA': 'PLZA',
'POCKET': 'PKT', 'POINT': 'PNT', 'PORT': 'PORT', 'PROMENADE': 'PROM', 'QUADRANT': 'QDRT',
'QUAYS': 'QYS', 'RAMBLE': 'RMBL', 'REST': 'REST', 'RETREAT': 'RTT', 'RIDGE': 'RDGE',
'RISE': 'RISE', 'ROAD': 'RD', 'ROTARY': 'RTY', 'ROUTE': 'RTE', 'ROW': 'ROW', 'RUE': 'RUE',
Expand Down
20 changes: 10 additions & 10 deletions addressnet/model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,12 +18,12 @@ def model_fn(features: Dict[str, tf.Tensor], labels: tf.Tensor, mode: str, param
rnn_size = params.get("rnn_size", 128)
rnn_layers = params.get("rnn_layers", 3)

embeddings = tf.get_variable("embeddings", dtype=tf.float32, initializer=tf.random_normal(shape=(len(vocab), 8)))
encoded_strings = tf.nn.embedding_lookup(embeddings, encoded_text)
embeddings = tf.compat.v1.get_variable("embeddings", dtype=tf.float32, initializer=tf.random.normal(shape=(len(vocab), 8)))
encoded_strings = tf.nn.embedding_lookup(params=embeddings, ids=encoded_text)

logits, loss = nnet(encoded_strings, lengths, rnn_layers, rnn_size, labels, mode == tf.estimator.ModeKeys.TRAIN)

predicted_classes = tf.argmax(logits, axis=2)
predicted_classes = tf.argmax(input=logits, axis=2)

if mode == tf.estimator.ModeKeys.PREDICT:
predictions = {
Expand All @@ -38,7 +38,7 @@ def model_fn(features: Dict[str, tf.Tensor], labels: tf.Tensor, mode: str, param
mode, loss=loss, eval_metric_ops=metrics)

if mode == tf.estimator.ModeKeys.TRAIN:
train_op = tf.train.AdamOptimizer(learning_rate=0.0001).minimize(loss, global_step=tf.train.get_global_step())
train_op = tf.compat.v1.train.AdamOptimizer(learning_rate=0.0001).minimize(loss, global_step=tf.compat.v1.train.get_global_step())
return tf.estimator.EstimatorSpec(mode, loss=loss, train_op=train_op)


Expand All @@ -57,19 +57,19 @@ def nnet(encoded_strings: tf.Tensor, lengths: tf.Tensor, rnn_layers: int, rnn_si

def rnn_cell():
probs = 0.8 if training else 1.0
return tf.contrib.rnn.DropoutWrapper(tf.contrib.cudnn_rnn.CudnnCompatibleGRUCell(rnn_size),
return tf.compat.v1.nn.rnn_cell.DropoutWrapper(tf.compat.v1.nn.rnn_cell.GRUCell(rnn_size),
state_keep_prob=probs, output_keep_prob=probs)

rnn_cell_fw = tf.nn.rnn_cell.MultiRNNCell([rnn_cell() for _ in range(rnn_layers)])
rnn_cell_bw = tf.nn.rnn_cell.MultiRNNCell([rnn_cell() for _ in range(rnn_layers)])
rnn_cell_fw = tf.compat.v1.nn.rnn_cell.MultiRNNCell([rnn_cell() for _ in range(rnn_layers)])
rnn_cell_bw = tf.compat.v1.nn.rnn_cell.MultiRNNCell([rnn_cell() for _ in range(rnn_layers)])

(rnn_output_fw, rnn_output_bw), states = tf.nn.bidirectional_dynamic_rnn(rnn_cell_fw, rnn_cell_bw, encoded_strings,
(rnn_output_fw, rnn_output_bw), states = tf.compat.v1.nn.bidirectional_dynamic_rnn(rnn_cell_fw, rnn_cell_bw, encoded_strings,
lengths, dtype=tf.float32)
rnn_output = tf.concat([rnn_output_fw, rnn_output_bw], axis=2)
logits = tf.layers.dense(rnn_output, n_labels, activation=tf.nn.elu)
logits = tf.compat.v1.layers.dense(rnn_output, n_labels, activation=tf.nn.elu)

loss = None
if labels is not None:
mask = tf.sequence_mask(lengths, dtype=tf.float32)
loss = tf.losses.softmax_cross_entropy(labels, logits, weights=mask)
loss = tf.compat.v1.losses.softmax_cross_entropy(labels, logits, weights=mask)
return logits, loss
Loading