|
@@ 421-442 (lines=22) @@
|
| 418 |
|
saver.restore(session, os.path.join(summaries_dir, 'best.ckpt')) |
| 419 |
|
logger.debug('Total Epoch: %d, current batch %d', injector.num_epochs, injector.cur_batch) |
| 420 |
|
|
| 421 |
|
def predict_proba(self, x, session=None, writer=None, writer_id=None): |
| 422 |
|
"""Predict probability (Softmax) |
| 423 |
|
""" |
| 424 |
|
if session is None: |
| 425 |
|
if self.sess is None: |
| 426 |
|
session = tf.Session() |
| 427 |
|
self.sess = session |
| 428 |
|
else: |
| 429 |
|
session = self.sess |
| 430 |
|
targets = [self.y] |
| 431 |
|
if writer is not None: |
| 432 |
|
targets += [self.merged] |
| 433 |
|
results = session.run(targets, |
| 434 |
|
feed_dict={self.x: x.reshape(tuple([1]) + x.shape), |
| 435 |
|
self.length: np.array([x.shape[0]], dtype=np.int), |
| 436 |
|
self.initial_state_c: np.zeros((1, self.num_units)), |
| 437 |
|
self.initial_state_h: np.zeros((1, self.num_units))}) |
| 438 |
|
if writer is not None: |
| 439 |
|
writer.add_summary(results[1], writer_id) |
| 440 |
|
batch_y = results[0] |
| 441 |
|
# Get result |
| 442 |
|
return batch_y[0, :, :] |
| 443 |
|
|
| 444 |
|
def predict(self, x, session=None, writer=None, writer_id=None): |
| 445 |
|
if session is None: |
|
@@ 444-463 (lines=20) @@
|
| 441 |
|
# Get result |
| 442 |
|
return batch_y[0, :, :] |
| 443 |
|
|
| 444 |
|
def predict(self, x, session=None, writer=None, writer_id=None): |
| 445 |
|
if session is None: |
| 446 |
|
if self.sess is None: |
| 447 |
|
session = tf.Session() |
| 448 |
|
self.sess = session |
| 449 |
|
else: |
| 450 |
|
session = self.sess |
| 451 |
|
targets = [self.y_class] |
| 452 |
|
if writer is not None: |
| 453 |
|
targets += [self.merged] |
| 454 |
|
results = session.run(targets, |
| 455 |
|
feed_dict={self.x: x.reshape(tuple([1]) + x.shape), |
| 456 |
|
self.length: np.array([x.shape[0]], dtype=np.int), |
| 457 |
|
self.initial_state_c: np.zeros((1, self.num_units)), |
| 458 |
|
self.initial_state_h: np.zeros((1, self.num_units))}) |
| 459 |
|
if writer is not None: |
| 460 |
|
writer.add_summary(results[1], writer_id) |
| 461 |
|
batch_y = results[0] |
| 462 |
|
# Get result |
| 463 |
|
return batch_y[0, :] |
| 464 |
|
|
| 465 |
|
def predict_accuracy(self, x, y, session=None, writer=None, writer_id=None, with_loss=False): |
| 466 |
|
"""Get Accuracy given feature array and corresponding labels |
TinghuiWang /
pyActLearn
