def main(unused_argv):
"""Main function that trains and evaluates the translation model"""
hparams = create_hparams(FLAGS)
os.environ['CUDA_VISIBLE_DEVICES'] = str(hparams.device)
print('Building models')
train_model, eval_model, encode_model = build_models(hparams)
train_loop(train_model, eval_model, encode_model, hparams)
def main(unused_argv):
"""Main function that trains and evaluats the translation model"""
hparams = create_hparams(FLAGS)
os.environ['CUDA_VISIBLE_DEVICES'] = '0,1,2,3'
train_model = build_models(hparams)
test_model = build_models(hparams, 'EVAL')
encode_model = build_models(hparams, 'ENCODE')
train_loop(train_model=train_model,
eval_model=test_model,
encoder_model=encode_model,
hparams=hparams)
def __init__(self,
model_dir=_default_model_dir,
use_gpu=True,
batch_size=256,
gpu_mem_frac=0.1,
beam_width=10,
num_top=1,
maximum_iterations=1000,
cpu_threads=5,
emb_activation=None):
"""Constructor for the inference model.
Args:
model_dir: Path to the model directory.
use_gpu: Flag for GPU usage.
batch_size: Number of samples to process per step.
gpu_mem_frac: If GPU is used, what memory fraction should be used?
beam_width: Width of the the window used for the beam search decoder.
num_top: Number of most probable sequnces as output of the beam search decoder.
emb_activation: Activation function used in the bottleneck layer.
Returns:
None
"""
self.num_top = num_top
self.use_gpu = use_gpu
parser = argparse.ArgumentParser()
add_arguments(parser)
flags = parser.parse_args([])
flags.hparams_from_file = True
flags.save_dir = model_dir
self.hparams = create_hparams(flags)
self.hparams.set_hparam("save_dir", model_dir)
self.hparams.set_hparam("batch_size", batch_size)
self.hparams.set_hparam("gpu_mem_frac", gpu_mem_frac)
self.hparams.add_hparam("beam_width", beam_width)
self.hparams.set_hparam("cpu_threads", cpu_threads)
self.encode_model, self.decode_model = build_models(
self.hparams, modes=["ENCODE", "DECODE"])
self.maximum_iterations = maximum_iterations
def main(unused_argv):
# Replace MODEL_DIR with the folder current run to resume training from a set of hyperparameters
# MODEL_DIR = '/Users/eduardolitonjua/Desktop/Retrieval-System/runs/1472130056'
hparams = hyperparameters.create_hparams()
model_fn = model.create_model_fn(
hparams,
model_impl=dual_encoder_model)
estimator = tf.contrib.learn.Estimator(
model_fn=model_fn,
model_dir=MODEL_DIR,
config=tf.contrib.learn.RunConfig())
input_fn_train = inputs.create_input_fn(
mode=tf.contrib.learn.ModeKeys.TRAIN,
input_files=[TRAIN_FILE],
batch_size=hparams.batch_size,
num_epochs=FLAGS.num_epochs)
input_fn_eval = inputs.create_input_fn(
mode=tf.contrib.learn.ModeKeys.EVAL,
input_files=[VALIDATION_FILE],
batch_size=hparams.eval_batch_size,
num_epochs=1)
eval_metrics = metrics.create_evaluation_metrics()
class EvaluationMonitor(tf.contrib.learn.monitors.EveryN):
def every_n_step_end(self, step, outputs):
self._estimator.evaluate(
input_fn=input_fn_eval,
metrics=eval_metrics,
steps=None)
eval_monitor = EvaluationMonitor(every_n_steps=FLAGS.eval_every, first_n_steps=-1)
estimator.fit(input_fn=input_fn_train, steps=None, monitors=[eval_monitor])
import tensorflow as tf
import os
import numpy as np
import time
import scipy
import nibabel as nib
import hyperparameters
from model_3D import deep_model_128in as model
from data_reformating_3D import reformat_eval_data_3D as reformat_data_3D
from loss import loss
# Get hyperparameters
hparams = hyperparameters.create_hparams()
# Set data input path
test_data_dir = hparams.input_dir + '/test'
# Load preprocessed input data
images_dict, labels_dict = load_data_eval(test_data_dir)
# Extract images from dict
images = []
for k, v in images_dict.items():
images.append(v.get_data())
# Extract labels from dict
labels = []