def create_train_model(hyper, **kwargs):
model_setup = dict(
g=tf.Graph(),
# per_process_gpu_memory_fraction=0.6,
# n_samples_per_ref=3,
block_size_x=8 * 3 * 50 // 2,
block_size_y=80,
num_blocks=1,
batch_size=32,
max_reach=8 * 20, # 160
queue_cap=300,
overwrite=False,
reuse=False,
shrink_factor=8,
dtype=tf.float32,
model_fn=model_fn,
in_data=input_readers.MinCallAlignedRaw(),
lr_fn=lambda global_step: tf.train.exponential_decay(
hyper['initial_lr'], global_step, 100000, hyper['decay_factor']),
hyper=hyper,
)
model_setup.update(kwargs)
return Model(**model_setup)
def create_train_model(hyper, **kwargs):
model_setup = model_setup_params(hyper)
model_setup.update(kwargs)
return Model(**model_setup)