def build_network(self, inputs, phase_train=True, nclass=1001):
try:
from official.recommendation import neumf_model # pylint: disable=g-import-not-at-top
except ImportError as e:
if 'neumf_model' not in e.message:
raise
raise ImportError('To use the experimental NCF model, you must clone the '
'repo https://github.com/tensorflow/models and add '
'tensorflow/models to the PYTHONPATH.')
del nclass
users, items, _ = inputs
params = {
'num_users': _NUM_USERS_20M,
'num_items': _NUM_ITEMS_20M,
'model_layers': (256, 256, 128, 64),
'mf_dim': 64,
'mf_regularization': 0,
'mlp_reg_layers': (0, 0, 0, 0),
'use_tpu': False
}
if self.data_type == tf.float32:
keras_model = neumf_model.construct_model(users, items, params)
logits = keras_model.output
else:
assert self.data_type == tf.float16
tf.keras.backend.set_floatx('float16')
# We cannot rely on the variable_scope's fp16 custom getter here, because
# the NCF model uses keras layers, which ignore variable scopes. So we use
# a variable_creator_scope instead.
with tf.variable_creator_scope(_fp16_variable_creator):
keras_model = neumf_model.construct_model(users, items, params)
logits = tf.cast(keras_model.output, tf.float32)
return model.BuildNetworkResult(logits=logits, extra_info=None)
def _get_keras_model(params):
"""Constructs and returns the model."""
batch_size = params['batch_size']
user_input = tf.keras.layers.Input(shape=(),
batch_size=batch_size,
name=movielens.USER_COLUMN,
dtype=rconst.USER_DTYPE)
item_input = tf.keras.layers.Input(shape=(),
batch_size=batch_size,
name=movielens.ITEM_COLUMN,
dtype=rconst.ITEM_DTYPE)
base_model = neumf_model.construct_model(user_input, item_input, params)
base_model_output = base_model.output
zeros = tf.keras.layers.Lambda(lambda x: x * 0)(base_model_output)
softmax_logits = tf.keras.layers.concatenate([zeros, base_model_output],
axis=-1)
keras_model = tf.keras.Model(inputs=[user_input, item_input],
outputs=softmax_logits)
keras_model.summary()
return keras_model
def _get_keras_model(params):
"""Constructs and returns the model."""
batch_size = params['batch_size']
user_input = tf.keras.layers.Input(
shape=(),
batch_size=batch_size,
name=movielens.USER_COLUMN,
dtype=rconst.USER_DTYPE)
item_input = tf.keras.layers.Input(
shape=(),
batch_size=batch_size,
name=movielens.ITEM_COLUMN,
dtype=rconst.ITEM_DTYPE)
base_model = neumf_model.construct_model(user_input, item_input, params)
base_model_output = base_model.output
zeros = tf.keras.layers.Lambda(
lambda x: x * 0)(base_model_output)
softmax_logits = tf.keras.layers.concatenate(
[zeros, base_model_output],
axis=-1)
keras_model = tf.keras.Model(
inputs=[user_input, item_input],
outputs=softmax_logits)
keras_model.summary()
return keras_model
def build_network(self, inputs, phase_train=True, nclass=1001,
data_type=tf.float32):
try:
from official.recommendation import neumf_model # pylint: disable=g-import-not-at-top
except ImportError as e:
if 'neumf_model' not in e.message:
raise
raise ImportError('To use the experimental NCF model, you must clone the '
'repo https://github.com/tensorflow/models and add '
'tensorflow/models to the PYTHONPATH.')
del nclass
if data_type != tf.float32:
raise ValueError('NCF model only supports float32 for now.')
users, items = inputs
params = {
'num_users': _NUM_USERS_20M,
'num_items': _NUM_ITEMS_20M,
'model_layers': (256, 256, 128, 64),
'mf_dim': 64,
'mf_regularization': 0,
'mlp_reg_layers': (0, 0, 0, 0),
}
logits = neumf_model.construct_model(users, items, params)
return model.BuildNetworkResult(logits=logits, extra_info=None)
def build_network(self,
images,
phase_train=True,
nclass=1001,
data_type=tf.float32):
try:
from official.recommendation import neumf_model # pylint: disable=g-import-not-at-top
except ImportError:
raise ImportError(
'To use the experimental NCF model, you must clone the '
'repo https://github.com/tensorflow/models and add '
'tensorflow/models to the PYTHONPATH.')
del nclass
if data_type != tf.float32:
raise ValueError('NCF model only supports float32 for now.')
batch_size = int(images.shape[0])
# Create synthetic users and items. tf_cnn_benchmarks only passes images to
# this function, which we cannot use in the NCF model. We use functions as
# initializers for XLA compatibility.
def users_init_val():
return tf.random_uniform((batch_size, ),
minval=0,
maxval=_NUM_USERS_20M,
dtype=tf.int32)
users = tf.Variable(users_init_val,
dtype=tf.int32,
trainable=False,
collections=[tf.GraphKeys.LOCAL_VARIABLES],
name='synthetic_users')
def items_init_val():
return tf.random_uniform((batch_size, ),
minval=0,
maxval=_NUM_ITEMS_20M,
dtype=tf.int32)
items = tf.Variable(items_init_val,
dtype=tf.int32,
trainable=False,
collections=[tf.GraphKeys.LOCAL_VARIABLES],
name='synthetic_items')
params = {
'num_users': _NUM_USERS_20M,
'num_items': _NUM_ITEMS_20M,
'model_layers': (256, 256, 128, 64),
'mf_dim': 64,
'mf_regularization': 0,
'mlp_reg_layers': (0, 0, 0, 0),
}
logits = neumf_model.construct_model(users, items, params)
return model.BuildNetworkResult(logits=logits, extra_info=None)
def _get_keras_model(params):
"""Constructs and returns the model."""
batch_size = params["batch_size"]
user_input = tf.keras.layers.Input(shape=(1, ),
name=movielens.USER_COLUMN,
dtype=tf.int32)
item_input = tf.keras.layers.Input(shape=(1, ),
name=movielens.ITEM_COLUMN,
dtype=tf.int32)
valid_pt_mask_input = tf.keras.layers.Input(shape=(1, ),
name=rconst.VALID_POINT_MASK,
dtype=tf.bool)
dup_mask_input = tf.keras.layers.Input(shape=(1, ),
name=rconst.DUPLICATE_MASK,
dtype=tf.int32)
label_input = tf.keras.layers.Input(shape=(1, ),
name=rconst.TRAIN_LABEL_KEY,
dtype=tf.bool)
base_model = neumf_model.construct_model(user_input, item_input, params)
logits = base_model.output
zeros = tf.keras.layers.Lambda(lambda x: x * 0)(logits)
softmax_logits = tf.keras.layers.concatenate([zeros, logits], axis=-1)
# Custom training loop calculates loss and metric as a part of
# training/evaluation step function.
if not params["keras_use_ctl"]:
softmax_logits = MetricLayer(
params["match_mlperf"])([softmax_logits, dup_mask_input])
# TODO(b/134744680): Use model.add_loss() instead once the API is well
# supported.
softmax_logits = LossLayer(batch_size)(
[softmax_logits, label_input, valid_pt_mask_input])
keras_model = tf.keras.Model(inputs={
movielens.USER_COLUMN: user_input,
movielens.ITEM_COLUMN: item_input,
rconst.VALID_POINT_MASK: valid_pt_mask_input,
rconst.DUPLICATE_MASK: dup_mask_input,
rconst.TRAIN_LABEL_KEY: label_input
},
outputs=softmax_logits)
keras_model.summary()
return keras_model
def _get_keras_model(params):
"""Constructs and returns the model."""
batch_size = params["batch_size"]
# The input layers are of shape (1, batch_size), to match the size of the
# input data. The first dimension is needed because the input data are
# required to be batched to use distribution strategies, and in this case, it
# is designed to be of batch_size 1 for each replica.
user_input = tf.keras.layers.Input(
shape=(batch_size,),
batch_size=params["batches_per_step"],
name=movielens.USER_COLUMN,
dtype=tf.int32)
item_input = tf.keras.layers.Input(
shape=(batch_size,),
batch_size=params["batches_per_step"],
name=movielens.ITEM_COLUMN,
dtype=tf.int32)
base_model = neumf_model.construct_model(
user_input, item_input, params, need_strip=True)
base_model_output = base_model.output
logits = tf.keras.layers.Lambda(
lambda x: tf.expand_dims(x, 0),
name="logits")(base_model_output)
zeros = tf.keras.layers.Lambda(
lambda x: x * 0)(logits)
softmax_logits = tf.keras.layers.concatenate(
[zeros, logits],
axis=-1)
keras_model = tf.keras.Model(
inputs=[user_input, item_input],
outputs=softmax_logits)
keras_model.summary()
return keras_model
def _get_keras_model(params):
"""Constructs and returns the model."""
batch_size = params['batch_size']
# The input layers are of shape (1, batch_size), to match the size of the
# input data. The first dimension is needed because the input data are
# required to be batched to use distribution strategies, and in this case, it
# is designed to be of batch_size 1 for each replica.
user_input = tf.keras.layers.Input(
shape=(batch_size,),
batch_size=params["batches_per_step"],
name=movielens.USER_COLUMN,
dtype=tf.int32)
item_input = tf.keras.layers.Input(
shape=(batch_size,),
batch_size=params["batches_per_step"],
name=movielens.ITEM_COLUMN,
dtype=tf.int32)
base_model = neumf_model.construct_model(
user_input, item_input, params, need_strip=True)
base_model_output = base_model.output
logits = tf.keras.layers.Lambda(
lambda x: tf.expand_dims(x, 0),
name="logits")(base_model_output)
zeros = tf.keras.layers.Lambda(
lambda x: x * 0)(logits)
softmax_logits = tf.keras.layers.concatenate(
[zeros, logits],
axis=-1)
keras_model = tf.keras.Model(
inputs=[user_input, item_input],
outputs=softmax_logits)
keras_model.summary()
return keras_model
def _get_keras_model(params):
"""Constructs and returns the model."""
batch_size = params["batch_size"]
# The input layers are of shape (1, batch_size), to match the size of the
# input data. The first dimension is needed because the input data are
# required to be batched to use distribution strategies, and in this case, it
# is designed to be of batch_size 1 for each replica.
user_input = tf.keras.layers.Input(
shape=(batch_size,),
batch_size=params["batches_per_step"],
name=movielens.USER_COLUMN,
dtype=tf.int32)
item_input = tf.keras.layers.Input(
shape=(batch_size,),
batch_size=params["batches_per_step"],
name=movielens.ITEM_COLUMN,
dtype=tf.int32)
valid_pt_mask_input = tf.keras.layers.Input(
shape=(batch_size,),
batch_size=params["batches_per_step"],
name=rconst.VALID_POINT_MASK,
dtype=tf.bool)
dup_mask_input = tf.keras.layers.Input(
shape=(batch_size,),
batch_size=params["batches_per_step"],
name=rconst.DUPLICATE_MASK,
dtype=tf.int32)
label_input = tf.keras.layers.Input(
shape=(batch_size, 1),
batch_size=params["batches_per_step"],
name=rconst.TRAIN_LABEL_KEY,
dtype=tf.bool)
base_model = neumf_model.construct_model(
user_input, item_input, params, need_strip=True)
base_model_output = base_model.output
logits = tf.keras.layers.Lambda(
lambda x: tf.expand_dims(x, 0),
name="logits")(base_model_output)
zeros = tf.keras.layers.Lambda(
lambda x: x * 0)(logits)
softmax_logits = tf.keras.layers.concatenate(
[zeros, logits],
axis=-1)
softmax_logits = MetricLayer(params)([softmax_logits, dup_mask_input])
keras_model = tf.keras.Model(
inputs={
movielens.USER_COLUMN: user_input,
movielens.ITEM_COLUMN: item_input,
rconst.VALID_POINT_MASK: valid_pt_mask_input,
rconst.DUPLICATE_MASK: dup_mask_input,
rconst.TRAIN_LABEL_KEY: label_input},
outputs=softmax_logits)
loss_obj = tf.keras.losses.SparseCategoricalCrossentropy(
from_logits=True,
reduction="sum")
keras_model.add_loss(loss_obj(
y_true=label_input,
y_pred=softmax_logits,
sample_weight=valid_pt_mask_input) * 1.0 / batch_size)
keras_model.summary()
return keras_model