def __init__(self, seed, lr, optimizer=None): self.lr = lr self.seed = seed self._optimizer = optimizer self.graph = tf.Graph() with self.graph.as_default(): tf.set_random_seed(123 + self.seed) self.features, self.labels, self.train_op, self.eval_metric_ops, self.loss = self.create_model( ) self.saver = tf.train.Saver() self.sess = tf.Session(graph=self.graph, config=tf.ConfigProto( inter_op_parallelism_threads=1, intra_op_parallelism_threads=1)) self.size = graph_size(self.graph) with self.graph.as_default(): self.sess.run(tf.global_variables_initializer()) metadata = tf.RunMetadata() opts = tf.profiler.ProfileOptionBuilder.float_operation() self.flops = tf.profiler.profile(self.graph, run_meta=metadata, cmd='scope', options=opts).total_float_ops np.random.seed(self.seed)
def __init__(self, seed, lr, optimizer=None, gpu_fraction=0.2): self.lr = lr self.seed = seed self._optimizer = optimizer self.graph = tf.Graph() with self.graph.as_default(): tf.set_random_seed(123 + self.seed) self.features, self.labels, self.train_op, self.eval_metric_ops, self.loss = self.create_model() self.saver = tf.train.Saver() config=tf.ConfigProto(log_device_placement=False) # config.gpu_options.per_process_gpu_memory_fraction = gpu_fraction self.sess = tf.Session(graph=self.graph, config=config) self.size = graph_size(self.graph) with self.graph.as_default(): self.sess.run(tf.global_variables_initializer()) metadata = tf.RunMetadata() opts = tf.profiler.ProfileOptionBuilder.float_operation() self.flops = tf.profiler.profile(self.graph, run_meta=metadata, cmd='scope', options=opts).total_float_ops np.random.seed(self.seed)
def __init__(self, lr): self.lr = lr self._optimizer = None self.graph = tf.Graph() with self.graph.as_default(): self.features, self.labels, self.is_train, self.train_op, self.eval_metric_ops, self.loss, self.var_grad = self.create_model() self.saver = tf.train.Saver() config = tf.ConfigProto() config.gpu_options.allow_growth=True self.sess = tf.Session(graph=self.graph, config=config) self.size = graph_size(self.graph) #self.profiler = tf.profiler.Profiler(self.sess.graph) #self.profiler = model_analyzer.Profiler(self.sess.graph) with self.graph.as_default(): self.sess.run(tf.global_variables_initializer()) metadata = tf.RunMetadata() opts = tf.profiler.ProfileOptionBuilder.float_operation() self.flops = tf.profiler.profile(self.graph, run_meta=metadata, cmd='scope', options=opts).total_float_ops
def __init__(self, seed, lr, optimizer=None): self.lr = lr self._optimizer = optimizer self.graph = tf.Graph() with self.graph.as_default(): tf.set_random_seed(123 + seed) self.features, self.labels, self.train_op, self.eval_metric_ops, self.loss = self.create_model( ) self.gradient_op = tf.gradients(self.loss, tf.trainable_variables()) self.saver = tf.train.Saver() self.sess = tf.Session(graph=self.graph) self.size = graph_size(self.graph) with self.graph.as_default(): self.sess.run(tf.global_variables_initializer()) metadata = tf.RunMetadata() opts = tf.profiler.ProfileOptionBuilder.float_operation() self.flops = tf.profiler.profile(self.graph, run_meta=metadata, cmd='scope', options=opts).total_float_ops
def __init__(self, lr, seed, max_batch_size): self.lr = lr self._optimizer = None self.rng = random.Random(seed) self.graph = tf.Graph() if seed is not None: self.graph.seed = seed with self.graph.as_default(): self.learning_rate_tensor = tf.placeholder(tf.float32, shape=[]) self.features, self.labels, self.loss_op, self.train_op, self.eval_metric_ops = self.create_model() self.saver = tf.train.Saver() self.sess = tf.Session(graph=self.graph) self.size = graph_size(self.graph) # largest batch size for which GPU will not run out of memory self.max_batch_size = max_batch_size if max_batch_size is not None else 2 ** 14 with self.graph.as_default(): self.sess.run(tf.global_variables_initializer()) # metadata = tf.RunMetadata() # opts = tf.profiler.ProfileOptionBuilder( # tf.profiler.ProfileOptionBuilder.float_operation() # ).with_empty_output().build() # self.flops = tf.profiler.profile(self.graph, run_meta=metadata, cmd='scope', options=opts).total_float_ops self.flops = 0
def __init__(self, seed, lr, optimizer=None): self.lr = lr self._optimizer = optimizer self.graph = tf.Graph() with self.graph.as_default(): tf.set_random_seed(seed) self.features, self.labels, self.train_op, self.eval_metric_ops, self.loss, self.pred_ops = self.create_model() self.saver = tf.train.Saver() self.sess = tf.Session(graph=self.graph) self.size = graph_size(self.graph) with self.graph.as_default(): self.sess.run(tf.global_variables_initializer())