def test_record_new_flags(self):
with utils_impl.record_new_flags() as hparam_flags:
flags.DEFINE_string('exp_name', 'name',
'Unique name for the experiment.')
flags.DEFINE_float('learning_rate', 0.1,
'Optimizer learning rate.')
self.assertCountEqual(hparam_flags, ['exp_name', 'learning_rate'])
def test_record_new_flags(self):
with utils_impl.record_new_flags() as hparam_flags:
flags.DEFINE_string('exp_name', 'name',
'Unique name for the experiment.')
flags.DEFINE_integer('random_seed', 0,
'Random seed for the experiment.')
self.assertCountEqual(hparam_flags, ['exp_name', 'random_seed'])
def test_convert_flag_names_to_odict(self):
with utils_impl.record_new_flags() as hparam_flags:
flags.DEFINE_integer('flag1', 1, 'This is the first flag.')
flags.DEFINE_float('flag2', 2.0, 'This is the second flag.')
hparam_odict = utils_impl.lookup_flag_values(hparam_flags)
expected_odict = collections.OrderedDict(flag1=1, flag2=2.0)
self.assertEqual(hparam_odict, expected_odict)
from absl import flags
from absl import logging
import pandas as pd
import tensorflow as tf
import tensorflow_federated as tff
from tensorflow_federated.python.research.optimization.shakespeare import dataset
from tensorflow_federated.python.research.optimization.shakespeare import models
from tensorflow_federated.python.research.optimization.shared import keras_callbacks
from tensorflow_federated.python.research.optimization.shared import keras_metrics
from tensorflow_federated.python.research.optimization.shared import optimizer_utils
from tensorflow_federated.python.research.utils import utils_impl
FLAGS = flags.FLAGS
with utils_impl.record_new_flags() as hparam_flags:
optimizer_utils.define_optimizer_flags('centralized')
flags.DEFINE_string(
'experiment_name', None,
'Name of the experiment. Part of the name of the output '
'directory.')
flags.DEFINE_integer('num_epochs', 60, 'Number of epochs to train.')
flags.DEFINE_integer('batch_size', 10,
'Size of batches for training and eval.')
flags.DEFINE_boolean('shuffle_train_data', True,
'Whether to shuffle the training data.')
flags.DEFINE_string(
'root_output_dir', '/tmp/tff/optimization/shakespeare/centralized',
'The top-level output directory experiment runs. --experiment_name will '
'be append, and the directory will contain tensorboard logs, metrics CSVs '
flags.DEFINE_float('clip', 0.05, 'Initial clip.')
flags.DEFINE_float('noise_multiplier', None,
'Noise multiplier. If None, no DP is used.')
flags.DEFINE_float('adaptive_clip_learning_rate', 0,
'Adaptive clip learning rate.')
flags.DEFINE_float('target_unclipped_quantile', 0.5,
'Target unclipped quantile.')
flags.DEFINE_float(
'clipped_count_budget_allocation', 0.1,
'Fraction of privacy budget to allocate for clipped counts.')
flags.DEFINE_boolean(
'per_vector_clipping', False, 'Use per-vector clipping'
'to indepednelty clip each weight tensor instead of the'
'entire model.')
with utils_impl.record_new_flags() as training_loop_flags:
flags.DEFINE_integer('total_rounds', 200, 'Number of total training rounds.')
flags.DEFINE_string(
'experiment_name', None, 'The name of this experiment. Will be append to '
'--root_output_dir to separate experiment results.')
flags.DEFINE_string('root_output_dir', '/tmp/differential_privacy/',
'Root directory for writing experiment output.')
flags.DEFINE_boolean(
'write_metrics_with_bz2', True, 'Whether to use bz2 '
'compression when writing output metrics to a csv file.')
flags.DEFINE_integer(
'rounds_per_eval', 1,
'How often to evaluate the global model on the validation dataset.')
flags.DEFINE_integer('rounds_per_checkpoint', 50,
'How often to checkpoint the global model.')
flags.DEFINE_integer(
