def create_prover(options: deephol_pb2.ProverOptions) -> Prover:
"""Creates a Prover object, initializing all dependencies."""
theorem_database = io_util.load_theorem_database_from_file(
str(options.path_theorem_database))
tactics = io_util.load_tactics_from_file(str(options.path_tactics),
str(options.path_tactics_replace))
if options.action_generator_options.asm_meson_no_params_only:
tf.logging.warn(
'Note: Using Meson action generator with no parameters.')
action_gen = action_generator.MesonActionGenerator()
else:
predictor = get_predictor(options)
emb_store = None
if options.HasField('theorem_embeddings'):
emb_store = embedding_store.TheoremEmbeddingStore(predictor)
emb_store.read_embeddings(str(options.theorem_embeddings))
assert emb_store.thm_embeddings.shape[0] == len(
theorem_database.theorems)
action_gen = action_generator.ActionGenerator(
theorem_database, tactics, predictor,
options.action_generator_options, options.model_architecture,
emb_store)
hol_wrapper = setup_prover(theorem_database)
tf.logging.info('DeepHOL dependencies initialization complete.')
if options.prover == 'bfs':
return BFSProver(options, hol_wrapper, action_gen, theorem_database)
return NoBacktrackProver(options, hol_wrapper, action_gen,
theorem_database)
def _verify_prover_options(prover_options: deephol_pb2.ProverOptions) -> None:
"""Asserts some (incomplete) consistency requirements over prover_options."""
for field_name in [
'path_tactics', 'path_tactics_replace', 'path_theorem_database',
'path_model_prefix'
]:
if not prover_options.HasField(field_name):
tf.logging.fatal('Missing field "%s" in ProverOptions', field_name)
if prover_options.prover not in ['nobacktrack', 'bfs']:
tf.logging.fatal('Unsupported proof strategy: "%s"', prover_options.prover)
history_dependent = [HIST_AVG, HIST_CONV, HIST_ATT]
if prover_options.model_architecture in history_dependent:
if not prover_options.path_emb_model_prefix:
tf.logging.fatal(
'History dependent model %s requires embeddings checkpoint '
'path_emb_model_prefix.',
deephol_pb2.ProverOptions.ModelArchitecture.Name(
prover_options.model_architecture))
# Light assertions on file naming conventions for embedding consistency.
# Embedding checkpoint number should be the end of the input file.
emb_checkpoint_num = next(
re.finditer(r'\d+$', prover_options.path_emb_model_prefix)).group(0)
if emb_checkpoint_num not in prover_options.path_model_prefix:
tf.logging.fatal(
'Embeddings checkpoint number (%s) was not found '
'in the path of predictions checkpoint (%s), indicating '
'it was trained with different embeddings.', emb_checkpoint_num,
prover_options.path_model_prefix)
def cache_embeddings(options: deephol_pb2.ProverOptions):
emb_path = str(options.theorem_embeddings)
if options.HasField(
'theorem_embeddings') and not tf.gfile.Exists(emb_path):
tf.logging.info(
'theorem_embeddings file "%s" does not exist, computing & saving.',
emb_path)
emb_store = embedding_store.TheoremEmbeddingStore(
get_predictor(options))
emb_store.compute_embeddings_for_thms_from_db_file(
str(options.path_theorem_database))
emb_store.save_embeddings(emb_path)
def _sample_bfs_options(prover_options: deephol_pb2.ProverOptions):
"""Sample parameters according the meta options."""
if not prover_options.HasField('bfs_options'):
return
options = prover_options.bfs_options
if options.HasField('meta_options'):
meta_options = options.meta_options
if meta_options.HasField('max_top_suggestions'):
options.max_top_suggestions = _sample_from_interval(
meta_options.max_top_suggestions)
if meta_options.HasField('max_successful_branches'):
options.max_successful_branches = _sample_from_interval(
meta_options.max_successful_branches)
if meta_options.HasField('max_explored_nodes'):
options.max_explored_nodes = _sample_from_interval(
meta_options.max_explored_nodes)
if meta_options.HasField('min_successful_branches'):
options.min_successful_branches = _sample_from_interval(
meta_options.min_successful_branches)
if meta_options.HasField('max_theorem_parameters'):
prover_options.action_generator_options.max_theorem_parameters = (
_sample_from_interval(meta_options.max_theorem_parameters))