def run(self, qual_name: str):
inputs, output, funcs, seqs = self.benchmark.unwrap()
iospec: IOSpec = IOSpec(inputs, output)
iospec.funcs = funcs
iospec.seqs = seqs
model = RelGraphInterface.from_model_dir(self.model_dir)
if self.cmd_args.use_old_featurization:
from autopandas_v2.ml.featurization_old.featurizer import RelationGraph
from autopandas_v2.ml.featurization_old.options import GraphOptions
else:
from autopandas_v2.ml.featurization.featurizer import RelationGraph
from autopandas_v2.ml.featurization.options import GraphOptions
options = GraphOptions()
graph: RelationGraph = RelationGraph(options)
graph.from_input_output(inputs, output)
encoding = graph.get_encoding(get_mapping=False)
str_seqs, probs = list(
zip(*model.predict_graphs([encoding], top_k=self.top_k)[0]))
ground_truth = ':'.join(funcs[i] for i in seqs[0])
pos = -1
for idx, pred in enumerate(str_seqs, 1):
if pred == ground_truth:
pos = idx
break
return {
'benchmark': qual_name,
'ground_truth': ground_truth,
'rank': pos
}
def run(self, qual_name: str):
inputs, output, funcs, seqs = self.benchmark.unwrap()
iospec: IOSpec = IOSpec(inputs, output)
iospec.funcs = funcs
iospec.seqs = seqs
engine: NeuralEngine = NeuralEngine(iospec,
self.function_model_dir,
top_k=self.cmd_args.top_k_function)
engine.max_depth = max(len(i) for i in seqs)
engine.stop_first_solution = True
engine.use_spec_funcs = True
engine.use_spec_seqs = True
engine.argument_engine = 'beam-search'
engine.arg_model_dir = self.arg_model_dir
engine.arg_top_k = self.cmd_args.top_k_args
engine.use_old_featurization = self.cmd_args.use_old_featurization
start_time = time.time()
solution_found = engine.search()
return {
'benchmark': qual_name,
'num_seqs_explored': engine.stats.num_seqs_explored,
'num_candidates_generated': dict(engine.stats.num_cands_generated),
'solution_found': solution_found,
'time': time.time() - start_time
}
def __init__(self,
func_seq: List[BaseGenerator],
cmd_args: ArgNamespace = None):
iospec: IOSpec = IOSpec(inputs=[], output=None)
super().__init__(iospec, cmd_args)
self.func_seq = func_seq
self.max_depth = len(func_seq)
self.stats = None
def check(self, inputs: List[Any], output: Any, funcs: List[str], seqs: List[List[int]], model_dir: str):
iospec: IOSpec = IOSpec(inputs, output)
iospec.funcs = funcs
iospec.seqs = seqs
engine: BFSEngine = BFSEngine(iospec)
engine.max_depth = max(len(i) for i in seqs)
engine.silent = True
engine.stop_first_solution = True
engine.use_spec_funcs = True
engine.use_spec_seqs = True
engine.argument_engine = 'beam-search'
engine.arg_model_dir = model_dir
# try:
# engine.search()
# except Exception as e:
# logging.exception(e)
return self.check_engine(engine)
def run(self, qual_name: str):
inputs, output, funcs, seqs = self.benchmark.unwrap()
iospec: IOSpec = IOSpec(inputs, output)
iospec.funcs = funcs
iospec.seqs = seqs
engine: BFSEngine = BFSEngine(iospec)
engine.max_depth = max(len(i) for i in seqs)
engine.silent = True
engine.stop_first_solution = True
engine.use_spec_funcs = True
engine.use_spec_seqs = True
engine.argument_engine = 'beam-search'
engine.arg_model_dir = self.model_dir
solution_found = engine.search()
return {
'benchmark': qual_name,
'num_candidates_generated': dict(engine.stats.num_cands_generated),
'solution_found': solution_found
}
def run(self, qual_name: str):
inputs, output, funcs, seqs = self.benchmark.unwrap()
iospec: IOSpec = IOSpec(inputs, output)
iospec.funcs = funcs
iospec.seqs = seqs
if self.cmd_args.engine == 'neural':
engine: NeuralEngine = NeuralEngine(
iospec,
self.function_model_dir,
top_k=self.cmd_args.top_k_function)
engine.max_depth = max(len(i) for i in seqs)
engine.stop_first_solution = True
engine.use_spec_funcs = True
engine.use_spec_seqs = True
engine.argument_engine = 'beam-search'
engine.arg_model_dir = self.arg_model_dir
engine.arg_top_k = self.cmd_args.top_k_args
engine.use_old_featurization = self.cmd_args.use_old_featurization
engine.model_store = self.model_store
else:
raise Exception("Engine {} not recognized".format(
self.cmd_args.engine))
start_time = time.time()
solution_found = engine.search()
return {
'benchmark': qual_name,
'ground_truth_depth': min(len(i) for i in seqs),
'num_seqs_explored': engine.stats.num_seqs_explored,
'num_candidates_generated': dict(engine.stats.num_cands_generated),
'solution_found': solution_found,
'solution':
repr(str(engine.solutions[0])) if solution_found else '',
'time': time.time() - start_time
}