async def start_next_training_api(
api_payload: schema.ExplanationTrainingPayload):
"""
Endpoint used to kick off training of a classifier via the next framework of learning from
explanations. Please refer to the docs or `json_schema.py` to understand both the supported
and required paramaters.
"""
params = api_payload.params
prepped_data = util_f.prepare_next_data(api_payload, lean_life=False)
label_space, unlabeled_docs, explanation_triples, ner_label_space = prepped_data
data = params.dict()
if len(ner_label_space) == 0:
ner_label_space = None
if len(unlabeled_docs) == 0:
unlabeled_docs = None
if len(explanation_triples) == 0:
explanation_triples = None
if hasattr(api_payload, "dev_data") and api_payload.dev_data is not None:
data["dev_data"] = api_payload.dev_data
for i, doc in enumerate(data["dev_data"]):
data["dev_data"][i] = doc.dict()
else:
data["dev_data"] = None
return schema.SavePathOutput(
save_path=train_next_framework(data, label_space, unlabeled_docs,
explanation_triples, ner_label_space))
async def start_trigger_training_lean_life(
lean_life_payload: schema.LeanLifeTriggerPayload,
background_tasks: BackgroundTasks):
"""
Endpoint hit by annotation tool's django api
"""
params = lean_life_payload.params
lean_life_data = lean_life_payload.lean_life_data
prepped_data = util_f.prepare_next_data(lean_life_data,
project_type=params.project_type)
_, unlabeled_docs, explanation_triples, _ = prepped_data
if len(unlabeled_docs) == 0:
unlabeled_docs = None
if len(explanation_triples) == 0:
explanation_triples = None
background_tasks.add_task(train_trigger_soft_match_lean_life,
params.__dict__, unlabeled_docs,
explanation_triples)
async def soft_match_data(api_payload: schema.LeanLifePayload):
"""
Endpoint that converts explanations into strict labeling functions and labels
a pool of unlabeled sentences. Please refer to the docs or `json_schema.py` to
understand the required paramaters.
"""
params = api_payload.params
lean_life_data = api_payload.lean_life_data
prepped_data = util_f.prepare_next_data(lean_life_data,
project_type=params.project_type)
label_space, unlabeled_docs, explanation_triples, ner_label_space = prepped_data
if len(ner_label_space) == 0:
ner_label_space = None
if len(unlabeled_docs) == 0:
unlabeled_docs = None
if len(explanation_triples) == 0:
explanation_triples = None
data = apply_soft_matching(params.__dict__, label_space, unlabeled_docs,
explanation_triples, ner_label_space)
return schema.SoftMatchData(scores=data)