def create_evaluate_task_by_train_job_id(train_job_id, evaluate_task_name, evaluate_task_desc, mark_job_ids, doc_term_ids, doc_relation_ids, use_rule=0):
"""
如果后面要加重新训练的逻辑，这部分要改，不能根据train_job_id去创建评估任务，而是根据train_task_id，
目前先保留，因为目前train_job和train_task是一一对应，不会有影响
"""
# get correspondent train_job, doc_type, train_task, nlp_task by train_job_id
train_job = TrainJobModel().get_by_id(train_job_id)
doc_type = DocTypeModel().get_by_id(train_job.doc_type_id)
doc_term_list = DocTermModel().get_by_filter(limit=99999, doc_type_id=doc_type.doc_type_id)
doc_type.doc_term_list = doc_term_list
nlp_task = NlpTaskEnum(doc_type.nlp_task_id)
_, train_task_list = TrainTaskModel().get_by_filter(train_job_id=train_job_id)
train_task = train_task_list[0]
# create evaluate_task
evaluate_task = EvaluateTaskModel().create(evaluate_task_name=evaluate_task_name,
evaluate_task_desc=evaluate_task_desc,
train_task_id=train_task.train_task_id,
evaluate_task_status=int(StatusEnum.processing))
# bulk create evaluate m2m mark
evaluate_m2m_mark_list = [{"evaluate_task_id": evaluate_task.evaluate_task_id, "mark_job_id": _id} for _id in mark_job_ids]
EvaluateM2mMarkModel().bulk_create(evaluate_m2m_mark_list)
# push to evaluate redis queue
doc_term_ids = [str(t.doc_term_id) for t in RelationM2mTermModel().get_by_filter(limit=99999, doc_relation_ids=[int(rl) for rl in doc_relation_ids])]
push_evaluate_task_to_redis(nlp_task, evaluate_task, train_task, doc_type, mark_job_ids, doc_term_ids, doc_relation_ids, use_rule)
session.commit()
return evaluate_task
def get_evaluate_task_list_by_train_job_id(train_job_id, order_by, order_by_desc, offset, limit):
count, evaluate_task_list = EvaluateTaskModel().get_by_train_job_id(train_job_id=train_job_id, order_by=order_by, order_by_desc=order_by_desc, offset=offset, limit=limit)
# assign train_job_id to evaluate_task for dumping
for evaluate_task in evaluate_task_list:
evaluate_task.train_job_id = train_job_id
evaluate_task.mark_job_ids = [m2m.mark_job_id for m2m in EvaluateM2mMarkModel().get_by_filter(limit=99999, evaluate_task_id=evaluate_task.evaluate_task_id)]
return count, evaluate_task_list
def get_doc_type_info_by_nlp_task_by_user(nlp_task_id, current_user: CurrentUser):
"""
获取管理大厅首页的doc_type信息
"""
result = []
# get doc_type list by user
_, doc_type_list = DocTypeModel().get_by_nlp_task_id_by_user(nlp_task_id=nlp_task_id, current_user=current_user)
for doc_type, terms in doc_type_list:
doc_type.doc_terms = [int(t) for t in terms.split(",")] if terms is not None else []
doc_type_list = [d[0] for d in doc_type_list]
doc_type_list = [{"doc_type": DocTypeSchema().dump(doc_type)} for doc_type in doc_type_list]
# get all job count and approved job count
all_status, all_marked_status = MarkTaskModel().count_status_by_user(nlp_task_id=nlp_task_id, current_user=current_user)
# calculate marked mark_job count and all mark_job for each doc_type
all_status_dict = Common().tuple_list2dict(all_status)
all_marked_status_dict = Common().tuple_list2dict(all_marked_status)
for doc_type in doc_type_list:
doc_type_id = doc_type["doc_type"]["doc_type_id"]
mark_job_count = len(all_status_dict.get(doc_type_id, {}))
marked_mark_job_count = 0
for _mark_job_id, _count_sum in all_status_dict.get(doc_type_id, {}).items():
if _count_sum == all_marked_status_dict.get(doc_type_id, {}).get(_mark_job_id, 0):
marked_mark_job_count += 1
doc_type.update(progress_state={"job_num": mark_job_count,
"labeled_job_number": marked_mark_job_count,
"progress_rate": round(marked_mark_job_count / mark_job_count, 2) if mark_job_count > 0 else 0})
# get latest evaluation result if exists
latest_evaluate = EvaluateTaskModel().get_latest_evaluate_by_doc_type_id(nlp_task_id=nlp_task_id,
doc_type_id=doc_type_id)
if latest_evaluate:
doc_type.update(evaluate=EvaluateTaskSchema().dump(latest_evaluate))
result.append(doc_type)
return result
def delete_evaluate_task_by_id(evaluate_task_id):
EvaluateTaskModel().delete(evaluate_task_id)
session.commit()
def update_evaluate_task_by_id(evaluate_task_id, args):
evaluate_task = EvaluateTaskModel().update(evaluate_task_id, **args)
session.commit()
return evaluate_task
def get_evaluate_task_by_id(evaluate_task_id):
evaluate_task = EvaluateTaskModel().get_by_id(evaluate_task_id)
evaluate_task.train_job_id = TrainTaskModel().get_by_id(evaluate_task.train_task_id).train_job_id
return evaluate_task
def get_by_nlp_task_id(nlp_task_id,
search,
current_user: CurrentUser,
order_by="created_time",
order_by_desc=True,
limit=10,
offset=0,
**kwargs):
"""
get (traintask, trainjob, doctype) tuple by nlp_task_id and other filters
"""
# Define allowed filter keys
accept_keys = ["train_job_status", "doc_type_id"]
# Compose query, select 3 tables related to a train job
q = session.query(TrainTask, TrainJob, DocType) \
.outerjoin(TrainJob, TrainTask.train_job_id == TrainJob.train_job_id) \
.outerjoin(DocType, DocType.doc_type_id == TrainJob.doc_type_id) \
.filter(DocType.nlp_task_id == nlp_task_id,
~DocType.is_deleted,
~TrainJob.is_deleted,
~TrainTask.is_deleted)
# auth
if current_user.user_role in [
RoleEnum.manager.value, RoleEnum.guest.value
]:
q = q.filter(DocType.group_id.in_(current_user.user_groups))
# Filter conditions
for key, val in kwargs.items():
if key in accept_keys:
q = q.filter(getattr(TrainJob, key) == val)
if search:
q = q.filter(TrainJob.train_job_name.like(f'%{search}%'))
# Order by key
if order_by_desc:
q = q.order_by(getattr(TrainJob, order_by).desc())
else:
q = q.order_by(getattr(TrainJob, order_by))
train_job_list = []
job_id_list = []
for train_task, train_job, doc_type in q.all():
train_task.mark_job_ids = [
m2m.mark_job_id for m2m in TrainM2mMarkbModel().get_by_filter(
limit=99999, train_job_id=train_task.train_job_id)
]
# assign train_task, doc_type to train_job
if train_task.train_job_id not in job_id_list:
job_id_list.append(train_task.train_job_id)
train_job.doc_type = doc_type
_, model_evaluate_list = EvaluateTaskModel(
).get_by_train_job_id(train_job_id=train_job.train_job_id,
evaluate_task_status=int(
StatusEnum.success))
if model_evaluate_list:
train_job.model_evaluate = model_evaluate_list[0]
train_job.model_version = train_task.model_version
train_job.train_list = [train_task]
train_job_list.append(train_job)
else:
train_job_list[job_id_list.index(
train_task.train_job_id)].train_list.append(train_task)
count = len(train_job_list)
return count, train_job_list[offset:offset + limit]