def rank2records_in_batch():
interval = 10
start = 20
end = 150 + interval
filter_var_range = range(start, end, interval)
for filter_var in tqdm(filter_var_range):
qa_rec_dn = ensemble_config.QA_RECORD_DIR_NAME_PATTERN.format(ensemble_config.MODEL_NAME,
filter_var,
ensemble_config.FILTER)
qa_rec_dp = join(path_parser.summary_rank, qa_rec_dn)
if exists(qa_rec_dp):
raise ValueError('qa_rec_dp exists: {}'.format(qa_rec_dp))
os.mkdir(qa_rec_dp)
for cid in cids:
retrieval_params = {
'model_name': ensemble_config.MODEL_NAME,
'cid': cid,
'filter_var': filter_var,
'filter': ensemble_config.FILTER,
'deduplicate': None,
}
retrieved_items = ir_tools.retrieve(**retrieval_params)
ir_tools.dump_retrieval(fp=join(qa_rec_dp, cid), retrieved_items=retrieved_items)
def qa_rank2records_in_batch():
if qa_config.FILTER == 'conf':
filter_var_range = np.arange(0.05, 1.05, 0.05)
else: # topK
interval = 10
if qa_config.ir_config.FILTER == 'topK':
start = interval
end = qa_config.ir_config.FILTER_VAR + interval
else:
start = 40
end = 150 + interval
filter_var_range = range(start, end, interval)
for filter_var in tqdm(filter_var_range):
qa_rec_dn = qa_config.QA_RECORD_DIR_NAME_PATTERN.format(
qa_config.QA_MODEL_NAME_BERT, filter_var, qa_config.FILTER)
qa_rec_dp = join(path_parser.summary_rank, qa_rec_dn)
if exists(qa_rec_dp):
raise ValueError('qa_rec_dp exists: {}'.format(qa_rec_dp))
os.mkdir(qa_rec_dp)
for cid in cids:
retrieval_params = {
'model_name': qa_config.QA_MODEL_NAME_BERT,
'cid': cid,
'filter_var': filter_var,
'filter': qa_config.FILTER,
'deduplicate': None,
}
retrieved_items = ir_tools.retrieve(**retrieval_params)
ir_tools.dump_retrieval(fp=join(qa_rec_dp, cid),
retrieved_items=retrieved_items)
def rank2records():
rec_dp = join(path_parser.summary_rank, ensemble_config.QA_RECORD_DIR_NAME)
if exists(rec_dp):
raise ValueError('rec_dp exists: {}'.format(rec_dp))
os.mkdir(rec_dp)
for cid in tqdm(cids):
retrieval_params = {
'model_name': ensemble_config.MODEL_NAME,
'cid': cid,
'filter_var': ensemble_config.FILTER_VAR,
'filter': ensemble_config.FILTER,
'deduplicate': None,
}
retrieved_items = ir_tools.retrieve(**retrieval_params)
ir_tools.dump_retrieval(fp=join(rec_dp, cid), retrieved_items=retrieved_items)
def ir_rank2records():
ir_rec_dp = join(path_parser.summary_rank,
ir_config.IR_RECORDS_DIR_NAME_TF)
if exists(ir_rec_dp):
raise ValueError('qa_rec_dp exists: {}'.format(ir_rec_dp))
os.mkdir(ir_rec_dp)
for cid in tqdm(cids):
retrieval_params = {
'model_name': ir_config.IR_MODEL_NAME_TF,
'cid': cid,
'filter_var': ir_config.FILTER_VAR,
'filter': ir_config.FILTER,
'deduplicate': ir_config.DEDUPLICATE,
'prune': True,
}
retrieved_items = ir_tools.retrieve(**retrieval_params)
ir_tools.dump_retrieval(fp=join(ir_rec_dp, cid),
retrieved_items=retrieved_items)
def ir_rank2records():
ir_rec_dp = join(path_parser.summary_rank,
ir_config.IR_RECORDS_DIR_NAME_TF)
assert not exists(ir_rec_dp), f'ir_rec_dp exists: {ir_rec_dp}'
os.mkdir(ir_rec_dp)
# cids = tools.get_test_cc_ids()
cids = [c_q_dict['cid'] for c_q_dict in test_cid_query_dicts]
for cid in tqdm(cids):
retrieval_params = {
'model_name': ir_config.IR_MODEL_NAME_TF,
'cid': cid,
'filter_var': ir_config.FILTER_VAR,
'filter': ir_config.FILTER,
'deduplicate': ir_config.DEDUPLICATE,
# 'prune': True,
'prune': False,
}
retrieved_items = ir_tools.retrieve(**retrieval_params)
ir_tools.dump_retrieval(fp=join(ir_rec_dp, cid),
retrieved_items=retrieved_items)
