examples/retrieval.py

import sys

sys.path.append('../')
from evaluation.evaluator import IREvaluator
from evaluation.encoders import Model
from adapter_fusion import AdapterEncoder
from reviewer_matching import ReviewerMatchingEvaluator

# default no control codes
# model = Model(base_checkpoint="allenai/specter")

# default control codes
# model = Model(base_checkpoint="../lightning_logs/full_run/scincl_ctrl/checkpoints/", task_id="[PRX]", use_ctrl_codes=True)


model = Model(base_checkpoint="malteos/scincl", variant="adapters",
              adapters_load_from="../../../phantasm/phantasm_new/lightning_logs/full_run/scincl_adapters/checkpoints/",
              task_id="[PRX]", all_tasks=["[PRX]"])
encoder = AdapterEncoder("malteos/scincl", ["[PRX]"],
                         "../../../phantasm/phantasm_new/lightning_logs/full_run/scincl_adapters/checkpoints/model/adapters")
model.encoder = encoder
model.encoder.cuda()
model.encoder.eval()
evaluator = IREvaluator("feeds_1", ("allenai/scirepeval", "feeds_1"), ("allenai/scirepeval_test", "feeds_1"), model,
                        metrics=("map", "ndcg",))
#
# embeddings = evaluator.generate_embeddings()
#
# evaluator.evaluate(embeddings)

# evaluator = IREvaluator("feeds_1", ("allenai/scirepeval", "feeds_title"), ("allenai/scirepeval_test", "feeds_title"),
#                         model, metrics=("map", "ndcg",))
# evaluator = ReviewerMatchingEvaluator("paper reviewer evaluation", ("allenai/scirepeval", "paper_reviewer_matching"),
#                                       ("allenai/scirepeval_test", "paper_reviewer_matching"),
#                                       ("allenai/scirepeval_test", "reviewers"), model, metrics=("map", "ndcg",))

embeddings = evaluator.generate_embeddings()

evaluator.evaluate(embeddings)