# path to pretrained model and tokenizer
model_name_or_path: microsoft/BiomedNLP-KRISSBERT-PubMed-UMLS-EL

test_data:
  _target_: utils.MedMentionsDataset
  dataset_path: MedMentions/full/data/
  split: test

# paths to encoded data
encoded_files: [
  prototypes/embeddings
]

encoded_umls_files: []

entity_list_ids:

entity_list_names: prototypes/name_cuis

index_path:

seed: 12345
batch_size: 256
max_length: 64
num_retrievals: 100
top_ks: [1, 5, 50, 100]