symptoms_debug.yaml

hydra:
  run:
    dir: ${output_dir}/${data.task_name}/${ue.dropout_type}_${ue.dropout_subs}/${data.subsample_perc}/${suffix}/${now:%Y-%m-%d}/${now:%H-%M-%S}-${repeat}
model:
    model_name_or_path: 'google/electra-base-discriminator'
    tokenizer_name: ''
    n_epochs: 4
    already_finetuned: false
    config_dir: '/home/avazhentsev/data/sbermed/experimental/ainesterov/symptom_checker/src/test_config.yaml'
    label_to_id_dir: '/home/avazhentsev/data/tmp/data/top3/l2i_niar_megi.pkl'
ue:
    dropout_type: 'MC'
    inference_prob: 0.1
    committee_size: 10
    dropout_subs: 'last'
    eval_bs: 1000
    use_cache: True
    eval_passes: False
    calibrate: False
    use_ood_sampling: False
    dropout:
        max_n: 100
        max_frac: 0.4
        mask_name: 'dpp'
        dry_run_dataset: 'eval'
        is_reused_mask: False
        mask_name_for_mask: 'rbf'
        committee_size: 20


training:
    seed: 101
    learning_rate: 5e-5
    per_device_eval_batch_size: 1000
    per_device_train_batch_size: 16
    num_train_epochs: 4

data:
    task_name: 'symptoms'
    data_dir: '../workdir/glue_data/symptoms/'
    subsample_perc: 0.
    subsample_perc_test: 0.05
    test_data_dir: '/home/avazhentsev/data/tmp/data/diseasecases_for_msk_body.csv'
    
cache_dir: '../workdir/cache/'
output_dir: '../workdir/results/'
repeat: 0
do_predict: True
seed: 42
suffix: ''
do_train: False
do_eval: True
reset_params: False