Introduce code from Rust monorepo. This is a mess

tts/NeMo/examples/nlp/entity_linking/conf/tiny_example_entity_linking_config.yaml

@@ -0,0 +1,90 @@
+project_dir: null
+name: SelfAlignmentPretrainingForMedicalEntityLinking
+trainer:
+  gpus: 1
+  num_nodes: 1
+  max_epochs: 2
+  max_steps: null
+  accumulate_grad_batches: 1
+  precision: 16
+  amp_level: O1
+  accelerator: ddp
+  gradient_clip_val: 0.0
+  log_every_n_steps: 1
+  val_check_interval: 2
+  checkpoint_callback: false
+  logger: false
+model:
+  nemo_path: ???
+  max_seq_length: 128
+  language_model:
+    pretrained_model_name: bert-base-uncased
+    config_file: null
+    config: null
+    lm_checkpoint: null
+  tokenizer:
+    tokenizer_name: ${model.language_model.pretrained_model_name}
+    vocab_file: null
+    tokenizer_model: null
+    do_lower_case: true
+  loss_params: null
+  train_ds:
+    data_file: ???
+    max_seq_length: ${model.max_seq_length}
+    batch_size: 8
+    shuffle: true
+    num_workers: 2
+    pin_memory: false
+    drop_last: false
+  validation_ds:
+    data_file: ??? 
+    max_seq_length: ${model.max_seq_length}
+    batch_size: 8
+    shuffle: false
+    num_workers: 2
+    pin_memory: false
+    drop_last: false
+  optim:
+    name: adam
+    lr: 3.0e-05
+    weight_decay: 0.0
+    sched:
+      name: CosineAnnealing
+      warmup_steps: null
+      warmup_ratio: 0.1
+      min_lr: 0.0
+      last_epoch: -1
+index:
+  dims: 768
+  nlist: 2
+  top_n: 3
+  query_num_factor: 20
+  index_save_name: ???
+  index_batch_size: 10
+  index_ds:
+    name: tiny_example
+    data_file: ???
+    max_seq_length: ${model.max_seq_length}
+    batch_size: 100
+    shuffle: false
+    num_workers: 2
+    pin_memory: false
+    drop_last: false
+  idx_to_id: ${project_dir}/idx_to_id.pkl
+  id_to_string: ${project_dir}/id_to_string.pkl
+  concept_id_save_name: ${project_dir}/tiny_example_concept_ids.pkl
+  embedding_save_name: ${project_dir}/tiny_example_concept_embeddings.hdf5
+  pca_embeddings_save_name: null
+  apply_pca: false
+  pca: null
+exp_manager:
+  exp_dir: .
+  name: ${project_dir}/SelfAlignmentPretrainingTinyExample
+  create_tensorboard_logger: true
+  create_checkpoint_callback: true
+hydra:
+  run:
+    dir: .
+  job_logging:
+    root:
+      handlers: null

tts/NeMo/examples/nlp/entity_linking/conf/umls_medical_entity_linking_config.yaml

@@ -0,0 +1,95 @@
+project_dir: ???
+name: SelfAlignmentPretrainingForMedicalEntityLinking
+trainer:
+  gpus: 1
+  num_nodes: 1
+  max_epochs: 2
+  max_steps: null
+  accumulate_grad_batches: 1
+  precision: 16
+  amp_level: O1
+  accelerator: ddp
+  gradient_clip_val: 0.0
+  log_every_n_steps: 1
+  val_check_interval: 1000
+  checkpoint_callback: false
+  logger: false
+model:
+  nemo_path: ${project_dir}/sap_bert_umls.nemo
+  raw_data: ${project_dir}/data/MRCONSO.RRF
+  max_seq_length: 128
+  language_model:
+    pretrained_model_name: bert-base-uncased
+    config_file: null
+    config: null
+    lm_checkpoint: null
+  tokenizer:
+    tokenizer_name: ${model.language_model.pretrained_model_name}
+    vocab_file: null
+    tokenizer_model: null
+    do_lower_case: true
+  train_ds:
+    data_file: ${project_dir}/data/umls_train_pairs.tsv
+    max_seq_length: ${model.max_seq_length}
+    batch_size: 128
+    shuffle: true
+    num_workers: 2
+    pin_memory: false
+    drop_last: false
+  validation_ds:
+    data_file: ${project_dir}/data/umls_validation_pairs.tsv
+    max_seq_length: ${model.max_seq_length}
+    batch_size: 128
+    shuffle: false
+    num_workers: 2
+    pin_memory: false
+    drop_last: false
+  optim:
+    name: adam
+    lr: 3.0e-05
+    weight_decay: 0.0
+    sched:
+      name: CosineAnnealing
+      warmup_steps: null
+      warmup_ratio: 0.1
+      min_lr: 0.0
+      last_epoch: -1
+index:
+  dims: 256
+  nlist: 300
+  top_n: 5
+  query_num_factor: 20
+  index_save_name: ${project_dir}/medical_entity_linking_index
+  index_batch_size: 1000
+  raw_data: ${model.raw_data}
+  index_ds:
+    name: umls
+    data_file: ${project_dir}/data/umls_index_concepts.tsv
+    max_seq_length: ${model.max_seq_length}
+    batch_size: 128
+    shuffle: false
+    num_workers: 2
+    pin_memory: false
+    drop_last: false
+  idx_to_id: ${project_dir}/data/idx_to_id.pkl
+  id_to_string: ${project_dir}/data/id_to_string.pkl
+  concept_id_save_name: ${project_dir}/data/concept_ids.pkl
+  embedding_save_name: ${project_dir}/data/medical_concept_embeddings.hdf5
+  pca_embeddings_save_name: ${project_dir}/data/medical_concept_reduced_${index.dims}dim_embeddings.hdf5
+  apply_pca: true
+  pca:
+    input_dim: 756
+    output_dim: ${index.dims}
+    sample_fraction: 0.5
+    pca_save_name: ${project_dir}/${index.pca.input_dim}_to_${index.pca.output_dim}_pca_model.pkl
+exp_manager:
+  exp_dir: ${project_dir}/medical_entity_linking_experiments
+  name: sap_bert_umls
+  create_tensorboard_logger: true
+  create_checkpoint_callback: true
+hydra:
+  run:
+    dir: .
+  job_logging:
+    root:
+      handlers: null