Initial Model Commit

config.json

@@ -0,0 +1 @@
+{"data_dir": "/home/amy/.cache/refined", "transformer_name": "roberta-base", "max_seq": 510, "learning_rate": 5e-05, "num_train_epochs": 1, "freeze_all_bert_layers": false, "gradient_accumulation_steps": 1, "per_gpu_batch_size": 22, "freeze_embedding_layers": false, "freeze_layers": [], "n_gpu": 4, "lr_ner_scale": 100, "ner_layer_dropout": 0.1, "ed_layer_dropout": 0.05, "max_candidates": 30, "warmup_steps": 5000, "logging_steps": 500, "save_steps": 500, "detach_ed_layer": true, "only_ner": false, "only_ed": false, "md_layer_dropout": 0.1, "debug": false, "sep_token_id": 2, "cls_token_id": 0, "mask_token_id": 50264, "pad_token_id": 1, "vocab_size": 50265, "ner_tag_to_ix": {"O": 0, "B-MENTION": 1, "I-MENTION": 2}}

fine_tuning_args.json

@@ -0,0 +1 @@
+{"class_name": "FineTuningArgs", "experiment_name": "wiki_model_llama", "device": "cuda:0", "el": false, "ed_dropout": 0.05, "et_dropout": 0.1, "gradient_accumulation_steps": 4, "epochs": 5, "lr": 1e-05, "batch_size": 32, "ed_threshold": 0.15, "num_warmup_steps": 10, "num_candidates_train": 30, "num_candidates_eval": 30, "use_precomputed_descriptions": false, "output_dir": "../", "restore_model_path": null, "entity_set": "wikipedia", "data_dir": "/home/amy/.cache/refined", "debug": false, "transformer_name": "roberta-base", "n_gpu": 1, "mask_prob": 0.7, "mask_random_prob": 0.05, "candidate_dropout": 0.0, "max_mentions": 25, "download_files": true, "checkpoint_every_n_steps": 1000000, "resume": false, "checkpoint_metric": "ed", "model_name": "wikipedia_model"}

model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42145fe6d9d17d76c12afc310af8ec16cb5e60167073ba016616ca0c17a81396
+size 725973509

optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e5d79e28943583476edfb09a7c8cb93b20a2a287045de2c7cf4f8e963b9a5b3
+size 1430176901

scaler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:43f8a19f895ed6b8e5a0be931211024892121f7a2098233691a39661e1f0cbef
+size 557

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:81430e567af733d864f684f5552266825e3c7ac818e519063d8f56c9eee5b063
+size 627