Push model using huggingface_hub.
Browse files- glirel_config.json +69 -0
- pytorch_model.bin +3 -0
glirel_config.json
ADDED
|
@@ -0,0 +1,69 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"lr_encoder": "1e-5",
|
| 3 |
+
"lr_others": "1e-4",
|
| 4 |
+
"weight_decay_encoder": 0.01,
|
| 5 |
+
"weight_decay_other": 0.01,
|
| 6 |
+
"num_steps": 100000,
|
| 7 |
+
"warmup_ratio": 0.1,
|
| 8 |
+
"train_batch_size": 1,
|
| 9 |
+
"eval_every": 3000,
|
| 10 |
+
"gradient_accumulation": 8,
|
| 11 |
+
"eval_batch_size": 32,
|
| 12 |
+
"num_layers_freeze": null,
|
| 13 |
+
"early_stopping_patience": null,
|
| 14 |
+
"early_stopping_delta": 0.0,
|
| 15 |
+
"save_at": [
|
| 16 |
+
200,
|
| 17 |
+
6000,
|
| 18 |
+
12000,
|
| 19 |
+
20000,
|
| 20 |
+
70000
|
| 21 |
+
],
|
| 22 |
+
"max_saves": 6,
|
| 23 |
+
"max_width": 12,
|
| 24 |
+
"model_name": "microsoft/deberta-v3-large",
|
| 25 |
+
"fine_tune": true,
|
| 26 |
+
"subtoken_pooling": "first",
|
| 27 |
+
"hidden_size": 768,
|
| 28 |
+
"scorer": "dot",
|
| 29 |
+
"rel_mode": "marker",
|
| 30 |
+
"span_marker_mode": "markerv1",
|
| 31 |
+
"refine_prompt": false,
|
| 32 |
+
"refine_relation": false,
|
| 33 |
+
"ffn_mul": 4,
|
| 34 |
+
"dropout": 0.4,
|
| 35 |
+
"scheduler": "cosine_with_warmup",
|
| 36 |
+
"loss_func": "binary_cross_entropy_loss",
|
| 37 |
+
"alpha": 0.6,
|
| 38 |
+
"gamma": 3,
|
| 39 |
+
"label_embed_strategy": "both",
|
| 40 |
+
"coref_classifier": false,
|
| 41 |
+
"coref_loss_weight": 10.0,
|
| 42 |
+
"dataset_name": "zero_rel",
|
| 43 |
+
"root_dir": "ablation_backbone",
|
| 44 |
+
"train_data": [
|
| 45 |
+
"data/zero_rel_all.jsonl"
|
| 46 |
+
],
|
| 47 |
+
"prev_path": "none",
|
| 48 |
+
"size_sup": -1,
|
| 49 |
+
"num_train_rel_types": 25,
|
| 50 |
+
"num_unseen_rel_types": 15,
|
| 51 |
+
"top_k": 1,
|
| 52 |
+
"random_drop": true,
|
| 53 |
+
"max_len": 512,
|
| 54 |
+
"eval_threshold": [
|
| 55 |
+
0.01,
|
| 56 |
+
0.1,
|
| 57 |
+
0.2,
|
| 58 |
+
0.3,
|
| 59 |
+
0.5
|
| 60 |
+
],
|
| 61 |
+
"max_entity_pair_distance": null,
|
| 62 |
+
"fixed_relation_types": true,
|
| 63 |
+
"name": "large",
|
| 64 |
+
"log_dir": "logs/zero_rel/zero_rel-2024-11-02__10-23-22",
|
| 65 |
+
"eval_data": "data/wiki_zsl_all.jsonl",
|
| 66 |
+
"coreference_label": "SELF",
|
| 67 |
+
"entity_start_token": "[E]",
|
| 68 |
+
"entity_end_token": "[/E]"
|
| 69 |
+
}
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:840aa21bb68a676ebe17451aa2ce8a7e5b987af8db1c0e28850bf0e395b713de
|
| 3 |
+
size 1866504702
|