Push model using huggingface_hub.
Browse files- glirel_config.json +32 -0
- pytorch_model.bin +3 -0
glirel_config.json
ADDED
|
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"lr_encoder": "1e-5",
|
| 3 |
+
"lr_others": "1e-4",
|
| 4 |
+
"num_steps": 10000,
|
| 5 |
+
"warmup_ratio": 0.1,
|
| 6 |
+
"train_batch_size": 8,
|
| 7 |
+
"eval_every": 500,
|
| 8 |
+
"gradient_accumulation": null,
|
| 9 |
+
"eval_batch_size": 32,
|
| 10 |
+
"max_width": 12,
|
| 11 |
+
"model_name": "microsoft/deberta-v3-large",
|
| 12 |
+
"fine_tune": true,
|
| 13 |
+
"subtoken_pooling": "first",
|
| 14 |
+
"hidden_size": 768,
|
| 15 |
+
"span_mode": "marker",
|
| 16 |
+
"dropout": 0.4,
|
| 17 |
+
"scheduler": "cosine_with_warmup",
|
| 18 |
+
"loss_func": "binary_cross_entropy_loss",
|
| 19 |
+
"dataset_name": "few_rel",
|
| 20 |
+
"root_dir": "ablation_backbone",
|
| 21 |
+
"train_data": "data/few_rel_all.jsonl",
|
| 22 |
+
"prev_path": "none",
|
| 23 |
+
"size_sup": -1,
|
| 24 |
+
"num_train_rel_types": 25,
|
| 25 |
+
"num_unseen_rel_types": 15,
|
| 26 |
+
"top_k": 1,
|
| 27 |
+
"random_drop": true,
|
| 28 |
+
"max_len": 384,
|
| 29 |
+
"eval_threshold": 0.1,
|
| 30 |
+
"name": "large",
|
| 31 |
+
"log_dir": null
|
| 32 |
+
}
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ba1641e51dcc0123bbd26b4ea8b9946fdf5ad2d30c6c048ccc92dbfb069459e8
|
| 3 |
+
size 1866500606
|