Upload folder using huggingface_hub
Browse files- gliner_config.json +37 -0
- pytorch_model.bin +3 -0
gliner_config.json
ADDED
|
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_name": "microsoft/mdeberta-v3-base",
|
| 3 |
+
"name": "token level gliner",
|
| 4 |
+
"max_width": 100,
|
| 5 |
+
"hidden_size": 768,
|
| 6 |
+
"dropout": 0.0,
|
| 7 |
+
"fine_tune": true,
|
| 8 |
+
"subtoken_pooling": "first",
|
| 9 |
+
"span_mode": "token_level",
|
| 10 |
+
"num_steps": 1000000,
|
| 11 |
+
"train_batch_size": 4,
|
| 12 |
+
"eval_every": 20000,
|
| 13 |
+
"warmup_ratio": 0.1,
|
| 14 |
+
"scheduler_type": "cosine",
|
| 15 |
+
"gradient_accumulation_steps": 2,
|
| 16 |
+
"loss_alpha": 0.25,
|
| 17 |
+
"loss_gamma": 2,
|
| 18 |
+
"label_smoothing": 0,
|
| 19 |
+
"loss_reduction": "sum",
|
| 20 |
+
"lr_encoder": "3e-5",
|
| 21 |
+
"lr_others": "3e-5",
|
| 22 |
+
"weight_decay_encoder": 0.01,
|
| 23 |
+
"weight_decay_other": 0.01,
|
| 24 |
+
"root_dir": "gliner_logs",
|
| 25 |
+
"train_data": "/home/ltngoc/Move Gliner/data/viner_train.json",
|
| 26 |
+
"val_data_dir": "/home/ltngoc/Move Gliner/data/viner",
|
| 27 |
+
"log_dir": "mdeberta",
|
| 28 |
+
"prev_path": "/home/ltngoc/Move Gliner/mdeberta/model_20000",
|
| 29 |
+
"save_total_limit": 10,
|
| 30 |
+
"size_sup": -1,
|
| 31 |
+
"max_types": 13000,
|
| 32 |
+
"shuffle_types": true,
|
| 33 |
+
"random_drop": true,
|
| 34 |
+
"max_neg_type_ratio": 1,
|
| 35 |
+
"max_len": 768,
|
| 36 |
+
"freeze_token_rep": false
|
| 37 |
+
}
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7d41167da86cf121a1ae83f8a5e1812e6cc61fe17f2b090bace95dd60fb5343a
|
| 3 |
+
size 1162215113
|