better version
Browse files- gliner_config.json +36 -0
- pytorch_model.bin +3 -0
gliner_config.json
ADDED
|
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_name": "microsoft/deberta-v3-base",
|
| 3 |
+
"name": "token level gliner",
|
| 4 |
+
"max_width": 100,
|
| 5 |
+
"hidden_size": 768,
|
| 6 |
+
"dropout": 0.1,
|
| 7 |
+
"fine_tune": true,
|
| 8 |
+
"subtoken_pooling": "first",
|
| 9 |
+
"span_mode": "token_level",
|
| 10 |
+
"num_steps": 30000,
|
| 11 |
+
"train_batch_size": 8,
|
| 12 |
+
"eval_every": 2500,
|
| 13 |
+
"warmup_ratio": 0.1,
|
| 14 |
+
"scheduler_type": "cosine",
|
| 15 |
+
"loss_alpha": -1,
|
| 16 |
+
"loss_gamma": 0,
|
| 17 |
+
"loss_reduction": "sum",
|
| 18 |
+
"lr_encoder": "1e-6",
|
| 19 |
+
"lr_others": "1e-5",
|
| 20 |
+
"weight_decay_encoder": 0.01,
|
| 21 |
+
"weight_decay_other": 0.01,
|
| 22 |
+
"root_dir": "token_gliner_instruct_logs",
|
| 23 |
+
"train_data": "/gpfswork/rech/bwq/upa43yu/LLMexp/instruct_1k.json",
|
| 24 |
+
"val_data_dir": "/gpfswork/rech/ohy/upa43yu/NER_datasets",
|
| 25 |
+
"prev_path": "/gpfswork/rech/bwq/upa43yu/gliner-pip/train_gliner/token_gliner_ft_logs/deberta-v3-base_2/model_30000",
|
| 26 |
+
"size_sup": -1,
|
| 27 |
+
"max_types": 10,
|
| 28 |
+
"shuffle_types": true,
|
| 29 |
+
"random_drop": true,
|
| 30 |
+
"max_neg_type_ratio": 1,
|
| 31 |
+
"max_len": 384,
|
| 32 |
+
"freeze_token_rep": false,
|
| 33 |
+
"gpu": "pds@a100",
|
| 34 |
+
"run_time": "3:30:00",
|
| 35 |
+
"log_dir": "/gpfsdswork/projects/rech/bwq/upa43yu/gliner-pip/train_gliner/token_gliner_instruct_logs/deberta-v3-base_0"
|
| 36 |
+
}
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3bb29e95c04700d1db8974a2c37dc19ec391d9d03da500daf0df8a8db6ba7609
|
| 3 |
+
size 787119650
|