Upload ./finetunedXLNetmodel/savedmodel300_12\config.json with huggingface_hub
Browse files
finetunedXLNetmodel/savedmodel300_12//config.json
ADDED
|
@@ -0,0 +1,52 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "./savedmodel/11/",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"XLNetLMHeadModel"
|
| 5 |
+
],
|
| 6 |
+
"attn_type": "bi",
|
| 7 |
+
"bi_data": false,
|
| 8 |
+
"bos_token_id": 1,
|
| 9 |
+
"clamp_len": -1,
|
| 10 |
+
"d_head": 64,
|
| 11 |
+
"d_inner": 3072,
|
| 12 |
+
"d_model": 768,
|
| 13 |
+
"dropout": 0.1,
|
| 14 |
+
"end_n_top": 5,
|
| 15 |
+
"eos_token_id": 2,
|
| 16 |
+
"eval_steps": 10,
|
| 17 |
+
"evaluation_strategy": "steps",
|
| 18 |
+
"ff_activation": "gelu",
|
| 19 |
+
"id2label": {
|
| 20 |
+
"0": "LABEL_0"
|
| 21 |
+
},
|
| 22 |
+
"initializer_range": 0.02,
|
| 23 |
+
"label2id": {
|
| 24 |
+
"LABEL_0": 0
|
| 25 |
+
},
|
| 26 |
+
"layer_norm_eps": 1e-12,
|
| 27 |
+
"load_best_model_at_end": true,
|
| 28 |
+
"mem_len": null,
|
| 29 |
+
"model_type": "xlnet",
|
| 30 |
+
"n_head": 12,
|
| 31 |
+
"n_layer": 12,
|
| 32 |
+
"pad_token_id": 5,
|
| 33 |
+
"reuse_len": null,
|
| 34 |
+
"same_length": false,
|
| 35 |
+
"save_total_limit": 5,
|
| 36 |
+
"start_n_top": 5,
|
| 37 |
+
"summary_activation": "tanh",
|
| 38 |
+
"summary_last_dropout": 0.1,
|
| 39 |
+
"summary_type": "last",
|
| 40 |
+
"summary_use_proj": true,
|
| 41 |
+
"task_specific_params": {
|
| 42 |
+
"text-generation": {
|
| 43 |
+
"do_sample": true,
|
| 44 |
+
"max_length": 250
|
| 45 |
+
}
|
| 46 |
+
},
|
| 47 |
+
"transformers_version": "4.5.1",
|
| 48 |
+
"untie_r": true,
|
| 49 |
+
"use_mems_eval": true,
|
| 50 |
+
"use_mems_train": false,
|
| 51 |
+
"vocab_size": 32000
|
| 52 |
+
}
|