Upload config.json with huggingface_hub
Browse files- config.json +87 -0
config.json
ADDED
|
@@ -0,0 +1,87 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_num_labels": 3,
|
| 3 |
+
"activation_dropout": 0.0,
|
| 4 |
+
"activation_function": "gelu",
|
| 5 |
+
"add_final_layer_norm": false,
|
| 6 |
+
"architectures": [
|
| 7 |
+
"BartForSequenceClassification"
|
| 8 |
+
],
|
| 9 |
+
"attention_dropout": 0.0,
|
| 10 |
+
"bos_token_id": 0,
|
| 11 |
+
"classif_dropout": 0.0,
|
| 12 |
+
"classifier_dropout": 0.0,
|
| 13 |
+
"d_model": 1024,
|
| 14 |
+
"decoder_attention_heads": 16,
|
| 15 |
+
"decoder_ffn_dim": 4096,
|
| 16 |
+
"decoder_layerdrop": 0.0,
|
| 17 |
+
"decoder_layers": 12,
|
| 18 |
+
"decoder_start_token_id": 2,
|
| 19 |
+
"dropout": 0.1,
|
| 20 |
+
"encoder_attention_heads": 16,
|
| 21 |
+
"encoder_ffn_dim": 4096,
|
| 22 |
+
"encoder_layerdrop": 0.0,
|
| 23 |
+
"encoder_layers": 12,
|
| 24 |
+
"eos_token_id": 2,
|
| 25 |
+
"forced_eos_token_id": 2,
|
| 26 |
+
"gradient_checkpointing": false,
|
| 27 |
+
"id2label": {
|
| 28 |
+
"0": "Simple Problem solving",
|
| 29 |
+
"1": "Simple Writing",
|
| 30 |
+
"2": "Simple Coding",
|
| 31 |
+
"3": "Complex Text Summarization",
|
| 32 |
+
"4": "Complex Coding",
|
| 33 |
+
"5": "Simple Classification",
|
| 34 |
+
"6": "Simple Logical Reasoning",
|
| 35 |
+
"7": "Simple Open ended QA",
|
| 36 |
+
"8": "Complex Writing",
|
| 37 |
+
"9": "Complex Classification",
|
| 38 |
+
"10": "Simple Roleplaying",
|
| 39 |
+
"11": "Complex Roleplaying",
|
| 40 |
+
"12": "Complex Logical Reasoning",
|
| 41 |
+
"13": "Complex Data manipulation",
|
| 42 |
+
"14": "Simple Language Translation",
|
| 43 |
+
"15": "Complex Open ended QA",
|
| 44 |
+
"16": "Simple Text Summarization",
|
| 45 |
+
"17": "Complex Language Translation",
|
| 46 |
+
"18": "Simple Data manipulation",
|
| 47 |
+
"19": "Complex Problem solving",
|
| 48 |
+
"20": "complex Classification"
|
| 49 |
+
},
|
| 50 |
+
"init_std": 0.02,
|
| 51 |
+
"is_encoder_decoder": true,
|
| 52 |
+
"label2id": {
|
| 53 |
+
"Complex Classification": 9,
|
| 54 |
+
"Complex Coding": 4,
|
| 55 |
+
"Complex Data manipulation": 13,
|
| 56 |
+
"Complex Language Translation": 17,
|
| 57 |
+
"Complex Logical Reasoning": 12,
|
| 58 |
+
"Complex Open ended QA": 15,
|
| 59 |
+
"Complex Problem solving": 19,
|
| 60 |
+
"Complex Roleplaying": 11,
|
| 61 |
+
"Complex Text Summarization": 3,
|
| 62 |
+
"Complex Writing": 8,
|
| 63 |
+
"Simple Classification": 5,
|
| 64 |
+
"Simple Coding": 2,
|
| 65 |
+
"Simple Data manipulation": 18,
|
| 66 |
+
"Simple Language Translation": 14,
|
| 67 |
+
"Simple Logical Reasoning": 6,
|
| 68 |
+
"Simple Open ended QA": 7,
|
| 69 |
+
"Simple Problem solving": 0,
|
| 70 |
+
"Simple Roleplaying": 10,
|
| 71 |
+
"Simple Text Summarization": 16,
|
| 72 |
+
"Simple Writing": 1,
|
| 73 |
+
"complex Classification": 20
|
| 74 |
+
},
|
| 75 |
+
"max_position_embeddings": 1024,
|
| 76 |
+
"model_type": "bart",
|
| 77 |
+
"normalize_before": false,
|
| 78 |
+
"num_hidden_layers": 12,
|
| 79 |
+
"output_past": false,
|
| 80 |
+
"pad_token_id": 1,
|
| 81 |
+
"problem_type": "single_label_classification",
|
| 82 |
+
"scale_embedding": false,
|
| 83 |
+
"torch_dtype": "float32",
|
| 84 |
+
"transformers_version": "4.51.3",
|
| 85 |
+
"use_cache": true,
|
| 86 |
+
"vocab_size": 50265
|
| 87 |
+
}
|