Upload folder using huggingface_hub
Browse files- bilstm/large/config.json +94 -0
- bilstm/large/model.onnx +3 -0
- bilstm/large/vocab.json +246 -0
- bilstm/medium/config.json +1 -1
- bilstm/medium/model.onnx +2 -2
bilstm/large/config.json
ADDED
|
@@ -0,0 +1,94 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"seed": 42,
|
| 3 |
+
"data": {
|
| 4 |
+
"train_files": [
|
| 5 |
+
"/kaggle/working/data/processed/processed.txt"
|
| 6 |
+
],
|
| 7 |
+
"val_files": null,
|
| 8 |
+
"test_files": null,
|
| 9 |
+
"batch_size": 256,
|
| 10 |
+
"num_workers": 12,
|
| 11 |
+
"cache_dir": ".cache/",
|
| 12 |
+
"cache_format": "npz",
|
| 13 |
+
"max_length": 512,
|
| 14 |
+
"val_split": 0.01,
|
| 15 |
+
"test_split": 0,
|
| 16 |
+
"diacritic_keep_probs": [
|
| 17 |
+
0.2,
|
| 18 |
+
0.0,
|
| 19 |
+
0.0
|
| 20 |
+
]
|
| 21 |
+
},
|
| 22 |
+
"modeling_config": {
|
| 23 |
+
"architecture": {
|
| 24 |
+
"name": "bilstm",
|
| 25 |
+
"embedding_dim": 128,
|
| 26 |
+
"hidden_dim": 256,
|
| 27 |
+
"num_layers": 3,
|
| 28 |
+
"dropout": 0.3
|
| 29 |
+
},
|
| 30 |
+
"loss": {
|
| 31 |
+
"name": "cross_entropy",
|
| 32 |
+
"label_smoothing": 0.1,
|
| 33 |
+
"ignore_index": 0
|
| 34 |
+
},
|
| 35 |
+
"optimizer": {
|
| 36 |
+
"name": "adamw",
|
| 37 |
+
"lr": 0.0006,
|
| 38 |
+
"weight_decay": 0.01
|
| 39 |
+
},
|
| 40 |
+
"scheduler": {
|
| 41 |
+
"name": "cosine",
|
| 42 |
+
"t_max": 50
|
| 43 |
+
}
|
| 44 |
+
},
|
| 45 |
+
"trainer": {
|
| 46 |
+
"max_epochs": 40,
|
| 47 |
+
"accelerator": "auto",
|
| 48 |
+
"devices": 1,
|
| 49 |
+
"precision": "32",
|
| 50 |
+
"log_every_n_steps": 100,
|
| 51 |
+
"gradient_clip_val": 1.0,
|
| 52 |
+
"loggers": [
|
| 53 |
+
{
|
| 54 |
+
"logger_name": "tensorboard",
|
| 55 |
+
"save_dir": "lightning_logs/"
|
| 56 |
+
},
|
| 57 |
+
{
|
| 58 |
+
"logger_name": "wandb",
|
| 59 |
+
"project": "Arabic-Diacritizer",
|
| 60 |
+
"log_model": "all",
|
| 61 |
+
"name": "main"
|
| 62 |
+
}
|
| 63 |
+
],
|
| 64 |
+
"callbacks": [
|
| 65 |
+
{
|
| 66 |
+
"name": "model_checkpoint",
|
| 67 |
+
"monitor": "val_der",
|
| 68 |
+
"mode": "min",
|
| 69 |
+
"save_top_k": 3,
|
| 70 |
+
"filename": "diacritizer-epoch={epoch:02d}-val-der={val_der:.4f}"
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"name": "early_stopping",
|
| 74 |
+
"monitor": "val_loss",
|
| 75 |
+
"mode": "min",
|
| 76 |
+
"patience": 10
|
| 77 |
+
},
|
| 78 |
+
{
|
| 79 |
+
"name": "lr_monitor",
|
| 80 |
+
"logging_interval": "step"
|
| 81 |
+
}
|
| 82 |
+
]
|
| 83 |
+
},
|
| 84 |
+
"export": {
|
| 85 |
+
"output_dir": "artifacts/",
|
| 86 |
+
"dummy_input_length": 50,
|
| 87 |
+
"onnx_opset_version": 15,
|
| 88 |
+
"use_torch_dynamo": false
|
| 89 |
+
},
|
| 90 |
+
"git": {
|
| 91 |
+
"branch": "main",
|
| 92 |
+
"commit_hash": "7d57f90a318c335a9486f2b4680389a4978aead9"
|
| 93 |
+
}
|
| 94 |
+
}
|
bilstm/large/model.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0e30e02a115e125357f7afb65cfca492295cc72836f0989a589854b8ced8e98f
|
| 3 |
+
size 15864947
|
bilstm/large/vocab.json
ADDED
|
@@ -0,0 +1,246 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"char2id": {
|
| 3 |
+
"<PAD>": 0,
|
| 4 |
+
"<UNK>": 1,
|
| 5 |
+
" ": 2,
|
| 6 |
+
"!": 3,
|
| 7 |
+
"\"": 4,
|
| 8 |
+
"(": 5,
|
| 9 |
+
")": 6,
|
| 10 |
+
",": 7,
|
| 11 |
+
"-": 8,
|
| 12 |
+
".": 9,
|
| 13 |
+
"/": 10,
|
| 14 |
+
":": 11,
|
| 15 |
+
";": 12,
|
| 16 |
+
"[": 13,
|
| 17 |
+
"]": 14,
|
| 18 |
+
"«": 15,
|
| 19 |
+
"»": 16,
|
| 20 |
+
"،": 17,
|
| 21 |
+
"؛": 18,
|
| 22 |
+
"؟": 19,
|
| 23 |
+
"ء": 20,
|
| 24 |
+
"آ": 21,
|
| 25 |
+
"أ": 22,
|
| 26 |
+
"ؤ": 23,
|
| 27 |
+
"إ": 24,
|
| 28 |
+
"ئ": 25,
|
| 29 |
+
"ا": 26,
|
| 30 |
+
"ب": 27,
|
| 31 |
+
"ة": 28,
|
| 32 |
+
"ت": 29,
|
| 33 |
+
"ث": 30,
|
| 34 |
+
"ج": 31,
|
| 35 |
+
"ح": 32,
|
| 36 |
+
"خ": 33,
|
| 37 |
+
"د": 34,
|
| 38 |
+
"ذ": 35,
|
| 39 |
+
"ر": 36,
|
| 40 |
+
"ز": 37,
|
| 41 |
+
"س": 38,
|
| 42 |
+
"ش": 39,
|
| 43 |
+
"ص": 40,
|
| 44 |
+
"ض": 41,
|
| 45 |
+
"ط": 42,
|
| 46 |
+
"ظ": 43,
|
| 47 |
+
"ع": 44,
|
| 48 |
+
"غ": 45,
|
| 49 |
+
"ػ": 46,
|
| 50 |
+
"ؼ": 47,
|
| 51 |
+
"ؽ": 48,
|
| 52 |
+
"ؾ": 49,
|
| 53 |
+
"ؿ": 50,
|
| 54 |
+
"ـ": 51,
|
| 55 |
+
"ف": 52,
|
| 56 |
+
"ق": 53,
|
| 57 |
+
"ك": 54,
|
| 58 |
+
"ل": 55,
|
| 59 |
+
"م": 56,
|
| 60 |
+
"ن": 57,
|
| 61 |
+
"ه": 58,
|
| 62 |
+
"و": 59,
|
| 63 |
+
"ى": 60,
|
| 64 |
+
"ي": 61,
|
| 65 |
+
"ً": 62,
|
| 66 |
+
"ٌ": 63,
|
| 67 |
+
"ٍ": 64,
|
| 68 |
+
"َ": 65,
|
| 69 |
+
"ُ": 66,
|
| 70 |
+
"ِ": 67,
|
| 71 |
+
"ّ": 68,
|
| 72 |
+
"ْ": 69,
|
| 73 |
+
"ٓ": 70,
|
| 74 |
+
"ٰ": 71,
|
| 75 |
+
"پ": 72,
|
| 76 |
+
"چ": 73,
|
| 77 |
+
"ڤ": 74,
|
| 78 |
+
"گ": 75,
|
| 79 |
+
"…": 76
|
| 80 |
+
},
|
| 81 |
+
"id2char": {
|
| 82 |
+
"0": "<PAD>",
|
| 83 |
+
"1": "<UNK>",
|
| 84 |
+
"2": " ",
|
| 85 |
+
"3": "!",
|
| 86 |
+
"4": "\"",
|
| 87 |
+
"5": "(",
|
| 88 |
+
"6": ")",
|
| 89 |
+
"7": ",",
|
| 90 |
+
"8": "-",
|
| 91 |
+
"9": ".",
|
| 92 |
+
"10": "/",
|
| 93 |
+
"11": ":",
|
| 94 |
+
"12": ";",
|
| 95 |
+
"13": "[",
|
| 96 |
+
"14": "]",
|
| 97 |
+
"15": "«",
|
| 98 |
+
"16": "»",
|
| 99 |
+
"17": "،",
|
| 100 |
+
"18": "؛",
|
| 101 |
+
"19": "؟",
|
| 102 |
+
"20": "ء",
|
| 103 |
+
"21": "آ",
|
| 104 |
+
"22": "أ",
|
| 105 |
+
"23": "ؤ",
|
| 106 |
+
"24": "إ",
|
| 107 |
+
"25": "ئ",
|
| 108 |
+
"26": "ا",
|
| 109 |
+
"27": "ب",
|
| 110 |
+
"28": "ة",
|
| 111 |
+
"29": "ت",
|
| 112 |
+
"30": "ث",
|
| 113 |
+
"31": "ج",
|
| 114 |
+
"32": "ح",
|
| 115 |
+
"33": "خ",
|
| 116 |
+
"34": "د",
|
| 117 |
+
"35": "ذ",
|
| 118 |
+
"36": "ر",
|
| 119 |
+
"37": "ز",
|
| 120 |
+
"38": "س",
|
| 121 |
+
"39": "ش",
|
| 122 |
+
"40": "ص",
|
| 123 |
+
"41": "ض",
|
| 124 |
+
"42": "ط",
|
| 125 |
+
"43": "ظ",
|
| 126 |
+
"44": "ع",
|
| 127 |
+
"45": "غ",
|
| 128 |
+
"46": "ػ",
|
| 129 |
+
"47": "ؼ",
|
| 130 |
+
"48": "ؽ",
|
| 131 |
+
"49": "ؾ",
|
| 132 |
+
"50": "ؿ",
|
| 133 |
+
"51": "ـ",
|
| 134 |
+
"52": "ف",
|
| 135 |
+
"53": "ق",
|
| 136 |
+
"54": "ك",
|
| 137 |
+
"55": "ل",
|
| 138 |
+
"56": "م",
|
| 139 |
+
"57": "ن",
|
| 140 |
+
"58": "ه",
|
| 141 |
+
"59": "و",
|
| 142 |
+
"60": "ى",
|
| 143 |
+
"61": "ي",
|
| 144 |
+
"62": "ً",
|
| 145 |
+
"63": "ٌ",
|
| 146 |
+
"64": "ٍ",
|
| 147 |
+
"65": "َ",
|
| 148 |
+
"66": "ُ",
|
| 149 |
+
"67": "ِ",
|
| 150 |
+
"68": "ّ",
|
| 151 |
+
"69": "ْ",
|
| 152 |
+
"70": "ٓ",
|
| 153 |
+
"71": "ٰ",
|
| 154 |
+
"72": "پ",
|
| 155 |
+
"73": "چ",
|
| 156 |
+
"74": "ڤ",
|
| 157 |
+
"75": "گ",
|
| 158 |
+
"76": "…"
|
| 159 |
+
},
|
| 160 |
+
"diacritic2id": {
|
| 161 |
+
"": 0,
|
| 162 |
+
"ً": 1,
|
| 163 |
+
"ٌ": 2,
|
| 164 |
+
"ٍ": 3,
|
| 165 |
+
"َ": 4,
|
| 166 |
+
"ُ": 5,
|
| 167 |
+
"ِ": 6,
|
| 168 |
+
"ًّ": 7,
|
| 169 |
+
"ٌّ": 8,
|
| 170 |
+
"ٍّ": 9,
|
| 171 |
+
"َّ": 10,
|
| 172 |
+
"ُّ": 11,
|
| 173 |
+
"ِّ": 12,
|
| 174 |
+
"ْ": 13,
|
| 175 |
+
"ٓ": 14,
|
| 176 |
+
"ٰ": 15
|
| 177 |
+
},
|
| 178 |
+
"id2diacritic": {
|
| 179 |
+
"0": "",
|
| 180 |
+
"1": "ً",
|
| 181 |
+
"2": "ٌ",
|
| 182 |
+
"3": "ٍ",
|
| 183 |
+
"4": "َ",
|
| 184 |
+
"5": "ُ",
|
| 185 |
+
"6": "ِ",
|
| 186 |
+
"7": "ًّ",
|
| 187 |
+
"8": "ٌّ",
|
| 188 |
+
"9": "ٍّ",
|
| 189 |
+
"10": "َّ",
|
| 190 |
+
"11": "ُّ",
|
| 191 |
+
"12": "ِّ",
|
| 192 |
+
"13": "ْ",
|
| 193 |
+
"14": "ٓ",
|
| 194 |
+
"15": "ٰ"
|
| 195 |
+
},
|
| 196 |
+
"pad_idx": 0,
|
| 197 |
+
"unk_idx": 1,
|
| 198 |
+
"arabic_letters": [
|
| 199 |
+
"ء",
|
| 200 |
+
"آ",
|
| 201 |
+
"أ",
|
| 202 |
+
"ؤ",
|
| 203 |
+
"إ",
|
| 204 |
+
"ئ",
|
| 205 |
+
"ا",
|
| 206 |
+
"ب",
|
| 207 |
+
"ة",
|
| 208 |
+
"ت",
|
| 209 |
+
"ث",
|
| 210 |
+
"ج",
|
| 211 |
+
"ح",
|
| 212 |
+
"خ",
|
| 213 |
+
"د",
|
| 214 |
+
"ذ",
|
| 215 |
+
"ر",
|
| 216 |
+
"ز",
|
| 217 |
+
"س",
|
| 218 |
+
"ش",
|
| 219 |
+
"ص",
|
| 220 |
+
"ض",
|
| 221 |
+
"ط",
|
| 222 |
+
"ظ",
|
| 223 |
+
"ع",
|
| 224 |
+
"غ",
|
| 225 |
+
"ػ",
|
| 226 |
+
"ؼ",
|
| 227 |
+
"ؽ",
|
| 228 |
+
"ؾ",
|
| 229 |
+
"ؿ",
|
| 230 |
+
"ـ",
|
| 231 |
+
"ف",
|
| 232 |
+
"ق",
|
| 233 |
+
"ك",
|
| 234 |
+
"ل",
|
| 235 |
+
"م",
|
| 236 |
+
"ن",
|
| 237 |
+
"ه",
|
| 238 |
+
"و",
|
| 239 |
+
"ى",
|
| 240 |
+
"ي",
|
| 241 |
+
"پ",
|
| 242 |
+
"چ",
|
| 243 |
+
"ڤ",
|
| 244 |
+
"گ"
|
| 245 |
+
]
|
| 246 |
+
}
|
bilstm/medium/config.json
CHANGED
|
@@ -23,7 +23,7 @@
|
|
| 23 |
"architecture": {
|
| 24 |
"name": "bilstm",
|
| 25 |
"embedding_dim": 128,
|
| 26 |
-
"hidden_dim":
|
| 27 |
"num_layers": 3,
|
| 28 |
"dropout": 0.3
|
| 29 |
},
|
|
|
|
| 23 |
"architecture": {
|
| 24 |
"name": "bilstm",
|
| 25 |
"embedding_dim": 128,
|
| 26 |
+
"hidden_dim": 128,
|
| 27 |
"num_layers": 3,
|
| 28 |
"dropout": 0.3
|
| 29 |
},
|
bilstm/medium/model.onnx
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ff3f1b2519f7e244895a92291076d71b97ce198351df3566c7af97789c99f1a4
|
| 3 |
+
size 4289635
|