Upload config.json
Browse files- config.json +242 -0
config.json
ADDED
|
@@ -0,0 +1,242 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model": "DAIR-Group/HTR-ConvText",
|
| 3 |
+
"generated_at": "2025-12-16T09:12:23.918807",
|
| 4 |
+
"artifacts": [
|
| 5 |
+
{
|
| 6 |
+
"name": "iam.pth",
|
| 7 |
+
"path": "checkpoints\\iam.pth",
|
| 8 |
+
"exists": true,
|
| 9 |
+
"size_bytes": 529444565,
|
| 10 |
+
"size_mb": 504.918,
|
| 11 |
+
"sha256": "9b2aba60e18d922a615562e23f4079cef4ba4bd559a38d1d6e53de8d7495c122",
|
| 12 |
+
"modified": "2025-12-16T07:24:08.462236",
|
| 13 |
+
"created": "2025-12-16T08:20:24.508913",
|
| 14 |
+
"dataset": "iam",
|
| 15 |
+
"best_cer": 0.02582327045571924,
|
| 16 |
+
"best_wer": 0.08769689229459345,
|
| 17 |
+
"nb_iter": 54000,
|
| 18 |
+
"args": {
|
| 19 |
+
"out_dir": "./output",
|
| 20 |
+
"exp_name": "iam",
|
| 21 |
+
"seed": 123,
|
| 22 |
+
"use_wandb": true,
|
| 23 |
+
"wandb_project": "iam",
|
| 24 |
+
"print_iter": 100,
|
| 25 |
+
"eval_iter": 1000,
|
| 26 |
+
"dataset": "iam",
|
| 27 |
+
"data_path": "/root/ws/dataset/iam/lines/",
|
| 28 |
+
"train_data_list": "/root/ws/dataset/iam/train.ln",
|
| 29 |
+
"val_data_list": "/root/ws/dataset/iam/val.ln",
|
| 30 |
+
"test_data_list": "/root/ws/dataset/iam/test.ln",
|
| 31 |
+
"nb_cls": 80,
|
| 32 |
+
"num_workers": 4,
|
| 33 |
+
"img_size": [
|
| 34 |
+
512,
|
| 35 |
+
64
|
| 36 |
+
],
|
| 37 |
+
"patch_size": [
|
| 38 |
+
4,
|
| 39 |
+
32
|
| 40 |
+
],
|
| 41 |
+
"train_bs": 128,
|
| 42 |
+
"accum_steps": 1,
|
| 43 |
+
"val_bs": 16,
|
| 44 |
+
"total_iter": 100001,
|
| 45 |
+
"warm_up_iter": 1000,
|
| 46 |
+
"max_lr": 0.001,
|
| 47 |
+
"weight_decay": 0.05,
|
| 48 |
+
"ema_decay": 0.9999,
|
| 49 |
+
"alpha": 1.0,
|
| 50 |
+
"model_type": "ctc",
|
| 51 |
+
"cos_temp": 8,
|
| 52 |
+
"proj": 8.0,
|
| 53 |
+
"attn_mask_ratio": 0.1,
|
| 54 |
+
"use_masking": false,
|
| 55 |
+
"mask_ratio": 0.4,
|
| 56 |
+
"max_span_length": 8,
|
| 57 |
+
"spacing": 0,
|
| 58 |
+
"r_rand": 0.6,
|
| 59 |
+
"r_block": 0.6,
|
| 60 |
+
"block_span": 4,
|
| 61 |
+
"r_span": 0.4,
|
| 62 |
+
"max_span": 8,
|
| 63 |
+
"dpi_min_factor": 0.5,
|
| 64 |
+
"dpi_max_factor": 1.5,
|
| 65 |
+
"perspective_low": 0.0,
|
| 66 |
+
"perspective_high": 0.4,
|
| 67 |
+
"elastic_distortion_min_kernel_size": 3,
|
| 68 |
+
"elastic_distortion_max_kernel_size": 3,
|
| 69 |
+
"elastic_distortion_max_magnitude": 20,
|
| 70 |
+
"elastic_distortion_min_alpha": 0.5,
|
| 71 |
+
"elastic_distortion_max_alpha": 1,
|
| 72 |
+
"elastic_distortion_min_sigma": 1,
|
| 73 |
+
"elastic_distortion_max_sigma": 10,
|
| 74 |
+
"dila_ero_max_kernel": 2,
|
| 75 |
+
"dila_ero_iter": 1,
|
| 76 |
+
"jitter_contrast": 0.4,
|
| 77 |
+
"jitter_brightness": 0.4,
|
| 78 |
+
"jitter_saturation": 0.4,
|
| 79 |
+
"jitter_hue": 0.2,
|
| 80 |
+
"blur_min_kernel": 3,
|
| 81 |
+
"blur_max_kernel": 5,
|
| 82 |
+
"blur_min_sigma": 3,
|
| 83 |
+
"blur_max_sigma": 5,
|
| 84 |
+
"sharpen_min_alpha": 0,
|
| 85 |
+
"sharpen_max_alpha": 1,
|
| 86 |
+
"sharpen_min_strength": 0,
|
| 87 |
+
"sharpen_max_strength": 1,
|
| 88 |
+
"zoom_min_h": 0.8,
|
| 89 |
+
"zoom_max_h": 1,
|
| 90 |
+
"zoom_min_w": 0.99,
|
| 91 |
+
"zoom_max_w": 1,
|
| 92 |
+
"proba": 0.5,
|
| 93 |
+
"decoder_layers": 6,
|
| 94 |
+
"decoder_heads": 8,
|
| 95 |
+
"max_seq_len": 256,
|
| 96 |
+
"label_smoothing": 0.1,
|
| 97 |
+
"beam_size": 5,
|
| 98 |
+
"generation_method": "nucleus",
|
| 99 |
+
"generation_temperature": 0.7,
|
| 100 |
+
"repetition_penalty": 1.3,
|
| 101 |
+
"top_p": 0.9,
|
| 102 |
+
"tcm_enable": true,
|
| 103 |
+
"tcm_lambda": 1.0,
|
| 104 |
+
"ctc_lambda": 0.1,
|
| 105 |
+
"tcm_sub_len": 5,
|
| 106 |
+
"tcm_warmup_iters": 0,
|
| 107 |
+
"resume": null,
|
| 108 |
+
"load_model": null,
|
| 109 |
+
"load_encoder_only": false,
|
| 110 |
+
"strict_loading": true,
|
| 111 |
+
"save_dir": "./output/iam"
|
| 112 |
+
},
|
| 113 |
+
"load_error": null
|
| 114 |
+
},
|
| 115 |
+
{
|
| 116 |
+
"name": "lam.pth",
|
| 117 |
+
"path": "checkpoints\\lam.pth",
|
| 118 |
+
"exists": false,
|
| 119 |
+
"size_bytes": null,
|
| 120 |
+
"size_mb": null,
|
| 121 |
+
"sha256": null,
|
| 122 |
+
"modified": null,
|
| 123 |
+
"created": null,
|
| 124 |
+
"dataset": null,
|
| 125 |
+
"best_cer": null,
|
| 126 |
+
"best_wer": null,
|
| 127 |
+
"nb_iter": null,
|
| 128 |
+
"args": {},
|
| 129 |
+
"load_error": null
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"name": "read.pth",
|
| 133 |
+
"path": "checkpoints\\read.pth",
|
| 134 |
+
"exists": true,
|
| 135 |
+
"size_bytes": 529516245,
|
| 136 |
+
"size_mb": 504.986,
|
| 137 |
+
"sha256": "01d3552f42356eecb1625dfc8692306f49efab1c7a3e116d819c9cc42d317d1f",
|
| 138 |
+
"modified": "2025-12-16T07:24:15.937379",
|
| 139 |
+
"created": "2025-12-16T08:20:24.959519",
|
| 140 |
+
"dataset": "read",
|
| 141 |
+
"best_cer": 0.039916476472587774,
|
| 142 |
+
"best_wer": 0.17772215269086358,
|
| 143 |
+
"nb_iter": 40000,
|
| 144 |
+
"args": {
|
| 145 |
+
"out_dir": "./output",
|
| 146 |
+
"exp_name": "read",
|
| 147 |
+
"seed": 123,
|
| 148 |
+
"use_wandb": true,
|
| 149 |
+
"wandb_project": "read2016",
|
| 150 |
+
"print_iter": 100,
|
| 151 |
+
"eval_iter": 1000,
|
| 152 |
+
"dataset": "read2016",
|
| 153 |
+
"data_path": "/root/ws/dataset/read/lines/",
|
| 154 |
+
"train_data_list": "/root/ws/dataset/read/train.ln",
|
| 155 |
+
"val_data_list": "/root/ws/dataset/read/val.ln",
|
| 156 |
+
"test_data_list": "/root/ws/dataset/read/test.ln",
|
| 157 |
+
"nb_cls": 90,
|
| 158 |
+
"num_workers": 4,
|
| 159 |
+
"img_size": [
|
| 160 |
+
512,
|
| 161 |
+
64
|
| 162 |
+
],
|
| 163 |
+
"patch_size": [
|
| 164 |
+
4,
|
| 165 |
+
32
|
| 166 |
+
],
|
| 167 |
+
"train_bs": 128,
|
| 168 |
+
"accum_steps": 1,
|
| 169 |
+
"val_bs": 16,
|
| 170 |
+
"total_iter": 100001,
|
| 171 |
+
"warm_up_iter": 1000,
|
| 172 |
+
"max_lr": 0.001,
|
| 173 |
+
"weight_decay": 0.05,
|
| 174 |
+
"ema_decay": 0.9999,
|
| 175 |
+
"alpha": 1.0,
|
| 176 |
+
"model_type": "ctc",
|
| 177 |
+
"cos_temp": 8,
|
| 178 |
+
"proj": 8.0,
|
| 179 |
+
"attn_mask_ratio": 0.1,
|
| 180 |
+
"use_masking": false,
|
| 181 |
+
"mask_ratio": 0.4,
|
| 182 |
+
"max_span_length": 8,
|
| 183 |
+
"spacing": 0,
|
| 184 |
+
"r_rand": 0.6,
|
| 185 |
+
"r_block": 0.6,
|
| 186 |
+
"block_span": 4,
|
| 187 |
+
"r_span": 0.4,
|
| 188 |
+
"max_span": 8,
|
| 189 |
+
"dpi_min_factor": 0.5,
|
| 190 |
+
"dpi_max_factor": 1.5,
|
| 191 |
+
"perspective_low": 0.0,
|
| 192 |
+
"perspective_high": 0.4,
|
| 193 |
+
"elastic_distortion_min_kernel_size": 3,
|
| 194 |
+
"elastic_distortion_max_kernel_size": 3,
|
| 195 |
+
"elastic_distortion_max_magnitude": 20,
|
| 196 |
+
"elastic_distortion_min_alpha": 0.5,
|
| 197 |
+
"elastic_distortion_max_alpha": 1,
|
| 198 |
+
"elastic_distortion_min_sigma": 1,
|
| 199 |
+
"elastic_distortion_max_sigma": 10,
|
| 200 |
+
"dila_ero_max_kernel": 2,
|
| 201 |
+
"dila_ero_iter": 1,
|
| 202 |
+
"jitter_contrast": 0.4,
|
| 203 |
+
"jitter_brightness": 0.4,
|
| 204 |
+
"jitter_saturation": 0.4,
|
| 205 |
+
"jitter_hue": 0.2,
|
| 206 |
+
"blur_min_kernel": 3,
|
| 207 |
+
"blur_max_kernel": 5,
|
| 208 |
+
"blur_min_sigma": 3,
|
| 209 |
+
"blur_max_sigma": 5,
|
| 210 |
+
"sharpen_min_alpha": 0,
|
| 211 |
+
"sharpen_max_alpha": 1,
|
| 212 |
+
"sharpen_min_strength": 0,
|
| 213 |
+
"sharpen_max_strength": 1,
|
| 214 |
+
"zoom_min_h": 0.8,
|
| 215 |
+
"zoom_max_h": 1,
|
| 216 |
+
"zoom_min_w": 0.99,
|
| 217 |
+
"zoom_max_w": 1,
|
| 218 |
+
"proba": 0.5,
|
| 219 |
+
"decoder_layers": 6,
|
| 220 |
+
"decoder_heads": 8,
|
| 221 |
+
"max_seq_len": 256,
|
| 222 |
+
"label_smoothing": 0.1,
|
| 223 |
+
"beam_size": 5,
|
| 224 |
+
"generation_method": "nucleus",
|
| 225 |
+
"generation_temperature": 0.7,
|
| 226 |
+
"repetition_penalty": 1.3,
|
| 227 |
+
"top_p": 0.9,
|
| 228 |
+
"tcm_enable": true,
|
| 229 |
+
"tcm_lambda": 1.0,
|
| 230 |
+
"ctc_lambda": 0.1,
|
| 231 |
+
"tcm_sub_len": 5,
|
| 232 |
+
"tcm_warmup_iters": 0,
|
| 233 |
+
"resume": null,
|
| 234 |
+
"load_model": null,
|
| 235 |
+
"load_encoder_only": false,
|
| 236 |
+
"strict_loading": true,
|
| 237 |
+
"save_dir": "./output/read"
|
| 238 |
+
},
|
| 239 |
+
"load_error": null
|
| 240 |
+
}
|
| 241 |
+
]
|
| 242 |
+
}
|