k0ry commited on
Commit
47e7021
·
1 Parent(s): 2aecb2d

Update model configuration and replace artifacts;

Browse files
vn.pth → checkpoints/hands-vnondb.pth RENAMED
File without changes
iam.pth → checkpoints/iam.pth RENAMED
File without changes
read.pth → checkpoints/read.pth RENAMED
File without changes
config.json CHANGED
@@ -1,6 +1,6 @@
1
  {
2
  "model": "DAIR-Group/HTR-ConvText",
3
- "generated_at": "2025-12-16T09:12:23.918807",
4
  "artifacts": [
5
  {
6
  "name": "iam.pth",
@@ -9,8 +9,8 @@
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,
@@ -113,19 +113,113 @@
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
  {
@@ -135,8 +229,8 @@
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,
 
1
  {
2
  "model": "DAIR-Group/HTR-ConvText",
3
+ "generated_at": "2025-12-16T09:19:08.351480",
4
  "artifacts": [
5
  {
6
  "name": "iam.pth",
 
9
  "size_bytes": 529444565,
10
  "size_mb": 504.918,
11
  "sha256": "9b2aba60e18d922a615562e23f4079cef4ba4bd559a38d1d6e53de8d7495c122",
12
+ "modified": "2025-12-16T09:16:22.604113",
13
+ "created": "2025-12-16T09:16:22.379708",
14
  "dataset": "iam",
15
  "best_cer": 0.02582327045571924,
16
  "best_wer": 0.08769689229459345,
 
113
  "load_error": null
114
  },
115
  {
116
+ "name": "hands-vnondb.pth",
117
+ "path": "checkpoints\\hands-vnondb.pth",
118
+ "exists": true,
119
+ "size_bytes": 530033301,
120
+ "size_mb": 505.479,
121
+ "sha256": "7208c41f3e58cf02bc228e29aa669836393bb1c08ae33a077e42dfa119645a61",
122
+ "modified": "2025-12-16T09:16:19.392341",
123
+ "created": "2025-12-16T09:16:19.160248",
124
+ "dataset": "hands-vnondb",
125
+ "best_cer": 0.035116817411641174,
126
+ "best_wer": 0.0871608236974248,
127
+ "nb_iter": 65000,
128
+ "args": {
129
+ "out_dir": "./output",
130
+ "exp_name": "vnondb",
131
+ "seed": 123,
132
+ "use_wandb": true,
133
+ "wandb_project": "vnondb",
134
+ "print_iter": 100,
135
+ "eval_iter": 1000,
136
+ "dataset": "vnondb",
137
+ "data_path": "/root/ws/dataset/vnondb/lines/",
138
+ "train_data_list": "/root/ws/dataset/vnondb/train.ln",
139
+ "val_data_list": "/root/ws/dataset/vnondb/valid.ln",
140
+ "test_data_list": "/root/ws/dataset/vnondb/test.ln",
141
+ "nb_cls": 162,
142
+ "num_workers": 4,
143
+ "img_size": [
144
+ 512,
145
+ 64
146
+ ],
147
+ "patch_size": [
148
+ 4,
149
+ 32
150
+ ],
151
+ "train_bs": 128,
152
+ "accum_steps": 1,
153
+ "val_bs": 16,
154
+ "total_iter": 100001,
155
+ "warm_up_iter": 1000,
156
+ "max_lr": 0.001,
157
+ "weight_decay": 0.05,
158
+ "ema_decay": 0.9999,
159
+ "alpha": 1.0,
160
+ "model_type": "ctc",
161
+ "cos_temp": 8,
162
+ "proj": 8.0,
163
+ "attn_mask_ratio": 0.1,
164
+ "use_masking": false,
165
+ "mask_ratio": 0.4,
166
+ "max_span_length": 8,
167
+ "spacing": 0,
168
+ "r_rand": 0.6,
169
+ "r_block": 0.6,
170
+ "block_span": 4,
171
+ "r_span": 0.4,
172
+ "max_span": 8,
173
+ "dpi_min_factor": 0.5,
174
+ "dpi_max_factor": 1.5,
175
+ "perspective_low": 0.0,
176
+ "perspective_high": 0.4,
177
+ "elastic_distortion_min_kernel_size": 3,
178
+ "elastic_distortion_max_kernel_size": 3,
179
+ "elastic_distortion_max_magnitude": 20,
180
+ "elastic_distortion_min_alpha": 0.5,
181
+ "elastic_distortion_max_alpha": 1,
182
+ "elastic_distortion_min_sigma": 1,
183
+ "elastic_distortion_max_sigma": 10,
184
+ "dila_ero_max_kernel": 2,
185
+ "dila_ero_iter": 1,
186
+ "jitter_contrast": 0.4,
187
+ "jitter_brightness": 0.4,
188
+ "jitter_saturation": 0.4,
189
+ "jitter_hue": 0.2,
190
+ "blur_min_kernel": 3,
191
+ "blur_max_kernel": 5,
192
+ "blur_min_sigma": 3,
193
+ "blur_max_sigma": 5,
194
+ "sharpen_min_alpha": 0,
195
+ "sharpen_max_alpha": 1,
196
+ "sharpen_min_strength": 0,
197
+ "sharpen_max_strength": 1,
198
+ "zoom_min_h": 0.8,
199
+ "zoom_max_h": 1,
200
+ "zoom_min_w": 0.99,
201
+ "zoom_max_w": 1,
202
+ "proba": 0.5,
203
+ "decoder_layers": 6,
204
+ "decoder_heads": 8,
205
+ "max_seq_len": 256,
206
+ "label_smoothing": 0.1,
207
+ "beam_size": 5,
208
+ "generation_method": "nucleus",
209
+ "generation_temperature": 0.7,
210
+ "repetition_penalty": 1.3,
211
+ "top_p": 0.9,
212
+ "tcm_enable": true,
213
+ "tcm_lambda": 1.0,
214
+ "ctc_lambda": 0.1,
215
+ "tcm_sub_len": 5,
216
+ "tcm_warmup_iters": 0,
217
+ "resume": null,
218
+ "load_model": null,
219
+ "load_encoder_only": false,
220
+ "strict_loading": true,
221
+ "save_dir": "./output/vnondb"
222
+ },
223
  "load_error": null
224
  },
225
  {
 
229
  "size_bytes": 529516245,
230
  "size_mb": 504.986,
231
  "sha256": "01d3552f42356eecb1625dfc8692306f49efab1c7a3e116d819c9cc42d317d1f",
232
+ "modified": "2025-12-16T09:16:17.590814",
233
+ "created": "2025-12-16T09:16:17.347489",
234
  "dataset": "read",
235
  "best_cer": 0.039916476472587774,
236
  "best_wer": 0.17772215269086358,