Hugging Face's logo

topdu
/
MDiff4STR

Model card Files
xet
Community
MDiff4STR
File size: 4,231 Bytes 
6730f86
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
 
Global:
  device: gpu
  epoch_num: 40
  log_smooth_window: 20
  print_batch_step: 10
  output_dir: ./output/rec/u14m_filter/svtrv2_llada_test_lr00005_fs_reflect_onedecoder_semiar_withoutmask_sample3/
  save_epoch_step:
  - 10
  - 1
  eval_batch_step:
  - 0
  - 500
  eval_epoch_step:
  - 0
  - 1
  cal_metric_during_train: false
  pretrained_model: null
  checkpoints: null
  use_tensorboard: false
  infer_img: ./img_44_1.jpg
  character_dict_path: ./tools/utils/EN_symbol_dict.txt
  max_text_length: 25
  use_space_char: false
  save_res_path: ./output/rec/u14m_filter/predicts_svtrv2_nrtr.txt
  use_amp: true
  grad_clip_val: 20.0
  distributed: true
Optimizer:
  name: AdamW
  lr: 0.0005
  weight_decay: 0.05
  filter_bias_and_bn: true
LRScheduler:
  name: OneCycleLR
  warmup_epoch: 1.5
  cycle_momentum: false
Architecture:
  model_type: rec
  algorithm: NRTR
  in_channels: 3
  Transform: null
  Encoder:
    name: SVTRv2LNConvTwo33
    use_pos_embed: false
    dims:
    - 128
    - 256
    - 384
    depths:
    - 6
    - 6
    - 6
    num_heads:
    - 4
    - 8
    - 12
    mixer:
    - - Conv
      - Conv
      - Conv
      - Conv
      - Conv
      - Conv
    - - Conv
      - Conv
      - FGlobal
      - Global
      - Global
      - Global
    - - Global
      - Global
      - Global
      - Global
      - Global
      - Global
    local_k:
    - - 5
      - 5
    - - 5
      - 5
    - - -1
      - -1
    sub_k:
    - - 1
      - 1
    - - 2
      - 1
    - - -1
      - -1
    last_stage: false
    feat2d: false
  Decoder:
    name: MDiffDecoder
    num_decoder_layers: 6
    nhead: 6
    max_len: 25
    parallel_decoding: false
    autoregressive_decoding: false
    low_confidence_decoding: false
    random_mask_decoding: false
    semi_autoregressive_decoding: true
    cloze_mask_decoding: false
    sampler_step: 3
    sample_k: 3
Loss:
  name: NoneLoss
PostProcess:
  name: ARLabelDecode
  character_dict_path: ./tools/utils/EN_symbol_dict.txt
  use_space_char: false
Metric:
  name: RecMetric
  main_indicator: acc
  is_filter: true
Train:
  dataset:
    name: RatioDataSetTVResize
    ds_width: true
    padding: false
    data_dir_list:
    - /data/Union14M-L-LMDB-Filtered/filter_train_challenging
    - /data/Union14M-L-LMDB-Filtered/filter_train_hard
    - /data/Union14M-L-LMDB-Filtered/filter_train_medium
    - /data/Union14M-L-LMDB-Filtered/filter_train_normal
    - /data/Union14M-L-LMDB-Filtered/filter_train_easy
    transforms:
    - DecodeImagePIL:
        img_mode: RGB
    - PARSeqAugPIL: null
    - LLaDALabelEncode:
        character_dict_path: ./tools/utils/EN_symbol_dict.txt
        use_space_char: false
        max_text_length: 25
        train_all_layer: true
        sample_num: 3
    - KeepKeys:
        keep_keys:
        - image
        - label
        - reflect_ids
        - noisy_batch
        - masked_indices
        - p_mask
        - length
  sampler:
    name: RatioSampler
    scales:
    - - 128
      - 32
    first_bs: 256
    fix_bs: false
    divided_factor:
    - 4
    - 16
    is_training: true
  loader:
    shuffle: true
    batch_size_per_card: 256
    drop_last: true
    max_ratio: 4
    num_workers: 4
Eval:
  dataset:
    name: RatioDataSetTVResize
    ds_width: true
    padding: false
    data_dir_list:
    - /data/evaluation/CUTE80
    - /data/evaluation/IC13_857
    - /data/evaluation/IC15_1811
    - /data/evaluation/IIIT5k_3000
    - /data/evaluation/SVT
    - /data/evaluation/SVTP
    transforms:
    - DecodeImagePIL:
        img_mode: RGB
    - ARLabelEncode:
        character_dict_path: ./tools/utils/EN_symbol_dict.txt
        use_space_char: false
        max_text_length: 25
    - KeepKeys:
        keep_keys:
        - image
        - label
        - length
  sampler:
    name: RatioSampler
    scales:
    - - 128
      - 32
    first_bs: 256
    fix_bs: false
    divided_factor:
    - 4
    - 16
    is_training: false
  loader:
    shuffle: false
    drop_last: false
    batch_size_per_card: 256
    max_ratio: 4
    num_workers: 4
filename: svtrv2_llada_fs_reflect_onedecoder_reflectwithoutmask_sample3
config: configs/rec/llada/svtrv2_llada_fs_reflect_onedecoder_reflectwithoutmask_sample3.yml
local_rank: '2'
eval: true