VOOZH about

URL: https://huggingface.co/merve/rfdetr-docvqa-media3-trainval-agree1-medium

⇱ merve/rfdetr-docvqa-media3-trainval-agree1-medium · Hugging Face

rfdetr-docvqa-media3-trainval-agree1-medium

This model is a fine-tuned version of Roboflow/rf-detr-medium on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 17.1433
  • Map: 0.2703
  • Map 50: 0.4858
  • Map 75: 0.2504
  • Map Small: -1.0
  • Map Medium: 0.0548
  • Map Large: 0.2829
  • Mar 1: 0.3017
  • Mar 10: 0.5088
  • Mar 100: 0.5802
  • Mar Small: -1.0
  • Mar Medium: 0.0818
  • Mar Large: 0.6086
  • Map Chart: 0.2876
  • Mar 100 Chart: 0.67
  • Map Image: 0.3634
  • Mar 100 Image: 0.6505
  • Map Signature: 0.1597
  • Mar 100 Signature: 0.42

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 5e-05
  • train_batch_size: 16
  • eval_batch_size: 8
  • seed: 42
  • optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 0.05
  • num_epochs: 30

Training results

Training Loss Epoch Step Validation Loss Map Map 50 Map 75 Map Small Map Medium Map Large Mar 1 Mar 10 Mar 100 Mar Small Mar Medium Mar Large Map Chart Mar 100 Chart Map Image Mar 100 Image Map Signature Mar 100 Signature
53.1977 1.0 42 28.6751 0.0006 0.0023 0.0 -1.0 0.0 0.0006 0.0017 0.0242 0.0754 -1.0 0.0 0.0776 0.0015 0.1425 0.0002 0.0838 0.0 0.0
18.3829 2.0 84 22.3031 0.0001 0.0003 0.0 -1.0 0.0 0.0001 0.0003 0.0111 0.0505 -1.0 0.0 0.0543 0.0 0.0 0.0003 0.1515 0.0 0.0
12.6798 3.0 126 18.0869 0.0013 0.0048 0.0006 -1.0 0.0001 0.0014 0.0116 0.0877 0.2137 -1.0 0.0429 0.224 0.0018 0.1675 0.0019 0.4525 0.0002 0.0211
12.3407 4.0 168 16.4894 0.023 0.0532 0.0209 -1.0 0.0028 0.0253 0.0933 0.2201 0.3409 -1.0 0.1701 0.3613 0.0072 0.0375 0.0465 0.6374 0.0153 0.3478
11.5202 5.0 210 16.6346 0.0759 0.1353 0.072 -1.0 0.0238 0.0804 0.1753 0.3318 0.4685 -1.0 0.2994 0.4833 0.0221 0.415 0.1495 0.6293 0.056 0.3611
10.8266 6.0 252 16.7964 0.0478 0.1013 0.0416 -1.0 0.0747 0.0492 0.1604 0.3358 0.5535 -1.0 0.2266 0.5731 0.0164 0.61 0.0912 0.6606 0.0358 0.39
10.5741 7.0 294 14.9717 0.0782 0.1557 0.0682 -1.0 0.0566 0.0819 0.2013 0.4125 0.5793 -1.0 0.2123 0.6014 0.0235 0.6475 0.1371 0.6626 0.0739 0.4278
10.4029 8.0 336 15.0839 0.0885 0.1835 0.0812 -1.0 0.0595 0.0932 0.2132 0.4473 0.5749 -1.0 0.1299 0.5994 0.0341 0.6925 0.161 0.6545 0.0704 0.3778
9.9259 9.0 378 14.9783 0.1416 0.2663 0.1187 -1.0 0.0359 0.1495 0.2359 0.4447 0.6108 -1.0 0.2318 0.6339 0.0716 0.69 0.2244 0.699 0.1288 0.4433
9.5176 10.0 420 15.0226 0.1601 0.302 0.1381 -1.0 0.1018 0.1691 0.2342 0.4814 0.6045 -1.0 0.3084 0.6235 0.0783 0.69 0.2668 0.6869 0.1352 0.4367
9.3748 11.0 462 16.2226 0.1894 0.3387 0.1733 -1.0 0.0955 0.1964 0.2599 0.4897 0.6145 -1.0 0.2682 0.6361 0.159 0.7025 0.2695 0.7111 0.1397 0.43
9.2042 12.0 504 15.7788 0.1941 0.3492 0.1943 -1.0 0.1027 0.2004 0.2632 0.5003 0.6026 -1.0 0.2773 0.6214 0.1436 0.7075 0.3131 0.7081 0.1255 0.3922
8.8070 13.0 546 15.3817 0.2003 0.3608 0.1933 -1.0 0.1393 0.2074 0.2804 0.5062 0.6149 -1.0 0.2708 0.6353 0.177 0.73 0.2941 0.7081 0.1297 0.4067
8.7009 14.0 588 16.3511 0.2263 0.4101 0.2175 -1.0 0.0849 0.2353 0.2889 0.5054 0.6006 -1.0 0.1864 0.6235 0.2408 0.715 0.288 0.6879 0.1502 0.3989
8.7887 15.0 630 15.6987 0.2323 0.4316 0.2118 -1.0 0.0478 0.2447 0.2927 0.4922 0.5853 -1.0 0.1922 0.6087 0.1911 0.6825 0.3247 0.6657 0.1812 0.4078
8.3357 16.0 672 16.1288 0.2687 0.4568 0.2566 -1.0 0.057 0.2795 0.2939 0.5004 0.5913 -1.0 0.2201 0.6122 0.311 0.7025 0.339 0.6859 0.1563 0.3856
8.1628 17.0 714 16.2324 0.2595 0.4474 0.2618 -1.0 0.0486 0.2733 0.2994 0.5051 0.6005 -1.0 0.0916 0.6281 0.2617 0.7375 0.3408 0.6596 0.1758 0.4044
7.9911 18.0 756 16.4372 0.2829 0.4742 0.2856 -1.0 0.0997 0.2936 0.3091 0.4983 0.5905 -1.0 0.2325 0.6124 0.3094 0.6675 0.3641 0.6939 0.1752 0.41
8.0354 19.0 798 16.2935 0.2817 0.4902 0.2885 -1.0 0.0872 0.2925 0.3144 0.5133 0.5951 -1.0 0.1903 0.6167 0.3241 0.705 0.3462 0.6869 0.1749 0.3933
7.6743 20.0 840 17.1765 0.2633 0.4703 0.2387 -1.0 0.054 0.2757 0.3124 0.506 0.5809 -1.0 0.0942 0.6073 0.3024 0.705 0.3233 0.6576 0.1643 0.38
7.5240 21.0 882 17.1978 0.2816 0.4809 0.2598 -1.0 0.0604 0.2951 0.2959 0.4943 0.5849 -1.0 0.0994 0.6134 0.2918 0.6725 0.3733 0.6667 0.1798 0.4156
7.3807 22.0 924 17.3134 0.2652 0.4629 0.2384 -1.0 0.0692 0.2771 0.2969 0.4931 0.5889 -1.0 0.1266 0.6138 0.2729 0.71 0.3514 0.6566 0.1714 0.4
7.1151 23.0 966 17.0971 0.2742 0.4702 0.2667 -1.0 0.0495 0.2865 0.3096 0.509 0.6008 -1.0 0.1045 0.6275 0.309 0.7175 0.3481 0.6616 0.1656 0.4233
6.9313 24.0 1008 17.1139 0.2742 0.4647 0.2899 -1.0 0.0612 0.2866 0.3096 0.5009 0.5832 -1.0 0.113 0.6094 0.3169 0.69 0.3259 0.6606 0.1799 0.3989
7.0055 25.0 1050 16.9771 0.2823 0.5006 0.2639 -1.0 0.0678 0.2947 0.3096 0.5099 0.5819 -1.0 0.1214 0.6074 0.2902 0.6775 0.3722 0.6505 0.1845 0.4178
6.8643 26.0 1092 17.1671 0.2717 0.4797 0.2636 -1.0 0.0522 0.2844 0.3102 0.5146 0.5999 -1.0 0.0981 0.628 0.292 0.705 0.3676 0.6626 0.1555 0.4322
6.7321 27.0 1134 17.1871 0.2832 0.4964 0.2763 -1.0 0.0493 0.2963 0.3067 0.4882 0.5683 -1.0 0.0682 0.5972 0.326 0.65 0.3708 0.6515 0.1528 0.4033
7.1077 28.0 1176 17.1420 0.2683 0.4745 0.2651 -1.0 0.0577 0.281 0.2997 0.5036 0.5741 -1.0 0.0818 0.6024 0.2823 0.6575 0.3686 0.6414 0.1539 0.4233
6.7518 29.0 1218 17.1374 0.276 0.4947 0.2527 -1.0 0.0548 0.2887 0.3107 0.5005 0.5766 -1.0 0.0818 0.6048 0.3023 0.6675 0.3673 0.6424 0.1585 0.42
6.9678 30.0 1260 17.1433 0.2703 0.4858 0.2504 -1.0 0.0548 0.2829 0.3017 0.5088 0.5802 -1.0 0.0818 0.6086 0.2876 0.67 0.3634 0.6505 0.1597 0.42

Framework versions

  • Transformers 5.12.1
  • Pytorch 2.12.0+cu130
  • Datasets 5.0.0
  • Tokenizers 0.22.2
Downloads last month
-
Safetensors
Model size
33.4M params
Tensor type
F32
·

Model tree for merve/rfdetr-docvqa-media3-trainval-agree1-medium

Finetuned
(4)
this model

Collection including merve/rfdetr-docvqa-media3-trainval-agree1-medium