Hugging Face's logo

holipori
/
saved_model_git-base

Image-Text-to-Text
Transformers
PyTorch
TensorBoard
git
Generated from Trainer
Eval Results
Inference Endpoints
Model card Files Metrics Community
saved_model_git-base / README.md
Xinyue Hu
update model card README.md
a5744d3
preview code
|
raw
history blame
7.05 kB
metadata
license: mit
tags:
  - generated_from_trainer
datasets:
  - imagefolder
metrics:
  - rouge
  - bleu
model-index:
  - name: saved_model_git-base
    results:
      - task:
          name: Causal Language Modeling
          type: text-generation
        dataset:
          name: imagefolder
          type: imagefolder
          config: default
          split: train
          args: default
        metrics:
          - name: Rouge1
            type: rouge
            value: 0.2892828477539953
          - name: Bleu
            type: bleu
            value: 0.09269389807461934

saved_model_git-base

This model is a fine-tuned version of microsoft/git-base on the imagefolder dataset. It achieves the following results on the evaluation set:

  • Loss: 0.3234
  • Wer Score: 2.8324
  • Rouge1: 0.2893
  • Rouge2: 0.1553
  • Rougel: 0.2580
  • Rougelsum: 0.2581
  • Meteor: 0.4702
  • Bleu: 0.0927
  • Bleu precisions: [0.19773408749396806, 0.11271121828401212, 0.07041869841876251, 0.047040063834047824]

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: 112
  • eval_batch_size: 112
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 224
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 40
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Score Rouge1 Rouge2 Rougel Rougelsum Meteor Bleu Bleu precisions
0.7742 1.7 1000 0.2768 3.6442 0.2183 0.1135 0.1960 0.1960 0.4151 0.0767 [0.16930678895753237, 0.09557836133795246, 0.05745933755000988, 0.03719485051124419]
0.2757 3.4 2000 0.2530 3.5078 0.2344 0.1280 0.2111 0.2111 0.4438 0.0866 [0.18094996558095722, 0.1062941982267445, 0.0660909502265641, 0.04415642702225686]
0.2558 5.11 3000 0.2431 3.5981 0.2321 0.1292 0.2104 0.2104 0.4488 0.0871 [0.17891643241063893, 0.10662635003059268, 0.06724109416177619, 0.04494076673073175]
0.2435 6.81 4000 0.2340 3.5928 0.2351 0.1337 0.2139 0.2139 0.4576 0.0909 [0.18105927626402774, 0.11015864269439932, 0.07084035989457634, 0.04838043468996083]
0.2341 8.51 5000 0.2283 3.5825 0.2379 0.1374 0.2168 0.2168 0.4641 0.0935 [0.18283567465782452, 0.11268246996142758, 0.07331907529684983, 0.0505040077040832]
0.2265 10.21 6000 0.2246 3.5432 0.2424 0.1415 0.2211 0.2211 0.4718 0.0961 [0.1857985595024412, 0.11537502356732557, 0.07570186326781568, 0.05252149630642548]
0.2203 11.91 7000 0.2207 3.4183 0.2511 0.1480 0.2298 0.2298 0.4795 0.0973 [0.1857490628497766, 0.11647364168304132, 0.07701231893556303, 0.053783746414351505]
0.2143 13.62 8000 0.2182 3.1230 0.2718 0.1608 0.2485 0.2485 0.4922 0.0985 [0.1871246642347378, 0.11779841874899488, 0.07811713540422874, 0.054763192765544234]
0.2094 15.32 9000 0.2168 3.5355 0.2473 0.1472 0.2257 0.2257 0.4833 0.0998 [0.18881918229416875, 0.1190513020004344, 0.0792775774175806, 0.05572707164637924]
0.2038 17.02 10000 0.2160 3.4545 0.2543 0.1509 0.2310 0.2310 0.4891 0.1007 [0.1902420103501397, 0.11991652989168618, 0.07994424738353746, 0.05637425111890453]
0.1967 18.72 11000 0.2173 3.3373 0.2641 0.1559 0.2388 0.2388 0.4945 0.1016 [0.19316197072023297, 0.12102576025295267, 0.08048056261991486, 0.056670825893034946]
0.1884 20.43 12000 0.2221 3.0745 0.2862 0.1681 0.2584 0.2584 0.5010 0.1039 [0.19918408147408226, 0.12401633746143571, 0.08199896497939006, 0.05751641776005364]
0.179 22.13 13000 0.2294 2.9232 0.3017 0.1747 0.2710 0.2710 0.5036 0.1027 [0.19968719829990308, 0.12253582577044991, 0.08059124511363425, 0.05632795013912335]
0.1676 23.83 14000 0.2358 2.8863 0.3017 0.1741 0.2714 0.2714 0.5022 0.1041 [0.20393480253606525, 0.12448069284204198, 0.08151878403794971, 0.056659933056666174]
0.1562 25.53 15000 0.2478 2.8259 0.3036 0.1730 0.2728 0.2728 0.4980 0.1039 [0.20674579460384884, 0.12461471833607483, 0.08099930020993702, 0.055934016820835084]
0.1463 27.23 16000 0.2588 2.7604 0.3066 0.1728 0.2748 0.2748 0.4965 0.1021 [0.2053763571102686, 0.12266641124678603, 0.07922919125284283, 0.05440373665093972]
0.1367 28.94 17000 0.2681 2.8889 0.2945 0.1650 0.2642 0.2643 0.4872 0.0987 [0.2002396944625925, 0.11880872600001989, 0.0763248805512799, 0.052217189262938515]
0.1276 30.64 18000 0.2811 2.7054 0.3062 0.1696 0.2743 0.2743 0.4891 0.1009 [0.2075584522604372, 0.12181684833069355, 0.07765563632441028, 0.05272764092006992]
0.1198 32.34 19000 0.2925 2.8151 0.2945 0.1616 0.2633 0.2633 0.4811 0.0972 [0.20216857878455405, 0.11754981616789531, 0.07455271799398129, 0.0503977409292061]
0.1133 34.04 20000 0.3031 2.8563 0.2910 0.1580 0.2599 0.2599 0.4758 0.0955 [0.20084995564059396, 0.11580160477587004, 0.07298742511539595, 0.04909252992133042]
0.1066 35.74 21000 0.3114 2.8931 0.2864 0.1548 0.2557 0.2557 0.4715 0.0930 [0.1969209496796176, 0.11291130154563171, 0.07089499015389361, 0.047528935183720734]
0.1018 37.45 22000 0.3191 2.8514 0.2884 0.1552 0.2572 0.2572 0.4708 0.0931 [0.19810372685233638, 0.11312981011747879, 0.0707889148571459, 0.04735648678945608]
0.0982 39.15 23000 0.3234 2.8324 0.2893 0.1553 0.2580 0.2581 0.4702 0.0927 [0.19773408749396806, 0.11271121828401212, 0.07041869841876251, 0.047040063834047824]

Framework versions

  • Transformers 4.29.2
  • Pytorch 2.0.0
  • Datasets 2.12.0
  • Tokenizers 0.13.3