lilt-GottBERT-base-xfund-de

This model is a fine-tuned version of moritzbur/lilt-GottBERT-base on the xfund dataset. It achieves the following results on the evaluation set:

  • Loss: 1.7402
  • Answer: {'precision': 0.7931914893617021, 'recall': 0.8589861751152074, 'f1': 0.8247787610619469, 'number': 1085}
  • Header: {'precision': 0.5581395348837209, 'recall': 0.41379310344827586, 'f1': 0.4752475247524752, 'number': 58}
  • Question: {'precision': 0.7877906976744186, 'recall': 0.7465564738292011, 'f1': 0.7666195190947666, 'number': 726}
  • Overall Precision: 0.7859
  • Overall Recall: 0.8015
  • Overall F1: 0.7936
  • Overall Accuracy: 0.7255

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: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.1
  • training_steps: 2000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Answer Header Question Overall Precision Overall Recall Overall F1 Overall Accuracy
0.0373 20.0 200 1.8211 {'precision': 0.7350565428109854, 'recall': 0.8387096774193549, 'f1': 0.7834696513129574, 'number': 1085} {'precision': 0.5135135135135135, 'recall': 0.3275862068965517, 'f1': 0.4, 'number': 58} {'precision': 0.7130102040816326, 'recall': 0.7699724517906336, 'f1': 0.7403973509933776, 'number': 726} 0.7227 0.7961 0.7576 0.7076
0.0345 40.0 400 2.1454 {'precision': 0.7412698412698413, 'recall': 0.8608294930875576, 'f1': 0.796588486140725, 'number': 1085} {'precision': 0.48148148148148145, 'recall': 0.4482758620689655, 'f1': 0.4642857142857143, 'number': 58} {'precision': 0.6554809843400448, 'recall': 0.8071625344352618, 'f1': 0.7234567901234568, 'number': 726} 0.7002 0.8272 0.7584 0.6866
0.0114 60.0 600 2.0185 {'precision': 0.8492723492723493, 'recall': 0.7529953917050691, 'f1': 0.7982413287738153, 'number': 1085} {'precision': 0.7857142857142857, 'recall': 0.3793103448275862, 'f1': 0.5116279069767441, 'number': 58} {'precision': 0.7317073170731707, 'recall': 0.7851239669421488, 'f1': 0.7574750830564784, 'number': 726} 0.7965 0.7539 0.7746 0.7294
0.0043 80.0 800 1.7402 {'precision': 0.7931914893617021, 'recall': 0.8589861751152074, 'f1': 0.8247787610619469, 'number': 1085} {'precision': 0.5581395348837209, 'recall': 0.41379310344827586, 'f1': 0.4752475247524752, 'number': 58} {'precision': 0.7877906976744186, 'recall': 0.7465564738292011, 'f1': 0.7666195190947666, 'number': 726} 0.7859 0.8015 0.7936 0.7255
0.0013 100.0 1000 1.8975 {'precision': 0.8072727272727273, 'recall': 0.8184331797235023, 'f1': 0.8128146453089244, 'number': 1085} {'precision': 0.5, 'recall': 0.41379310344827586, 'f1': 0.4528301886792453, 'number': 58} {'precision': 0.7246022031823746, 'recall': 0.8154269972451791, 'f1': 0.7673363577446531, 'number': 726} 0.7654 0.8047 0.7846 0.7248
0.0009 120.0 1200 1.8875 {'precision': 0.8050314465408805, 'recall': 0.8258064516129032, 'f1': 0.8152866242038216, 'number': 1085} {'precision': 0.6666666666666666, 'recall': 0.3793103448275862, 'f1': 0.48351648351648346, 'number': 58} {'precision': 0.7094017094017094, 'recall': 0.800275482093664, 'f1': 0.7521035598705502, 'number': 726} 0.7628 0.8020 0.7820 0.7334
0.0003 140.0 1400 1.9918 {'precision': 0.8246575342465754, 'recall': 0.832258064516129, 'f1': 0.8284403669724771, 'number': 1085} {'precision': 0.4716981132075472, 'recall': 0.43103448275862066, 'f1': 0.45045045045045046, 'number': 58} {'precision': 0.7354430379746836, 'recall': 0.800275482093664, 'f1': 0.766490765171504, 'number': 726} 0.7786 0.8074 0.7928 0.7316
0.0003 160.0 1600 2.4537 {'precision': 0.7632850241545893, 'recall': 0.8737327188940092, 'f1': 0.8147829823807479, 'number': 1085} {'precision': 0.6857142857142857, 'recall': 0.41379310344827586, 'f1': 0.5161290322580646, 'number': 58} {'precision': 0.7536231884057971, 'recall': 0.7878787878787878, 'f1': 0.7703703703703704, 'number': 726} 0.7583 0.8261 0.7908 0.6903
0.0004 180.0 1800 2.1619 {'precision': 0.785593220338983, 'recall': 0.8543778801843318, 'f1': 0.8185430463576159, 'number': 1085} {'precision': 0.5641025641025641, 'recall': 0.3793103448275862, 'f1': 0.4536082474226804, 'number': 58} {'precision': 0.7718579234972678, 'recall': 0.778236914600551, 'f1': 0.7750342935528121, 'number': 726} 0.7760 0.8101 0.7927 0.7197
0.0003 200.0 2000 2.1507 {'precision': 0.7948051948051948, 'recall': 0.8460829493087557, 'f1': 0.8196428571428571, 'number': 1085} {'precision': 0.631578947368421, 'recall': 0.41379310344827586, 'f1': 0.5, 'number': 58} {'precision': 0.7438551099611902, 'recall': 0.7920110192837465, 'f1': 0.7671781187458305, 'number': 726} 0.7716 0.8117 0.7911 0.7207

Framework versions

  • Transformers 4.45.1
  • Pytorch 2.4.0
  • Datasets 3.0.1
  • Tokenizers 0.20.0
Downloads last month
51
Safetensors
Model size
132M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for moritzbur/lilt-GottBERT-base-xfund-de

Finetuned
(1)
this model