metadata

license: mit
base_model: nielsr/lilt-xlm-roberta-base
tags:
  - generated_from_trainer
datasets:
  - xfun
metrics:
  - precision
  - recall
  - f1
  - accuracy
model-index:
  - name: LiLT-SER-ZH
    results:
      - task:
          name: Token Classification
          type: token-classification
        dataset:
          name: xfun
          type: xfun
          config: xfun.zh
          split: validation
          args: xfun.zh
        metrics:
          - name: Precision
            type: precision
            value: 0.8408488063660478
          - name: Recall
            type: recall
            value: 0.9347968545216252
          - name: F1
            type: f1
            value: 0.8853374709076804
          - name: Accuracy
            type: accuracy
            value: 0.8116519985331867

LiLT-SER-ZH

This model is a fine-tuned version of nielsr/lilt-xlm-roberta-base on the xfun dataset. It achieves the following results on the evaluation set:

Loss: 1.8792
Precision: 0.8408
Recall: 0.9348
F1: 0.8853
Accuracy: 0.8117

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: 8
eval_batch_size: 2
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
training_steps: 10000

Training results

Training Loss	Epoch	Step	Accuracy	F1	Validation Loss	Precision	Recall
0.2166	10.64	500	0.7724	0.8544	1.1239	0.7932	0.9260
0.0238	21.28	1000	0.8201	0.8624	1.0535	0.8572	0.8676
0.0034	31.91	1500	0.8057	0.8599	1.4675	0.8088	0.9178
0.0163	42.55	2000	0.8232	0.8729	1.2837	0.8572	0.8893
0.0037	53.19	2500	0.8114	0.8627	1.5315	0.8142	0.9174
0.0003	63.83	3000	0.8137	0.8652	1.4604	0.8471	0.8840
0.0005	74.47	3500	0.8115	0.8767	1.5980	0.8409	0.9158
0.0005	85.11	4000	0.8129	0.8634	1.5108	0.8261	0.9043
0.0004	95.74	4500	0.8161	0.8817	1.7719	0.8397	0.9282
0.0001	106.38	5000	0.8203	0.8813	1.4313	0.8600	0.9037
0.0001	117.02	5500	0.8181	0.8832	1.5232	0.8509	0.9181
0.0	127.66	6000	0.8069	0.8808	1.6845	0.8532	0.9102
0.0179	138.3	6500	0.8192	0.8793	1.6293	0.8398	0.9227
0.0	148.94	7000	0.8081	0.8815	1.8209	0.8381	0.9296
0.0	159.57	7500	1.8224	0.8443	0.9184	0.8798	0.8070
0.0	170.21	8000	1.7810	0.8450	0.9305	0.8857	0.8127
0.0	180.85	8500	1.7531	0.8454	0.9230	0.8825	0.8088
0.0	191.49	9000	1.8757	0.8394	0.9302	0.8825	0.8070
0.0	202.13	9500	1.8757	0.8417	0.9338	0.8854	0.8123
0.0	212.77	10000	1.8792	0.8408	0.9348	0.8853	0.8117

Framework versions

Transformers 4.35.2
Pytorch 2.1.0+cu121
Datasets 2.18.0
Tokenizers 0.15.1