README.md

metadata

language:
  - en
license: mit
base_model: microsoft/deberta-v3-base
tags:
  - nycu-112-2-datamining-hw2
  - generated_from_trainer
datasets:
  - DandinPower/review_mergeallfeaturetotext
metrics:
  - accuracy
model-index:
  - name: deberta-v3-base-maftt
    results:
      - task:
          name: Text Classification
          type: text-classification
        dataset:
          name: DandinPower/review_mergeallfeaturetotext
          type: DandinPower/review_mergeallfeaturetotext
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.6288571428571429

deberta-v3-base-maftt

This model is a fine-tuned version of microsoft/deberta-v3-base on the DandinPower/review_mergeallfeaturetotext dataset. It achieves the following results on the evaluation set:

• Loss: 1.4616
• Accuracy: 0.6289
• Macro F1: 0.6302

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: 4.5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1500
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Macro F1
1.0302	0.14	500	1.0771	0.5511	0.5499
1.0412	0.29	1000	0.9406	0.5966	0.6030
0.9494	0.43	1500	0.9546	0.5949	0.5602
0.898	0.57	2000	1.0436	0.5957	0.5872
0.9171	0.71	2500	0.9004	0.622	0.6074
0.8856	0.86	3000	0.8741	0.6137	0.5990
0.9359	1.0	3500	0.8821	0.6267	0.6245
0.8626	1.14	4000	0.8859	0.6213	0.6200
0.7953	1.29	4500	0.8606	0.6337	0.6271
0.8206	1.43	5000	0.8543	0.6169	0.6202
0.8184	1.57	5500	0.9360	0.6266	0.6165
0.8044	1.71	6000	0.8606	0.6234	0.6227
0.7094	1.86	6500	0.8842	0.6434	0.6387
0.8264	2.0	7000	0.9063	0.612	0.6128
0.6951	2.14	7500	0.8782	0.6386	0.6415
0.704	2.29	8000	0.9510	0.6326	0.6308
0.6806	2.43	8500	0.8709	0.6413	0.6455
0.6983	2.57	9000	0.8977	0.6426	0.6436
0.6852	2.71	9500	0.9686	0.5984	0.6010
0.6761	2.86	10000	0.8961	0.6386	0.6406
0.6804	3.0	10500	0.9378	0.6307	0.6332
0.5329	3.14	11000	1.1209	0.6341	0.6382
0.5461	3.29	11500	1.0323	0.6393	0.6377
0.5725	3.43	12000	1.0678	0.6334	0.6366
0.5499	3.57	12500	1.0547	0.6374	0.6394
0.5218	3.71	13000	1.0524	0.6453	0.6460
0.5022	3.86	13500	1.1100	0.6363	0.6358
0.534	4.0	14000	1.0378	0.6357	0.6386
0.3823	4.14	14500	1.3985	0.6357	0.6357
0.4518	4.29	15000	1.3265	0.6314	0.6318
0.4147	4.43	15500	1.3946	0.631	0.6324
0.3936	4.57	16000	1.4649	0.6279	0.6308
0.4339	4.71	16500	1.5322	0.6286	0.6314
0.4448	4.86	17000	1.4890	0.629	0.6302
0.4006	5.0	17500	1.4616	0.6289	0.6302

Framework versions

• Transformers 4.39.3
• Pytorch 2.2.2+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2