README.md

metadata

base_model: microsoft/wavlm-base-plus
tags:
  - generated_from_trainer
model-index:
  - name: wavlm_base-plus_emodb
    results: []

wavlm_base-plus_emodb

This model is a fine-tuned version of microsoft/wavlm-base-plus on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.1390
• Uar: 0.6759
• Acc: 0.7426

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Uar	Acc
No log	0.31	1	1.3804	0.3148	0.4559
No log	0.62	2	1.3739	0.2593	0.4118
No log	0.92	3	1.3586	0.25	0.4044
1.4729	1.23	4	1.3445	0.25	0.4044
1.4729	1.54	5	1.3265	0.3056	0.4485
1.4729	1.85	6	1.3054	0.4167	0.5368
1.3428	2.15	7	1.2888	0.4352	0.5515
1.3428	2.46	8	1.2719	0.4630	0.5735
1.3428	2.77	9	1.2511	0.5093	0.6103
1.2214	3.08	10	1.2465	0.5833	0.6691
1.2214	3.38	11	1.2409	0.5370	0.6324
1.2214	3.69	12	1.2366	0.5000	0.6029
1.2214	4.0	13	1.2346	0.5185	0.6176
0.7965	4.31	14	1.2130	0.6574	0.7279
0.7965	4.62	15	1.1881	0.7222	0.7794
0.7965	4.92	16	1.1775	0.7407	0.7941
0.9522	5.23	17	1.1707	0.7315	0.7868
0.9522	5.54	18	1.1667	0.7222	0.7794
0.9522	5.85	19	1.1636	0.7130	0.7721
0.8702	6.15	20	1.1628	0.7037	0.7647
0.8702	6.46	21	1.1557	0.7037	0.7647
0.8702	6.77	22	1.1444	0.7130	0.7721
0.7803	7.08	23	1.1378	0.7130	0.7721
0.7803	7.38	24	1.1331	0.7130	0.7721
0.7803	7.69	25	1.1339	0.7037	0.7647
0.7803	8.0	26	1.1363	0.6944	0.7574
0.5654	8.31	27	1.1382	0.6759	0.7426
0.5654	8.62	28	1.1394	0.6759	0.7426
0.5654	8.92	29	1.1395	0.6759	0.7426
0.7148	9.23	30	1.1390	0.6759	0.7426

Framework versions

• Transformers 4.32.0
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.13.3