README.md

metadata

base_model: ylacombe/w2v-bert-2.0
tags:
  - generated_from_trainer
datasets:
  - common_voice_16_0
metrics:
  - wer
model-index:
  - name: wav2vec2-common_voice-tr-demo
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: common_voice_16_0
          type: common_voice_16_0
          config: tr
          split: test
          args: tr
        metrics:
          - name: Wer
            type: wer
            value: 1

wav2vec2-common_voice-tr-demo

This model is a fine-tuned version of ylacombe/w2v-bert-2.0 on the common_voice_16_0 dataset. It achieves the following results on the evaluation set:

• Loss: 3.0815
• Wer: 1.0

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: 0.00072018208512399
• train_batch_size: 20
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 40
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 5000
• num_epochs: 15.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.27	300	7.2663	1.0
10.5256	0.55	600	3.0893	1.0
10.5256	0.82	900	3.0612	1.0
2.9795	1.1	1200	2.9937	1.0
2.9564	1.37	1500	3.2424	1.0
2.9564	1.64	1800	3.2866	1.0
3.1552	1.92	2100	3.6339	1.0
3.1552	2.19	2400	3.1185	1.0
3.2079	2.47	2700	3.1832	1.0
3.1275	2.74	3000	3.3952	1.0
3.1275	3.01	3300	3.2982	1.0
3.0987	3.29	3600	3.1036	1.0
3.0987	3.56	3900	3.1223	1.0
2.9301	3.84	4200	3.1145	1.0
2.9197	4.11	4500	3.0324	1.0
2.9197	4.38	4800	2.9994	1.9599
2.9023	4.66	5100	2.9917	1.8240
2.9023	4.93	5400	2.9946	1.9589
2.9007	5.21	5700	3.1955	1.0
3.1887	5.48	6000	3.1902	1.0
3.1887	5.75	6300	3.1672	1.0
3.135	6.03	6600	3.2076	1.0
3.135	6.3	6900	3.2120	1.0
3.1482	6.58	7200	3.1832	1.0
3.1546	6.85	7500	3.1799	1.0
3.1546	7.12	7800	3.2452	1.0
3.1567	7.4	8100	3.2319	1.0
3.1567	7.67	8400	3.2228	1.0
3.1719	7.95	8700	3.2055	1.0
3.168	8.22	9000	3.2553	1.0
3.168	8.49	9300	3.1975	1.0
3.1643	8.77	9600	3.2446	1.0
3.1643	9.04	9900	3.2781	1.0
3.169	9.32	10200	3.2597	1.0
3.1789	9.59	10500	3.2586	1.0
3.1789	9.86	10800	3.2690	1.0
3.1701	10.14	11100	3.2737	1.0
3.1701	10.41	11400	3.2738	1.0
3.1698	10.68	11700	3.2595	1.0
3.1595	10.96	12000	3.2467	1.0
3.1595	11.23	12300	3.2524	1.0
3.15	11.51	12600	3.2327	1.0
3.15	11.78	12900	3.2196	1.0
3.1444	12.05	13200	3.1943	1.0
3.132	12.33	13500	3.1911	1.0
3.132	12.6	13800	3.2075	1.0
3.1153	12.88	14100	3.1938	1.0
3.1153	13.15	14400	3.1639	1.0
3.1039	13.42	14700	3.1515	1.0
3.0839	13.7	15000	3.1535	1.0
3.0839	13.97	15300	3.1307	1.0
3.0632	14.25	15600	3.1138	1.0
3.0632	14.52	15900	3.1289	1.0
3.0518	14.79	16200	3.0815	1.0

Framework versions

• Transformers 4.37.0.dev0
• Pytorch 2.1.0+cu121
• Datasets 2.14.5
• Tokenizers 0.15.0