wav2vec2-burak-new-300-v2-3

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2910
• Wer: 0.2712

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: 32
• 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: 500
• num_epochs: 71

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
7.5554	4.9	500	3.1912	1.0
3.1116	9.8	1000	2.5931	1.0
1.291	14.71	1500	0.4394	0.5575
0.6214	19.61	2000	0.3243	0.4370
0.4501	24.51	2500	0.2890	0.3517
0.3659	29.41	3000	0.2809	0.3310
0.3065	34.31	3500	0.2882	0.3076
0.2739	39.22	4000	0.2632	0.2870
0.2516	44.12	4500	0.2872	0.2897
0.2316	49.02	5000	0.2875	0.2767
0.2159	53.92	5500	0.2953	0.2746
0.1957	58.82	6000	0.2955	0.2787
0.1997	63.73	6500	0.2963	0.2822
0.1852	68.63	7000	0.2910	0.2712

Framework versions

• Transformers 4.22.2
• Pytorch 1.12.1+cu113
• Datasets 2.5.1
• Tokenizers 0.12.1