asr-africa/W2V2_Bert_BIG_C_Bemba_167hr_v1

facebook/w2v-bert-2.0

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

• Loss: 0.4054
• Wer: 0.4003
• Cer: 0.0766

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-06
• train_batch_size: 2
• eval_batch_size: 1
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.025
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.0933	1.0	41178	0.5653	0.4240	0.1120
0.5359	2.0	82356	0.5139	0.3772	0.1026
0.4943	3.0	123534	0.4832	0.3560	0.0996
0.4599	4.0	164712	0.4774	0.3378	0.0948
0.4331	5.0	205890	0.4882	0.3305	0.0931
0.4092	6.0	247068	0.4580	0.3281	0.0921
0.3826	7.0	288246	0.4873	0.3232	0.0903
0.3536	8.0	329424	0.5067	0.3227	0.0908
0.3231	9.0	370602	0.5101	0.3274	0.0938
0.2924	10.0	411780	0.5481	0.3290	0.0927
0.263	11.0	452958	0.5684	0.3320	0.0927
0.2364	12.0	494136	0.5973	0.3362	0.0935
0.2135	13.0	535314	0.6344	0.3405	0.0951
0.1941	14.0	576492	0.7075	0.3370	0.0939
0.1765	15.0	617670	0.7800	0.3398	0.0947
0.1615	16.0	658848	0.8164	0.3389	0.0941
0.1482	17.0	700026	0.8562	0.3410	0.0949

Framework versions

• Transformers 4.45.2
• Pytorch 2.2.0+cu121
• Datasets 3.0.1
• Tokenizers 0.20.1