wav2vec2_xls_r_300m_BIG-C_Bemba_10hr_v4

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.7362
• Wer: 0.7282
• Cer: 0.1958

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-05
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 500
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
9.2663	0.3268	200	3.6743	1.0	1.0
3.1749	0.6536	400	2.8927	1.0	1.0
2.7952	0.9804	600	2.3885	1.0	0.9155
1.4097	1.3072	800	0.8953	0.8535	0.2054
0.9475	1.6340	1000	0.7894	0.7777	0.1771
0.8738	1.9608	1200	0.7285	0.7626	0.1766
0.8183	2.2876	1400	0.9577	0.9257	0.3696
0.6813	2.6144	1600	0.6992	0.7143	0.1708
0.6788	2.9412	1800	0.6921	0.7035	0.1652
0.6277	3.2680	2000	0.8239	0.8301	0.2821
0.5931	3.5948	2200	0.7037	0.7125	0.1818
0.6129	3.9216	2400	0.8998	0.8762	0.3413
0.5626	4.2484	2600	0.7700	0.8201	0.2259
0.5221	4.5752	2800	0.7325	0.7485	0.1917
0.534	4.9020	3000	0.7069	0.7325	0.1892
0.464	5.2288	3200	0.7738	0.7416	0.2102
0.4342	5.5556	3400	0.6993	0.6657	0.1726
0.4615	5.8824	3600	0.7637	0.6915	0.2038
0.3996	6.2092	3800	0.8508	0.7535	0.2423
0.4066	6.5359	4000	0.8229	0.7200	0.2134
0.3671	6.8627	4200	0.8800	0.7616	0.2736
0.3431	7.1895	4400	0.7773	0.6975	0.1983
0.3071	7.5163	4600	0.7757	0.6982	0.1816
0.3387	7.8431	4800	0.8184	0.6978	0.1948
0.318	8.1699	5000	0.8252	0.7157	0.2000
0.2712	8.4967	5200	0.9077	0.7203	0.2043
0.2907	8.8235	5400	0.9530	0.7550	0.2316
0.2574	9.1503	5600	0.8278	0.6963	0.1926
0.2475	9.4771	5800	0.9854	0.7638	0.2411
0.2602	9.8039	6000	1.0509	0.7938	0.2395
0.2572	10.1307	6200	0.9754	0.7698	0.2340
0.2224	10.4575	6400	0.8636	0.7174	0.1987

Framework versions

• Transformers 4.44.0
• Pytorch 2.2.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1