wav2vec2_xls_r_300m_BIG-C_Bemba_5hr_v2

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: 0.7579
• Model Preparation Time: 0.0073
• Wer: 0.9757
• Cer: 0.2009

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
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• 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	Model Preparation Time	Wer	Cer
18.1773	0.9870	38	16.6918	0.0073	1.0	0.9927
12.4914	2.0	77	7.1287	0.0073	1.0	1.0
5.4066	2.9870	115	4.4523	0.0073	1.0	1.0
4.0047	4.0	154	3.7664	0.0073	1.0	1.0
3.573	4.9870	192	3.3104	0.0073	1.0	1.0
3.1192	6.0	231	3.0163	0.0073	1.0	1.0
2.9907	6.9870	269	2.8848	0.0073	1.0	1.0
2.8352	8.0	308	2.8369	0.0073	1.0	1.0
2.8465	8.9870	346	2.7657	0.0073	1.0	1.0
2.6157	10.0	385	2.5019	0.0073	1.0	1.0
2.1828	10.9870	423	1.7628	0.0073	1.0	0.6116
1.147	12.0	462	1.0935	0.0073	0.9979	0.3337
0.8245	12.9870	500	1.0108	0.0073	1.0	0.3379
0.6669	14.0	539	0.9689	0.0073	1.0	0.3462
0.6124	14.9870	577	0.8346	0.0073	0.9937	0.2478
0.5405	16.0	616	0.8864	0.0073	1.0	0.2856
0.5005	16.9870	654	0.7386	0.0073	0.9811	0.2065
0.4501	18.0	693	0.7730	0.0073	0.9874	0.2220
0.4325	18.9870	731	0.8231	0.0073	0.9853	0.2136
0.3731	20.0	770	0.8173	0.0073	0.9832	0.1987
0.3322	20.9870	808	0.9589	0.0073	0.9853	0.2348
0.2948	22.0	847	0.9515	0.0073	0.9916	0.2257
0.2721	22.9870	885	0.8517	0.0073	0.9874	0.2014
0.2422	24.0	924	0.8960	0.0073	0.9916	0.2106
0.2328	24.9870	962	1.1535	0.0073	0.9916	0.2219
0.2091	26.0	1001	0.9904	0.0073	0.9853	0.2251
0.1883	26.9870	1039	0.8991	0.0073	0.9958	0.2045
0.1696	28.0	1078	1.0196	0.0073	0.9937	0.2296
0.16	28.9870	1116	1.0126	0.0073	0.9874	0.2193
0.1421	30.0	1155	1.0042	0.0073	0.9874	0.2185
0.1381	30.9870	1193	1.0587	0.0073	0.9958	0.2273
0.1262	32.0	1232	1.0195	0.0073	0.9979	0.2116
0.1195	32.9870	1270	1.0541	0.0073	0.9937	0.2224
0.1099	34.0	1309	1.1546	0.0073	0.9937	0.2376
0.1054	34.9870	1347	1.1324	0.0073	0.9853	0.2364
0.103	36.0	1386	1.1279	0.0073	0.9979	0.2349
0.1006	36.9870	1424	1.1162	0.0073	0.9895	0.2198

Framework versions

• Transformers 4.43.3
• Pytorch 2.2.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1