Modelpublic

CLEAR-Global/w2v-bert-2.0-igbo_naijavoices_500h

sourceHugging Facemitupdated 1y agoView on Hugging Face

1likes215downloads

Model Card

w2v-bert-2.0-igbonaijavoices500h

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the None dataset. It achieves the following results on the evaluation set:

—Loss: 0.2218
—Wer: 0.2455
—Cer: 0.1105

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: 3e-05
—trainbatchsize: 160
—evalbatchsize: 160
—seed: 42
—distributed_type: multi-GPU
—num_devices: 2
—totaltrainbatch_size: 320
—totalevalbatch_size: 320
—optimizer: Use OptimizerNames.ADAMWTORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizerargs=No additional optimizer arguments
—lrschedulertype: linear
—lrschedulerwarmup_ratio: 0.1
—num_epochs: 50.0
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.923	0.6901	1000	0.6313	0.5401	0.2069
0.4648	1.3803	2000	0.4222	0.3868	0.1549
0.3635	2.0704	3000	0.3582	0.3499	0.1434
0.3388	2.7605	4000	0.3368	0.3322	0.1383
0.2967	3.4507	5000	0.3141	0.3191	0.1366
0.2738	4.1408	6000	0.3041	0.3151	0.1328
0.3146	4.8309	7000	0.2972	0.3091	0.1297
0.2612	5.5210	8000	0.2856	0.2998	0.1312
0.282	6.2112	9000	0.2873	0.3001	0.1300
0.2989	6.9013	10000	0.2864	0.2959	0.1309
0.2633	7.5914	11000	0.2660	0.2883	0.1242
0.2471	8.2816	12000	0.2674	0.2905	0.1263
0.2746	8.9717	13000	0.2671	0.2822	0.1224
0.2754	9.6618	14000	0.2617	0.2833	0.1234
0.2881	10.3520	15000	0.2596	0.2833	0.1229
0.2717	11.0421	16000	0.2524	0.2760	0.1221
0.2204	11.7322	17000	0.2513	0.2720	0.1197
0.2429	12.4224	18000	0.2530	0.2738	0.1203
0.2429	13.1125	19000	0.2511	0.2745	0.1194
0.2449	13.8026	20000	0.2555	0.2748	0.1209
0.2053	14.4928	21000	0.2464	0.2719	0.1181
0.222	15.1829	22000	0.2428	0.2659	0.1195
0.1874	15.8730	23000	0.2418	0.2609	0.1156
0.1924	16.5631	24000	0.2363	0.2675	0.1176
0.1855	17.2533	25000	0.2336	0.2629	0.1151
0.2172	17.9434	26000	0.2302	0.2633	0.1153
0.2074	18.6335	27000	0.2345	0.2588	0.1161
0.1589	19.3237	28000	0.2204	0.2474	0.1137
0.1382	20.0138	29000	0.2279	0.2531	0.1127
0.1525	20.7039	30000	0.2284	0.2550	0.1128
0.149	21.3941	31000	0.2221	0.2509	0.1122
0.1193	22.0842	32000	0.2229	0.2482	0.1113
0.1245	22.7743	33000	0.2218	0.2455	0.1105

Framework versions

—Transformers 4.48.1
—Pytorch 2.7.1+cu126
—Datasets 4.0.0
—Tokenizers 0.21.2