This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b on the common_voice_16_1 dataset. It achieves the following results on the evaluation set:

Loss: 0.6040
Wer: 0.4719

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: 0.0005
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: 30
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.2653	0.29	400	1.1675	0.9024
1.0108	0.58	800	1.2504	0.9083
0.9497	0.88	1200	1.2197	0.9001
0.8842	1.17	1600	0.8955	0.8195
0.8603	1.46	2000	0.8623	0.8267
0.8271	1.75	2400	0.8324	0.7950
0.8037	2.05	2800	0.7933	0.7751
0.7488	2.34	3200	0.8006	0.7841
0.7683	2.63	3600	0.8042	0.7925
0.7554	2.92	4000	0.7955	0.7728
0.7078	3.21	4400	0.8038	0.7781
0.7161	3.51	4800	0.8470	0.7967
0.7114	3.8	5200	0.8424	0.8074
0.696	4.09	5600	0.7482	0.7617
0.6581	4.38	6000	0.7177	0.7427
0.6448	4.67	6400	0.7201	0.7292
0.6343	4.97	6800	0.6995	0.7265
0.6072	5.26	7200	0.7234	0.7549
0.598	5.55	7600	0.6919	0.7215
0.5938	5.84	8000	0.7364	0.7310
0.5641	6.14	8400	0.7075	0.7074
0.5557	6.43	8800	0.6785	0.7069
0.5537	6.72	9200	0.6434	0.7019
0.5475	7.01	9600	0.6415	0.6797
0.5053	7.3	10000	0.6402	0.6751
0.518	7.6	10400	0.6214	0.6618
0.5041	7.89	10800	0.6156	0.6607
0.4853	8.18	11200	0.6600	0.6879
0.4738	8.47	11600	0.6359	0.6659
0.4766	8.77	12000	0.6674	0.6980
0.4822	9.06	12400	0.6189	0.6545
0.4421	9.35	12800	0.6090	0.6434
0.4562	9.64	13200	0.6099	0.6383
0.4559	9.93	13600	0.6009	0.6511
0.4456	10.23	14000	0.6090	0.6409
0.4277	10.52	14400	0.5940	0.6374
0.4228	10.81	14800	0.5992	0.6402
0.4139	11.1	15200	0.6287	0.6344
0.3918	11.4	15600	0.6134	0.6326
0.3957	11.69	16000	0.5952	0.6275
0.4022	11.98	16400	0.5957	0.6350
0.3733	12.27	16800	0.5466	0.6010
0.3749	12.56	17200	0.5566	0.6044
0.3736	12.86	17600	0.5453	0.5994
0.3701	13.15	18000	0.5846	0.6203
0.359	13.44	18400	0.5880	0.5997
0.3501	13.73	18800	0.5738	0.6153
0.3403	14.02	19200	0.5766	0.5855
0.3302	14.32	19600	0.5507	0.5954
0.3295	14.61	20000	0.5467	0.5899
0.3371	14.9	20400	0.5571	0.5907
0.3253	15.19	20800	0.5266	0.5745
0.3054	15.49	21200	0.5211	0.5681
0.3044	15.78	21600	0.5409	0.5698
0.2982	16.07	22000	0.5467	0.5827
0.293	16.36	22400	0.5426	0.5706
0.2901	16.65	22800	0.5404	0.5793
0.2913	16.95	23200	0.5342	0.5647
0.269	17.24	23600	0.5309	0.5623
0.2803	17.53	24000	0.5300	0.5637
0.2697	17.82	24400	0.5103	0.5539
0.2727	18.12	24800	0.5414	0.5607
0.2505	18.41	25200	0.5472	0.5508
0.2554	18.7	25600	0.5260	0.5511
0.2552	18.99	26000	0.5246	0.5389
0.2333	19.28	26400	0.5392	0.5497
0.2315	19.58	26800	0.5230	0.5395
0.2366	19.87	27200	0.5303	0.5344
0.2317	20.16	27600	0.5348	0.5350
0.2229	20.45	28000	0.5138	0.5328
0.2243	20.75	28400	0.5147	0.5235
0.2169	21.04	28800	0.5494	0.5266
0.2068	21.33	29200	0.5361	0.5266
0.2073	21.62	29600	0.5660	0.5346
0.2035	21.91	30000	0.5048	0.5196
0.1954	22.21	30400	0.5498	0.5137
0.1974	22.5	30800	0.5338	0.5157
0.1914	22.79	31200	0.5311	0.5080
0.1874	23.08	31600	0.5600	0.5020
0.1792	23.37	32000	0.5428	0.5012
0.182	23.67	32400	0.5237	0.5013
0.1825	23.96	32800	0.5383	0.4999
0.1723	24.25	33200	0.5690	0.5063
0.1673	24.54	33600	0.5525	0.5030
0.165	24.84	34000	0.5519	0.5001
0.162	25.13	34400	0.5553	0.4966
0.1597	25.42	34800	0.5614	0.4938
0.1505	25.71	35200	0.5569	0.4932
0.157	26.0	35600	0.5629	0.4931
0.1468	26.3	36000	0.5808	0.4879
0.1438	26.59	36400	0.5675	0.4871
0.1462	26.88	36800	0.5568	0.4852
0.138	27.17	37200	0.5995	0.4821
0.1394	27.47	37600	0.5810	0.4798
0.1363	27.76	38000	0.5776	0.4771
0.1318	28.05	38400	0.5909	0.4763
0.128	28.34	38800	0.5967	0.4782
0.1304	28.63	39200	0.5866	0.4758
0.1284	28.93	39600	0.5904	0.4747
0.1207	29.22	40000	0.6023	0.4739
0.1275	29.51	40400	0.6038	0.4733
0.1241	29.8	40800	0.6040	0.4719

Framework versions

Transformers 4.38.1
Pytorch 2.2.0+cu121
Datasets 2.17.0
Tokenizers 0.15.2

tgrhn/wav2vec-turkish-300m-xls

wav2vec-turkish-300m-xls

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

tgrhn/
wav2vec-turkish-300m-xls