Amit65/whisper-small-multilingual

This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6283
• Wer: 80.0691

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

full fine tuning on custom data and evaluate on word error rate(WER)

Training procedure

Apply full fine tuning using hugging face trainer API

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.4481	0.0480	25	1.7935	138.3641
1.494	0.0960	50	1.3053	105.6452
1.4092	0.1440	75	1.1546	102.6498
1.1367	0.1919	100	1.0424	105.4147
0.9748	0.2399	125	1.0038	116.7051
0.9522	0.2879	150	1.0032	140.6682
0.9114	0.3359	175	0.9329	126.2673
0.9498	0.3839	200	0.9077	117.0507
0.8762	0.4319	225	0.9359	97.4654
0.9051	0.4798	250	0.8390	88.5945
0.7941	0.5278	275	0.8869	105.2995
0.8417	0.5758	300	0.8299	109.7926
0.9244	0.6238	325	0.8105	79.9539
0.855	0.6718	350	0.7960	87.5576
0.7516	0.7198	375	0.7844	88.9401
0.9119	0.7678	400	0.8116	87.4424
0.7478	0.8157	425	0.7593	79.0323
0.7125	0.8637	450	0.7280	84.2166
0.8235	0.9117	475	0.7171	88.9401
0.6975	0.9597	500	0.7029	74.8848
0.5599	1.0077	525	0.7060	76.6129
0.4681	1.0557	550	0.6891	100.8065
0.3496	1.1036	575	0.6995	104.9539
0.4196	1.1516	600	0.7102	82.4885
0.3884	1.1996	625	0.6856	104.7235
0.4788	1.2476	650	0.6745	81.6820
0.4237	1.2956	675	0.6722	81.9124
0.4001	1.3436	700	0.6740	83.2949
0.3909	1.3916	725	0.6823	71.8894
0.3435	1.4395	750	0.6934	75.1152
0.344	1.4875	775	0.6810	72.0046
0.3071	1.5355	800	0.6704	71.1982
0.3392	1.5835	825	0.6589	88.3641
0.3742	1.6315	850	0.6532	77.9954
0.4153	1.6795	875	0.6363	79.8387
0.3416	1.7274	900	0.6560	79.4931
0.3121	1.7754	925	0.6320	82.0276
0.2986	1.8234	950	0.6447	76.9585
0.3761	1.8714	975	0.6420	75.8065
0.4394	1.9194	1000	0.6234	77.5346
0.3094	1.9674	1025	0.6430	81.5668
0.3468	2.0154	1050	0.6266	78.5714
0.25	2.0633	1075	0.6251	79.0323
0.1969	2.1113	1100	0.6337	81.2212
0.157	2.1593	1125	0.6367	76.8433
0.2118	2.2073	1150	0.6414	74.4240
0.2207	2.2553	1175	0.6345	77.4194
0.1965	2.3033	1200	0.6414	76.9585
0.1959	2.3512	1225	0.6322	79.6083
0.1668	2.3992	1250	0.6394	81.5668
0.2128	2.4472	1275	0.6361	80.4147
0.173	2.4952	1300	0.6322	74.8848
0.152	2.5432	1325	0.6312	73.3871
0.1897	2.5912	1350	0.6334	79.0323
0.1666	2.6392	1375	0.6339	81.1060
0.202	2.6871	1400	0.6283	77.9954
0.1511	2.7351	1425	0.6296	80.8756
0.1616	2.7831	1450	0.6313	80.4147
0.1482	2.8311	1475	0.6289	80.5300
0.1672	2.8791	1500	0.6283	80.0691

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1