whisper-tiny-mongolian-ver_0.4

This model is a fine-tuned version of openai/whisper-tiny on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.2186
• Wer: 0.7197
• Cer: 0.3174

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: 3.5e-06
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 60
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
4.8371	0.3914	200	2.8192	1.4535	2.1342
2.5062	0.7828	400	1.8720	1.0734	0.6824
2.0403	1.1742	600	1.4946	0.9449	0.5009
1.7612	1.5656	800	1.3044	0.8948	0.4486
1.6561	1.9569	1000	1.1837	0.8693	0.3876
1.5081	2.3483	1200	1.1053	0.8484	0.3905
1.4746	2.7397	1400	1.0471	0.8246	0.3658
1.4252	3.1311	1600	1.0011	0.8285	0.3604
1.3403	3.5225	1800	0.9659	0.7957	0.3442
1.3305	3.9139	2000	0.9380	0.7909	0.3336
1.2562	4.3053	2200	0.9085	0.7811	0.3521
1.2316	4.6967	2400	0.8907	0.7902	0.3472
1.2032	5.0881	2600	0.8755	0.7733	0.3290
1.1524	5.4795	2800	0.8579	0.7638	0.3313
1.1564	5.8708	3000	0.8423	0.7519	0.3254
1.0937	6.2622	3200	0.8351	0.7495	0.3189
1.0801	6.6536	3400	0.8236	0.7407	0.3212
1.0562	7.0450	3600	0.8176	0.7326	0.3103
1.0198	7.4364	3800	0.8133	0.7226	0.3080
1.0191	7.8278	4000	0.8041	0.7254	0.3165
0.9625	8.2192	4200	0.8038	0.7200	0.2990
0.9475	8.6106	4400	0.7985	0.7167	0.3083
0.9789	9.0020	4600	0.7912	0.7090	0.3063
0.9008	9.3933	4800	0.7943	0.7065	0.2987
0.9077	9.7847	5000	0.7911	0.7094	0.3106
0.8888	10.1761	5200	0.7948	0.7051	0.3013
0.8615	10.5675	5400	0.7914	0.6983	0.2968
0.8720	10.9589	5600	0.7910	0.6980	0.2932
0.8315	11.3503	5800	0.7932	0.7053	0.3065
0.8060	11.7417	6000	0.7918	0.6909	0.3128
0.7898	12.1331	6200	0.7929	0.6962	0.3027
0.7662	12.5245	6400	0.7988	0.6953	0.3027
0.8050	12.9159	6600	0.7921	0.6948	0.2952
0.7373	13.3072	6800	0.7988	0.6938	0.2963
0.7363	13.6986	7000	0.8036	0.6912	0.2985
0.7271	14.0900	7200	0.8058	0.6885	0.3058
0.6968	14.4814	7400	0.8124	0.6959	0.2979
0.7177	14.8728	7600	0.8107	0.6922	0.3030
0.6778	15.2642	7800	0.8175	0.6917	0.3019
0.6630	15.6556	8000	0.8170	0.6875	0.2984
0.6608	16.0470	8200	0.8236	0.6846	0.2874
0.6240	16.4384	8400	0.8273	0.6889	0.2977
0.6362	16.8297	8600	0.8316	0.6945	0.2972
0.6123	17.2211	8800	0.8350	0.6879	0.2986
0.6093	17.6125	9000	0.8431	0.6897	0.2918
0.5941	18.0039	9200	0.8463	0.6909	0.2936
0.5517	18.3953	9400	0.8530	0.6936	0.2969
0.5874	18.7867	9600	0.8549	0.6922	0.2932
0.5540	19.1781	9800	0.8599	0.6931	0.2973
0.5494	19.5695	10000	0.8643	0.6934	0.2914
0.5468	19.9609	10200	0.8704	0.6931	0.2959
0.5036	20.3523	10400	0.8726	0.6905	0.2917
0.5092	20.7436	10600	0.8795	0.6901	0.2969
0.5206	21.1350	10800	0.8864	0.6961	0.2952
0.4607	21.5264	11000	0.8921	0.6882	0.2931
0.5045	21.9178	11200	0.8933	0.6861	0.2964
0.4640	22.3092	11400	0.9028	0.6943	0.2979
0.4564	22.7006	11600	0.9056	0.6909	0.2989
0.4696	23.0920	11800	0.9098	0.6903	0.2962
0.4310	23.4834	12000	0.9169	0.6892	0.2958
0.4418	23.8748	12200	0.9203	0.6958	0.2927
0.4165	24.2661	12400	0.9301	0.6963	0.2946
0.4102	24.6575	12600	0.9339	0.6949	0.2951
0.4240	25.0489	12800	0.9410	0.6945	0.2985
0.3811	25.4403	13000	0.9438	0.7021	0.2967
0.4038	25.8317	13200	0.9509	0.7104	0.3112
0.3897	26.2231	13400	0.9589	0.7009	0.3081
0.3697	26.6145	13600	0.9573	0.7106	0.3070
0.3685	27.0059	13800	0.9649	0.7032	0.3028
0.3524	27.3973	14000	0.9661	0.7023	0.3015
0.3493	27.7886	14200	0.9752	0.7124	0.3098
0.3395	28.1800	14400	0.9831	0.6992	0.3019
0.3350	28.5714	14600	0.9830	0.7004	0.2989
0.3342	28.9628	14800	0.9889	0.7176	0.3129
0.2998	29.3542	15000	0.9939	0.7100	0.3111
0.3283	29.7456	15200	1.0028	0.7024	0.3003
0.2989	30.1370	15400	1.0067	0.7193	0.3162
0.2975	30.5284	15600	1.0112	0.7049	0.3069
0.2962	30.9198	15800	1.0139	0.7052	0.3061
0.2789	31.3112	16000	1.0209	0.7141	0.3116
0.2888	31.7025	16200	1.0227	0.7089	0.3038
0.2702	32.0939	16400	1.0308	0.7051	0.3071
0.2643	32.4853	16600	1.0341	0.7053	0.3074
0.2603	32.8767	16800	1.0367	0.7061	0.3070
0.2560	33.2681	17000	1.0446	0.7068	0.3039
0.2546	33.6595	17200	1.0480	0.7050	0.3100
0.2461	34.0509	17400	1.0540	0.7039	0.3052
0.2364	34.4423	17600	1.0591	0.7053	0.3083
0.2370	34.8337	17800	1.0655	0.7130	0.3086
0.2207	35.2250	18000	1.0673	0.7053	0.3061
0.2163	35.6164	18200	1.0690	0.7059	0.3065
0.2275	36.0078	18400	1.0721	0.7094	0.3096
0.2090	36.3992	18600	1.0817	0.7068	0.3065
0.2061	36.7906	18800	1.0827	0.7107	0.3114
0.2094	37.1820	19000	1.0882	0.7080	0.3097
0.1921	37.5734	19200	1.0920	0.7095	0.3121
0.1995	37.9648	19400	1.0944	0.7090	0.3092
0.1881	38.3562	19600	1.1006	0.7101	0.3103
0.1831	38.7476	19800	1.1014	0.7083	0.3072
0.1840	39.1389	20000	1.1108	0.7135	0.3115
0.1838	39.5303	20200	1.1122	0.7107	0.3117
0.1756	39.9217	20400	1.1139	0.7103	0.3099
0.1671	40.3131	20600	1.1199	0.7127	0.3079
0.1640	40.7045	20800	1.1239	0.7126	0.3077
0.1721	41.0959	21000	1.1248	0.7171	0.3116
0.1588	41.4873	21200	1.1303	0.7122	0.3087
0.1631	41.8787	21400	1.1334	0.7139	0.3119
0.1560	42.2701	21600	1.1383	0.7169	0.3174
0.1491	42.6614	21800	1.1404	0.7139	0.3178
0.1500	43.0528	22000	1.1434	0.7148	0.3120
0.1419	43.4442	22200	1.1459	0.7138	0.3137
0.1470	43.8356	22400	1.1505	0.7120	0.3132
0.1444	44.2270	22600	1.1541	0.7143	0.3146
0.1369	44.6184	22800	1.1550	0.7137	0.3126
0.1367	45.0098	23000	1.1582	0.7159	0.3124
0.1308	45.4012	23200	1.1626	0.7199	0.3166
0.1303	45.7926	23400	1.1618	0.7211	0.3153
0.1332	46.1840	23600	1.1676	0.7164	0.3153
0.1255	46.5753	23800	1.1701	0.7161	0.3136
0.1246	46.9667	24000	1.1711	0.7154	0.3121
0.1180	47.3581	24200	1.1760	0.7198	0.3175
0.1225	47.7495	24400	1.1767	0.7158	0.3164
0.1218	48.1409	24600	1.1810	0.7180	0.3175
0.1140	48.5323	24800	1.1813	0.7179	0.3170
0.1206	48.9237	25000	1.1860	0.7174	0.3150
0.1106	49.3151	25200	1.1873	0.7184	0.3154
0.1118	49.7065	25400	1.1894	0.7197	0.3179
0.1131	50.0978	25600	1.1919	0.7189	0.3184
0.1085	50.4892	25800	1.1916	0.7173	0.3159
0.1093	50.8806	26000	1.1948	0.7204	0.3183
0.1064	51.2720	26200	1.1974	0.7203	0.3183
0.1024	51.6634	26400	1.1994	0.7185	0.3153
0.1070	52.0548	26600	1.2017	0.7207	0.3178
0.1005	52.4462	26800	1.2023	0.7207	0.3174
0.1036	52.8376	27000	1.2021	0.7195	0.3192
0.1013	53.2290	27200	1.2037	0.7193	0.3184
0.0993	53.6204	27400	1.2078	0.7193	0.3158
0.1011	54.0117	27600	1.2089	0.7178	0.3166
0.0987	54.4031	27800	1.2096	0.7216	0.3167
0.0945	54.7945	28000	1.2112	0.7210	0.3194
0.0989	55.1859	28200	1.2127	0.7217	0.3184
0.0938	55.5773	28400	1.2122	0.7213	0.3225
0.0967	55.9687	28600	1.2142	0.7197	0.3196
0.0940	56.3601	28800	1.2141	0.7175	0.3186
0.0922	56.7515	29000	1.2144	0.7200	0.3191
0.0960	57.1429	29200	1.2161	0.7210	0.3180
0.0917	57.5342	29400	1.2170	0.7205	0.3171
0.0941	57.9256	29600	1.2168	0.7195	0.3171
0.0932	58.3170	29800	1.2180	0.7203	0.3187
0.0901	58.7084	30000	1.2177	0.7204	0.3181
0.0918	59.0998	30200	1.2181	0.7202	0.3163
0.0914	59.4912	30400	1.2185	0.7201	0.3175
0.0924	59.8826	30600	1.2186	0.7197	0.3174

Framework versions

• Transformers 5.3.0.dev0
• Pytorch 2.7.1+cu128
• Datasets 3.6.0
• Tokenizers 0.22.2