Moroccan-Darija-STT-large-v1.5.2

This model is a fine-tuned version of openai/whisper-large-v3 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3054
• Wer: 117.4489
• Cer: 98.8307

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-07
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• 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: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 60

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
2.8823	0.6135	100	0.8263	92.5391	49.8173
2.81	1.2270	200	0.7744	92.9001	49.7929
2.6462	1.8405	300	0.6980	93.1408	49.4032
2.46	2.4540	400	0.6228	91.3357	49.5250
2.3255	3.0675	500	0.5622	91.3357	50.4750
2.1054	3.6810	600	0.5389	94.9458	57.1985
2.0243	4.2945	700	0.5230	95.6679	60.2436
1.9601	4.9080	800	0.5091	95.5475	63.3374
1.8677	5.5215	900	0.4955	101.0830	74.3727
1.793	6.1350	1000	0.4824	109.6270	80.2680
1.7006	6.7485	1100	0.4703	105.0542	84.5311
1.6489	7.3620	1200	0.4552	106.2575	87.8197
1.6782	7.9755	1300	0.4509	141.3959	119.3666
1.6137	8.5890	1400	0.4477	138.7485	118.9769
1.5515	9.2025	1500	0.4256	162.6955	137.9537
1.5284	9.8160	1600	0.4199	169.3141	140.8039
1.4676	10.4294	1700	0.4220	145.1264	120.5116
1.4378	11.0429	1800	0.4228	142.7196	120.9013
1.3755	11.6564	1900	0.4007	133.6943	112.8136
1.3006	12.2699	2000	0.3947	111.0710	93.3496
1.289	12.8834	2100	0.3747	110.8303	93.2034
1.3037	13.4969	2200	0.3641	121.5403	105.1888
1.2583	14.1104	2300	0.3621	111.7930	93.8611
1.2245	14.7239	2400	0.3598	135.6197	115.3228
1.1955	15.3374	2500	0.3530	122.1420	105.3350
1.2061	15.9509	2600	0.3466	121.4200	103.3130
1.1644	16.5644	2700	0.3409	119.7353	100.3898
1.1885	17.1779	2800	0.3421	121.0590	104.4336
1.115	17.7914	2900	0.3356	119.3742	102.8258
1.1449	18.4049	3000	0.3299	109.0253	89.6955
1.1613	19.0184	3100	0.3255	116.4862	96.3459
1.1437	19.6319	3200	0.3256	116.0048	95.2741
1.0599	20.2454	3300	0.3169	105.6558	86.2119
1.0489	20.8589	3400	0.3173	106.0168	85.7978
1.0779	21.4724	3500	0.3145	104.5728	86.9184
1.0808	22.0859	3600	0.3137	104.3321	86.6261
1.0569	22.6994	3700	0.3134	103.0084	86.0414
1.0907	23.3129	3800	0.3120	105.2948	87.5761
1.092	23.9264	3900	0.3075	103.7304	86.2850
1.012	24.5399	4000	0.3081	105.1745	87.9659
1.0455	25.1534	4100	0.3042	111.0710	89.8417
0.993	25.7669	4200	0.3100	111.1913	89.9147
1.0574	26.3804	4300	0.3007	111.0710	88.7698
0.9836	26.9939	4400	0.3042	109.9880	87.7467
1.0063	27.6074	4500	0.3054	102.1661	86.1389
1.0172	28.2209	4600	0.3042	102.1661	85.9196
0.9869	28.8344	4700	0.3047	110.8303	88.6480
1.0173	29.4479	4800	0.3042	111.0710	88.6967
1.007	30.0613	4900	0.3039	102.4067	85.9440
0.9428	30.6748	5000	0.3052	102.4067	86.5043
0.9475	31.2883	5100	0.3023	102.8881	86.3094
0.9538	31.9018	5200	0.3034	121.7810	103.7028
0.9769	32.5153	5300	0.3049	131.0469	105.9927
0.9242	33.1288	5400	0.3026	111.4320	88.2582
0.927	33.7423	5500	0.3049	127.1961	101.1206
0.9576	34.3558	5600	0.3030	122.5030	103.8246
0.9709	34.9693	5700	0.3041	122.8640	104.0926
0.9365	35.5828	5800	0.3048	122.8640	104.4823
0.9723	36.1963	5900	0.3029	122.0217	103.7759
0.9383	36.8098	6000	0.3040	122.6233	104.3118
0.948	37.4233	6100	0.3027	138.9892	117.2716
0.8709	38.0368	6200	0.3026	139.1095	117.5883
0.965	38.6503	6300	0.3033	122.6233	104.2875
0.9689	39.2638	6400	0.3036	138.9892	117.6370
0.9515	39.8773	6500	0.3044	139.3502	117.5639
0.8778	40.4908	6600	0.3042	121.9013	103.9951
0.9073	41.1043	6700	0.3057	139.4705	117.5883
0.8688	41.7178	6800	0.3047	118.7726	99.3910
0.9335	42.3313	6900	0.3047	139.2298	118.1242
0.9439	42.9448	7000	0.3037	138.6282	117.9050
0.8685	43.5583	7100	0.3042	117.0878	98.5627
0.9241	44.1718	7200	0.3049	117.9302	98.7820
0.9078	44.7853	7300	0.3041	117.5692	98.7333
0.8858	45.3988	7400	0.3064	117.0878	98.5140
0.871	46.0123	7500	0.3048	117.8099	99.1717
0.9462	46.6258	7600	0.3046	118.5319	99.3666
0.9373	47.2393	7700	0.3046	117.2082	98.6358
0.9336	47.8528	7800	0.3044	118.0505	99.2205
0.8448	48.4663	7900	0.3061	117.4489	99.1474
0.8969	49.0798	8000	0.3053	117.5692	98.8794
0.8706	49.6933	8100	0.3051	118.0505	98.8794
0.9211	50.3067	8200	0.3053	117.6895	98.8794
0.9141	50.9202	8300	0.3049	116.9675	98.0755
0.8708	51.5337	8400	0.3059	117.4489	99.0743
0.9129	52.1472	8500	0.3057	117.9302	99.1717
0.8726	52.7607	8600	0.3052	117.4489	98.2217
0.8889	53.3742	8700	0.3050	118.4116	98.9525
0.9593	53.9877	8800	0.3053	118.2912	98.8063
0.9276	54.6012	8900	0.3057	116.8472	97.9050
0.8554	55.2147	9000	0.3055	117.5692	98.7820
0.8811	55.8282	9100	0.3055	118.1709	98.5384
0.9238	56.4417	9200	0.3051	116.7268	97.8319
0.8927	57.0552	9300	0.3055	117.3285	98.0755
0.8539	57.6687	9400	0.3054	117.3285	98.0512
0.9293	58.2822	9500	0.3055	117.9302	98.7820
0.9069	58.8957	9600	0.3055	117.0878	98.5140
0.8801	59.5092	9700	0.3054	117.4489	98.8307

Framework versions

• Transformers 4.48.0.dev0
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.21.0