whisper-small-si-bank-v5

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: 0.6785
• Wer Ortho: 65.2174
• Wer: 54.6988

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: 1e-05
• 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: constant_with_warmup
• lr_scheduler_warmup_steps: 50
• training_steps: 250
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
No log	0.8	2	2.3673	183.6957	250.3614
No log	1.4	4	2.3670	173.9130	250.1205
No log	2.0	6	2.3480	173.3696	254.6988
No log	2.8	8	2.2897	184.2391	254.2169
No log	3.4	10	2.1773	179.8913	255.1807
No log	4.0	12	2.0397	202.7174	240.0
No log	4.8	14	1.9600	225.0	236.6265
No log	5.4	16	1.8787	227.1739	220.9639
No log	6.0	18	1.7866	264.1304	243.1325
No log	6.8	20	1.7069	217.3913	232.5301
No log	7.4	22	1.6312	167.3913	225.5422
No log	8.0	24	1.5601	157.6087	230.1205
1.679	8.8	26	1.5088	166.8478	243.8554
1.679	9.4	28	1.4717	178.8043	226.9880
1.679	10.0	30	1.4356	153.8043	203.6145
1.679	10.8	32	1.4001	147.8261	172.7711
1.679	11.4	34	1.3626	132.6087	164.0964
1.679	12.0	36	1.3208	104.8913	114.4578
1.679	12.8	38	1.2845	117.9348	128.4337
1.679	13.4	40	1.2408	104.3478	127.9518
1.679	14.0	42	1.2081	94.0217	113.9759
1.679	14.8	44	1.1600	94.5652	108.9157
1.679	15.4	46	1.1251	94.0217	93.2530
1.679	16.0	48	1.0849	90.7609	86.9880
0.8854	16.8	50	1.0486	85.3261	84.0964
0.8854	17.4	52	1.0228	85.3261	78.7952
0.8854	18.0	54	0.9948	90.7609	85.5422
0.8854	18.8	56	0.9657	83.1522	75.6627
0.8854	19.4	58	0.9494	82.6087	70.3614
0.8854	20.0	60	0.9217	82.0652	95.4217
0.8854	20.8	62	0.9007	83.6957	69.8795
0.8854	21.4	64	0.9024	80.4348	84.5783
0.8854	22.0	66	0.8837	78.2609	66.5060
0.8854	22.8	68	0.8545	78.2609	69.6386
0.8854	23.4	70	0.8554	77.7174	66.2651
0.8854	24.0	72	0.8362	79.8913	65.5422
0.8854	24.8	74	0.8236	78.2609	77.5904
0.2938	25.4	76	0.8270	79.3478	64.0964
0.2938	26.0	78	0.8338	73.3696	62.4096
0.2938	26.8	80	0.8291	77.7174	66.5060
0.2938	27.4	82	0.8466	76.0870	60.7229
0.2938	28.0	84	0.8509	75.5435	60.2410
0.2938	28.8	86	0.8416	76.6304	64.8193
0.2938	29.4	88	0.8729	81.5217	79.5181
0.2938	30.0	90	0.8373	72.2826	55.4217
0.2938	30.8	92	0.8513	73.3696	62.8916
0.2938	31.4	94	0.8573	80.9783	67.2289
0.2938	32.0	96	0.8631	73.9130	54.4578
0.2938	32.8	98	0.8339	76.0870	66.7470
0.112	33.4	100	0.8490	73.3696	60.9639
0.112	34.0	102	0.8454	71.1957	57.5904
0.112	34.8	104	0.8534	73.3696	59.7590
0.112	35.4	106	0.8456	72.8261	57.5904
0.112	36.0	108	0.8541	77.1739	61.6867
0.112	36.8	110	0.8482	75.5435	57.1084
0.112	37.4	112	0.8513	77.7174	63.3735
0.112	38.0	114	0.8452	76.6304	57.8313
0.112	38.8	116	0.8438	74.4565	56.8675
0.112	39.4	118	0.8620	79.8913	65.7831
0.112	40.0	120	0.8870	74.4565	60.7229
0.112	40.8	122	0.8372	74.4565	57.5904
0.112	41.4	124	0.8063	77.1739	61.9277
0.0736	42.0	126	0.8106	77.7174	56.1446
0.0736	42.8	128	0.7557	73.9130	58.3133
0.0736	43.4	130	0.7636	80.4348	70.3614
0.0736	44.0	132	0.6653	73.9130	56.3855
0.0736	44.8	134	0.6808	76.6304	55.9036
0.0736	45.4	136	0.6913	72.2826	54.2169
0.0736	46.0	138	0.6848	73.3696	55.1807
0.0736	46.8	140	0.7092	72.8261	55.6627
0.0736	47.4	142	0.6734	69.5652	51.8072
0.0736	48.0	144	0.6881	70.6522	52.5301
0.0736	48.8	146	0.7022	73.9130	63.8554
0.0736	49.4	148	0.6888	77.7174	57.8313
0.0155	50.0	150	0.6622	67.3913	49.1566
0.0155	50.8	152	0.6599	71.1957	59.5181
0.0155	51.4	154	0.6921	72.8261	64.0964
0.0155	52.0	156	0.6805	66.8478	54.2169
0.0155	52.8	158	0.6917	66.3043	51.8072
0.0155	53.4	160	0.7486	72.8261	57.3494
0.0155	54.0	162	0.6846	71.1957	54.9398
0.0155	54.8	164	0.6845	70.1087	49.1566
0.0155	55.4	166	0.6888	67.3913	54.9398
0.0155	56.0	168	0.6847	65.7609	53.4940
0.0155	56.8	170	0.6819	68.4783	55.1807
0.0155	57.4	172	0.6835	65.2174	53.7349
0.0155	58.0	174	0.6706	63.0435	50.6024
0.0025	58.8	176	0.6732	61.4130	44.3373
0.0025	59.4	178	0.6852	66.8478	49.8795
0.0025	60.0	180	0.6781	63.5870	43.3735
0.0025	60.8	182	0.6678	64.1304	46.0241
0.0025	61.4	184	0.6624	60.8696	44.5783
0.0025	62.0	186	0.6671	60.8696	47.4699
0.0025	62.8	188	0.6773	61.9565	47.2289
0.0025	63.4	190	0.6970	63.0435	48.6747
0.0025	64.0	192	0.7228	68.4783	56.8675
0.0025	64.8	194	0.7653	68.4783	57.3494
0.0025	65.4	196	0.6913	62.5	46.0241
0.0025	66.0	198	0.7189	66.3043	46.0241
0.0013	66.8	200	0.6960	61.4130	47.4699
0.0013	67.4	202	0.7476	70.1087	53.7349
0.0013	68.0	204	0.7047	70.1087	51.0843
0.0013	68.8	206	0.6862	64.1304	48.4337
0.0013	69.4	208	0.6858	62.5	46.2651
0.0013	70.0	210	0.6685	64.6739	46.9880
0.0013	70.8	212	0.6682	65.7609	48.9157
0.0013	71.4	214	0.6710	61.4130	45.0602
0.0013	72.0	216	0.7055	63.5870	47.4699
0.0013	72.8	218	0.7346	69.0217	55.6627
0.0013	73.4	220	0.7085	63.0435	49.1566
0.0013	74.0	222	0.7040	64.1304	48.6747
0.0013	74.8	224	0.6926	65.2174	49.6386
0.0022	75.4	226	0.6767	66.3043	50.1205
0.0022	76.0	228	0.6654	70.1087	51.0843
0.0022	76.8	230	0.6989	69.0217	52.2892
0.0022	77.4	232	0.6957	67.9348	51.3253
0.0022	78.0	234	0.6579	62.5	43.6145
0.0022	78.8	236	0.6690	65.2174	50.8434
0.0022	79.4	238	0.6625	66.3043	47.4699
0.0022	80.0	240	0.6495	66.8478	45.7831
0.0022	80.8	242	0.6464	61.9565	43.1325
0.0022	81.4	244	0.6555	63.5870	47.9518
0.0022	82.0	246	0.6533	59.7826	40.4819
0.0022	82.8	248	0.6579	60.3261	44.3373
0.0021	83.4	250	0.6785	65.2174	54.6988

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0