everypidigit/whisper-medium-initial

whisper-medium-v1

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

• Loss: 0.2038
• Wer: 0.1553

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-06
• train_batch_size: 4
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1
• training_steps: 156
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.3851	0.01	1	1.3618	0.7022
1.3592	0.03	2	1.3618	0.7022
1.3161	0.04	3	1.3618	0.7022
1.3345	0.06	4	1.1804	0.8059
1.4398	0.07	5	1.1408	0.5451
1.2986	0.08	6	1.1120	0.5328
1.6042	0.1	7	1.0866	0.5552
1.2198	0.11	8	1.0557	0.5451
0.9403	0.13	9	1.0209	0.5282
0.8978	0.14	10	0.9880	0.4724
1.6951	0.15	11	0.9619	0.4204
0.8423	0.17	12	0.9425	0.4132
1.3482	0.18	13	0.9270	0.4155
0.8783	0.2	14	0.9154	0.3754
0.9994	0.21	15	0.9050	0.3399
0.9805	0.23	16	0.8947	0.3368
0.9782	0.24	17	0.8844	0.3361
1.0236	0.25	18	0.8750	0.3345
0.8005	0.27	19	0.8647	0.3335
0.7419	0.28	20	0.8547	0.3322
1.0348	0.3	21	0.8451	0.3348
0.9592	0.31	22	0.8361	0.3322
1.23	0.32	23	0.8287	0.3330
0.8397	0.34	24	0.8225	0.3312
0.9839	0.35	25	0.8171	0.3296
0.862	0.37	26	0.8118	0.3286
0.6045	0.38	27	0.8071	0.3281
0.7948	0.39	28	0.8025	0.3281
0.6693	0.41	29	0.7977	0.3692
0.7191	0.42	30	0.7930	0.3680
1.0364	0.44	31	0.7888	0.3268
0.7458	0.45	32	0.7852	0.3286
0.6369	0.46	33	0.7817	0.3291
0.8618	0.48	34	0.7781	0.3291
0.7303	0.49	35	0.7747	0.3291
0.7667	0.51	36	0.7713	0.3268
0.8733	0.52	37	0.7680	0.3268
0.7262	0.54	38	0.7650	0.3227
0.746	0.55	39	0.7623	0.3194
0.7106	0.56	40	0.7598	0.3204
0.6862	0.58	41	0.7576	0.3209
1.2068	0.59	42	0.7555	0.3209
0.8285	0.61	43	0.7537	0.3206
0.9254	0.62	44	0.7521	0.3186
0.7495	0.63	45	0.7508	0.3188
0.7105	0.65	46	0.7496	0.3170
0.731	0.66	47	0.7486	0.3152
0.8015	0.68	48	0.7478	0.3152
0.931	0.69	49	0.7471	0.3152
0.6324	0.7	50	0.7466	0.3155
0.5111	0.72	51	0.7145	0.2716
0.5398	0.73	52	0.7013	0.2663
0.8491	0.75	53	0.7013	0.2663
0.6479	0.76	54	0.6906	0.2644
0.5161	0.77	55	0.6815	0.2608
0.7262	0.79	56	0.6739	0.2633
0.606	0.8	57	0.6665	0.2644
0.773	0.82	58	0.6594	0.2650
0.6815	0.83	59	0.6518	0.2730
0.5886	0.85	60	0.6436	0.2708
0.5243	0.86	61	0.6436	0.2708
0.4213	0.87	62	0.6367	0.2733
0.5213	0.89	63	0.6319	0.2760
0.8247	0.9	64	0.6277	0.2785
0.7444	0.92	65	0.6230	0.2816
0.7056	0.93	66	0.6171	0.2802
0.667	0.94	67	0.6105	0.2785
0.3999	0.96	68	0.6029	0.2749
0.6285	0.97	69	0.5966	0.2757
0.537	0.99	70	0.5917	0.2752
0.3534	1.0	71	0.5912	0.2699
0.3613	1.01	72	0.5918	0.2633
0.5215	1.03	73	0.5925	0.2666
0.8283	1.04	74	0.5910	0.2661
0.5645	1.06	75	0.5872	0.2669
0.3813	1.07	76	0.5814	0.2655
0.5916	1.08	77	0.5750	0.2647
0.5135	1.1	78	0.5690	0.2611
0.5852	1.11	79	0.5640	0.2619
0.3307	1.13	80	0.5610	0.2575
0.6206	1.14	81	0.5588	0.2558
0.4794	1.15	82	0.5562	0.2555
0.4213	1.17	83	0.5536	0.2564
0.5264	1.18	84	0.5506	0.2594
0.3956	1.2	85	0.5474	0.2583
0.3321	1.21	86	0.5448	0.2580
0.3627	1.23	87	0.5422	0.2591
0.3872	1.24	88	0.5400	0.2600
0.4919	1.25	89	0.5369	0.2616
0.4247	1.27	90	0.5338	0.2605
0.7103	1.28	91	0.5297	0.2586
0.3547	1.3	92	0.5260	0.2586
0.2676	1.31	93	0.5216	0.2580
0.8941	1.32	94	0.5169	0.2594
0.4403	1.34	95	0.5117	0.2594
0.3067	1.35	96	0.5070	0.2602
0.2869	1.37	97	0.5024	0.2619
0.4016	1.38	98	0.4976	0.2627
0.4335	1.39	99	0.4934	0.2644
0.5381	1.41	100	0.4877	0.2644
0.3677	1.42	101	0.4808	0.2583
0.499	1.44	102	0.4731	0.2583
0.5656	1.45	103	0.4663	0.2575
0.3133	1.46	104	0.4607	0.2572
0.6015	1.48	105	0.4557	0.2572
0.4748	1.49	106	0.4515	0.2564
0.4336	1.51	107	0.4479	0.2564
0.433	1.52	108	0.4443	0.2550
0.2138	1.54	109	0.4409	0.2533
0.221	1.55	110	0.4387	0.2542
0.3434	1.56	111	0.4379	0.2536
0.3193	1.58	112	0.4414	0.2508
0.3732	1.59	113	0.4448	0.2486
0.1977	1.61	114	0.4442	0.2494
0.2772	1.62	115	0.4401	0.2481
0.3893	1.63	116	0.4352	0.2478
0.3424	1.65	117	0.4311	0.2458
0.5052	1.66	118	0.4285	0.2456
0.5281	1.68	119	0.4266	0.2464
0.248	1.69	120	0.4256	0.2447
0.2911	1.7	121	0.4249	0.2442
0.1723	1.72	122	0.4247	0.2434
0.1987	1.73	123	0.4247	0.2475
0.4155	1.75	124	0.4249	0.2456
0.3602	1.76	125	0.4247	0.2450
0.4435	1.77	126	0.4243	0.2467
0.1951	1.79	127	0.4242	0.2472
0.2449	1.8	128	0.4242	0.2478
0.3351	1.82	129	0.4240	0.2530
0.3099	1.83	130	0.4240	0.2550
0.3729	1.85	131	0.4241	0.2542
0.5431	1.86	132	0.4237	0.2539
0.2896	1.87	133	0.4233	0.2544
0.2678	1.89	134	0.4227	0.2544
0.4447	1.9	135	0.4218	0.2519
0.3563	1.92	136	0.4208	0.2486
0.2154	1.93	137	0.4201	0.2486
0.4183	1.94	138	0.4193	0.2453
0.5911	1.96	139	0.4185	0.2456
0.1677	1.97	140	0.4180	0.2486
0.2831	1.99	141	0.4177	0.2470
0.502	2.0	142	0.4178	0.2494
0.2007	2.01	143	0.4177	0.2492
0.0994	2.03	144	0.4183	0.2539
0.0823	2.04	145	0.4192	0.2547
0.217	2.06	146	0.4199	0.2569
0.224	2.07	147	0.4206	0.2555
0.2767	2.08	148	0.4208	0.2542
0.1879	2.1	149	0.4203	0.2528
0.1974	2.11	150	0.4199	0.2578
0.1788	2.13	151	0.4196	0.2558
0.1505	2.14	152	0.4191	0.2483
0.1873	2.15	153	0.2135	0.1586
0.1603	2.17	154	0.2125	0.1591
0.2388	2.18	155	0.2040	0.1556
0.2209	2.2	156	0.2038	0.1553

Framework versions

• Transformers 4.39.2
• Pytorch 2.2.1
• Datasets 2.18.0
• Tokenizers 0.15.1