JunWorks/whisper-small-LIDmetric-4060-2lang-properLORAsettings

Whisper Small 2 lang LORA 2nd Settings

This model is a fine-tuned version of openai/whisper-small on the Common Voice 13.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2569

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: 4
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 20000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.9564	0.1002	250	2.1385
1.2046	0.2004	500	1.1713
0.8719	0.3006	750	0.9414
0.7799	0.4008	1000	0.8261
0.6956	0.5010	1250	0.6536
0.3536	0.6012	1500	0.2990
0.3167	0.7014	1750	0.2902
0.3252	0.8016	2000	0.2850
0.3133	0.9018	2250	0.2821
0.3764	1.0020	2500	0.2800
0.2942	1.1022	2750	0.2775
0.2937	1.2024	3000	0.2770
0.3478	1.3026	3250	0.2745
0.3776	1.4028	3500	0.2729
0.3099	1.5030	3750	0.2713
0.3087	1.6032	4000	0.2705
0.2998	1.7034	4250	0.2699
0.3226	1.8036	4500	0.2683
0.3589	1.9038	4750	0.2676
0.3411	2.0040	5000	0.2673
0.3084	2.1042	5250	0.2674
0.31	2.2044	5500	0.2663
0.3388	2.3046	5750	0.2657
0.2716	2.4048	6000	0.2652
0.3059	2.5050	6250	0.2652
0.27	2.6052	6500	0.2648
0.2954	2.7054	6750	0.2639
0.336	2.8056	7000	0.2641
0.2833	2.9058	7250	0.2631
0.2777	3.0060	7500	0.2624
0.2418	3.1062	7750	0.2618
0.3194	3.2064	8000	0.2623
0.3319	3.3066	8250	0.2623
0.3551	3.4068	8500	0.2615
0.3421	3.5070	8750	0.2619
0.3862	3.6072	9000	0.2616
0.2437	3.7074	9250	0.2609
0.2995	3.8076	9500	0.2604
0.3535	3.9078	9750	0.2603
0.2871	4.0080	10000	0.2601
0.2908	4.1082	10250	0.2604
0.3203	4.2084	10500	0.2599
0.2598	4.3086	10750	0.2594
0.2942	4.4088	11000	0.2593
0.3302	4.5090	11250	0.2590
0.3615	4.6092	11500	0.2584
0.3291	4.7094	11750	0.2582
0.2781	4.8096	12000	0.2588
0.3106	4.9098	12250	0.2585
0.2484	5.0100	12500	0.2583
0.2645	5.1102	12750	0.2583
0.3034	5.2104	13000	0.2581
0.2865	5.3106	13250	0.2576
0.3301	5.4108	13500	0.2580
0.3759	5.5110	13750	0.2579
0.3318	5.6112	14000	0.2581
0.2825	5.7114	14250	0.2579
0.2976	5.8116	14500	0.2578
0.2976	5.9118	14750	0.2577
0.3681	6.0120	15000	0.2575
0.3274	6.1122	15250	0.2575
0.2948	6.2124	15500	0.2577
0.2932	6.3126	15750	0.2576
0.2587	6.4128	16000	0.2578
0.2564	6.5130	16250	0.2573
0.2776	6.6132	16500	0.2569
0.2954	6.7134	16750	0.2569
0.2891	6.8136	17000	0.2568
0.2373	6.9138	17250	0.2569
0.3532	7.0140	17500	0.2569
0.2676	7.1142	17750	0.2569
0.2763	7.2144	18000	0.2569
0.2692	7.3146	18250	0.2571
0.3198	7.4148	18500	0.2570
0.2158	7.5150	18750	0.2571
0.277	7.6152	19000	0.2572
0.2308	7.7154	19250	0.2572
0.3166	7.8156	19500	0.2569
0.3064	7.9158	19750	0.2570
0.2743	8.0160	20000	0.2569

Framework versions

• PEFT 0.11.2.dev0
• Transformers 4.43.0.dev0
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1