metadata

license: apache-2.0
base_model: openai/whisper-small
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: Teamim-small_WeightDecay-0.05_Augmented_New-Data_date-19-07-2024_15-41
    results: []

Teamim-small_WeightDecay-0.05_Augmented_New-Data_date-19-07-2024_15-41

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

Loss: 0.2357
Wer: 12.4316
Avg Precision Exact: 0.8908
Avg Recall Exact: 0.8920
Avg F1 Exact: 0.8911
Avg Precision Letter Shift: 0.9098
Avg Recall Letter Shift: 0.9110
Avg F1 Letter Shift: 0.9100
Avg Precision Word Level: 0.9122
Avg Recall Word Level: 0.9134
Avg F1 Word Level: 0.9125
Avg Precision Word Shift: 0.9710
Avg Recall Word Shift: 0.9730
Avg F1 Word Shift: 0.9716
Precision Median Exact: 1.0
Recall Median Exact: 1.0
F1 Median Exact: 1.0
Precision Max Exact: 1.0
Recall Max Exact: 1.0
F1 Max Exact: 1.0
Precision Min Exact: 0.0
Recall Min Exact: 0.0
F1 Min Exact: 0.0
Precision Min Letter Shift: 0.0
Recall Min Letter Shift: 0.0
F1 Min Letter Shift: 0.0
Precision Min Word Level: 0.0
Recall Min Word Level: 0.0
F1 Min Word Level: 0.0
Precision Min Word Shift: 0.1429
Recall Min Word Shift: 0.1111
F1 Min Word Shift: 0.125

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: 8
eval_batch_size: 32
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
training_steps: 200000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Avg Precision Exact	Avg Recall Exact	Avg F1 Exact	Avg Precision Letter Shift	Avg Recall Letter Shift	Avg F1 Letter Shift	Avg Precision Word Level	Avg Recall Word Level	Avg F1 Word Level	Avg Precision Word Shift	Avg Recall Word Shift	Avg F1 Word Shift	Precision Median Exact	Recall Median Exact	F1 Median Exact	Precision Max Exact	Recall Max Exact	F1 Max Exact	Precision Min Word Shift	Recall Min Word Shift	F1 Min Word Shift
No log	0.0001	1	8.6313	200.3067	0.0004	0.0014	0.0004	0.0046	0.0050	0.0041	0.0031	0.0092	0.0033	0.0326	0.0360	0.0303	0.0	0.0	0.0	0.1111	1.0	0.2000	0.0	0.0	0.0
0.0158	0.9410	10000	0.1501	17.4538	0.8417	0.8453	0.8428	0.8645	0.8682	0.8656	0.8682	0.8723	0.8695	0.9449	0.9497	0.9463	0.9231	0.9231	0.9231	1.0	1.0	1.0	0.0909	0.0909	0.1111
0.0149	1.8820	20000	0.1635	15.7724	0.8627	0.8659	0.8639	0.8851	0.8885	0.8864	0.8881	0.8915	0.8894	0.9560	0.9594	0.9571	0.9286	0.9333	0.9412	1.0	1.0	1.0	0.0	0.0	0.0
0.0053	2.8230	30000	0.1755	14.9261	0.8767	0.8763	0.8761	0.8990	0.8986	0.8983	0.9016	0.9013	0.9010	0.9609	0.9621	0.9610	0.9333	0.9375	0.9474	1.0	1.0	1.0	0.1429	0.125	0.1333
0.0034	3.7640	40000	0.1817	15.5691	0.8584	0.8592	0.8584	0.8794	0.8804	0.8794	0.8823	0.8833	0.8824	0.9538	0.9566	0.9547	0.9333	0.9333	0.9444	1.0	1.0	1.0	0.0	0.0	0.0
0.0005	4.7050	50000	0.1857	14.4050	0.8793	0.8793	0.8789	0.8998	0.8999	0.8994	0.9024	0.9028	0.9022	0.9629	0.9637	0.9628	0.9474	1.0	0.9565	1.0	1.0	1.0	0.1429	0.1	0.1176
0.0021	5.6460	60000	0.1945	14.5196	0.8716	0.8762	0.8734	0.8930	0.8978	0.8949	0.8960	0.9006	0.8979	0.9607	0.9659	0.9628	0.9333	1.0	0.9565	1.0	1.0	1.0	0.0769	0.1	0.0952
0.0016	6.5870	70000	0.1987	14.0872	0.8773	0.8781	0.8772	0.8985	0.8993	0.8984	0.9015	0.9025	0.9015	0.9641	0.9661	0.9646	0.9412	1.0	0.9600	1.0	1.0	1.0	0.1429	0.1111	0.125
0.0005	7.5280	80000	0.2020	13.7288	0.8801	0.8825	0.8809	0.9015	0.9041	0.9024	0.9044	0.9069	0.9052	0.9673	0.9709	0.9687	1.0	1.0	0.9630	1.0	1.0	1.0	0.1429	0.1	0.1176
0.0021	8.4690	90000	0.2088	14.3829	0.8767	0.8782	0.8770	0.8980	0.8997	0.8984	0.9009	0.9026	0.9013	0.9629	0.9664	0.9640	0.9412	1.0	0.9600	1.0	1.0	1.0	0.0	0.0	0.0
0.0013	9.4100	100000	0.2041	13.4146	0.8826	0.8837	0.8828	0.9025	0.9037	0.9027	0.9052	0.9063	0.9054	0.9681	0.9697	0.9684	1.0	1.0	0.9630	1.0	1.0	1.0	0.1429	0.1111	0.125
0.0002	10.3510	110000	0.2110	13.5329	0.8876	0.8891	0.8880	0.9086	0.9101	0.9089	0.9112	0.9127	0.9115	0.9681	0.9710	0.9690	1.0	1.0	0.9630	1.0	1.0	1.0	0.1429	0.1	0.1176
0.0003	11.2920	120000	0.2123	13.5033	0.8822	0.8828	0.8821	0.9023	0.9029	0.9022	0.9048	0.9056	0.9048	0.9679	0.9698	0.9684	1.0	1.0	0.9677	1.0	1.0	1.0	0.1429	0.1	0.1176
0.0011	12.2330	130000	0.2082	13.5070	0.8884	0.8881	0.8878	0.9085	0.9083	0.9080	0.9109	0.9107	0.9104	0.9677	0.9683	0.9675	1.0	1.0	0.9677	1.0	1.0	1.0	0.0769	0.0769	0.0769
0.0	13.1740	140000	0.2181	13.4220	0.8835	0.8860	0.8844	0.9028	0.9054	0.9038	0.9056	0.9082	0.9065	0.9671	0.9706	0.9684	1.0	1.0	0.9677	1.0	1.0	1.0	0.1429	0.1111	0.125
0.0	14.1150	150000	0.2240	13.0266	0.8847	0.8860	0.8850	0.9043	0.9057	0.9046	0.9073	0.9086	0.9075	0.9692	0.9719	0.9701	1.0	1.0	0.9677	1.0	1.0	1.0	0.1429	0.1111	0.125
0.0002	15.0560	160000	0.2245	13.0155	0.8828	0.8831	0.8826	0.9019	0.9022	0.9017	0.9049	0.9053	0.9048	0.9701	0.9717	0.9704	1.0	1.0	0.9687	1.0	1.0	1.0	0.1429	0.1111	0.125
0.0001	15.9970	170000	0.2272	13.0155	0.8855	0.8861	0.8854	0.9043	0.9049	0.9042	0.9070	0.9078	0.9070	0.9693	0.9710	0.9697	1.0	1.0	0.9677	1.0	1.0	1.0	0.1429	0.1111	0.125
0.0	16.9380	180000	0.2312	12.6053	0.8875	0.8892	0.8880	0.9062	0.9079	0.9067	0.9089	0.9106	0.9093	0.9703	0.9728	0.9711	1.0	1.0	1.0	1.0	1.0	1.0	0.1429	0.1111	0.125
0.0	17.8790	190000	0.2324	12.5684	0.8898	0.8910	0.8900	0.9089	0.9102	0.9092	0.9113	0.9127	0.9116	0.9703	0.9726	0.9710	1.0	1.0	1.0	1.0	1.0	1.0	0.1429	0.1111	0.125
0.0	18.8200	200000	0.2357	12.4316	0.8908	0.8920	0.8911	0.9098	0.9110	0.9100	0.9122	0.9134	0.9125	0.9710	0.9730	0.9716	1.0	1.0	1.0	1.0	1.0	1.0	0.1429	0.1111	0.125

Framework versions

Transformers 4.41.2
Pytorch 2.2.1
Datasets 2.20.0
Tokenizers 0.19.1