Add CER, hyperparameters, and training logs to README

9369c80 verified about 1 month ago

11.9 kB

language:
  - lus
license: apache-2.0
base_model: Qwen/Qwen2-Audio-7B
tags:
  - generated_from_trainer
datasets:
  - andrewbawitlung/mizonal-v3
metrics:
  - wer
  - cer
model-index:
  - name: qwen3-asr-0.6b-mizonal3-E3-lus-v2026.06
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: MiZonal v3
          type: andrewbawitlung/mizonal-v3
          config: default
          split: test
        metrics:
          - name: Wer
            type: wer
            value: 18.985
          - name: Cer
            type: cer
            value: 4.217

qwen3-asr-0.6b-mizonal3-E3-lus-v2026.06

This model is a fine-tuned version of Qwen/Qwen2-Audio-7B on the MiZonal v3 dataset. Note: ~1 hour of conversational speech was added to this dataset version.

It achieves the following results on the evaluation set:

Wer: 18.9850
Cer: 4.2170

Model description

Experiment Configurations

This repository is part of a series of experiments. The different configurations are:

E1 (Baseline): Standard training configuration.
E2 (Noise): Training with background noise augmentation.
E3 (Speed): Training with speed perturbation augmentation.
E4 (SpecAug): Training with SpecAugment (time and frequency masking).
E5 (Combined): Training with a combination of all augmentations.

All Models in this Family

Experiment	Hugging Face Repository
E1 (Baseline)	andrewbawitlung/qwen3-asr-0.6b-mizonal3-E1-lus-v2026.06
E2 (Noise)	andrewbawitlung/qwen3-asr-0.6b-mizonal3-E2-lus-v2026.06
E3 (Speed)	andrewbawitlung/qwen3-asr-0.6b-mizonal3-E3-lus-v2026.06
E4 (SpecAug)	andrewbawitlung/qwen3-asr-0.6b-mizonal3-E4-lus-v2026.06
E5 (Combined)	andrewbawitlung/qwen3-asr-0.6b-mizonal3-E5-lus-v2026.06

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

More information needed

Training results

step	epoch	train_loss	eval_loss	eval_wer	eval_cer	learning_rate	grad_norm
200	0.12143290831815422	0.7348	1.0950071811676025	0.6606986474652136	0.21059513712185468	1.5075757575757577e-05	33.0
400	0.24286581663630843	0.2043	0.4572545289993286	0.3625571664882748	0.09278696635566865	1.979089219330855e-05	9.0
600	0.36429872495446264	0.1685	0.3414927124977112	0.2738153157536246	0.06606940910376025	1.948110285006196e-05	8.5
800	0.48573163327261687	0.1129	0.3106901943683624	0.24909993188673737	0.05797639242295731	1.9171313506815367e-05	5.5
1000	0.607164541590771	0.1239	0.28289029002189636	0.22876325776004672	0.05285199321458863	1.8861524163568775e-05	4.625
1200	0.7285974499089253	0.0943	0.26755592226982117	0.21767052641821544	0.052304212609556124	1.8551734820322184e-05	5.0
1400	0.8500303582270795	0.074	0.2625117599964142	0.21124841879926048	0.04889383658467628	1.8241945477075592e-05	4.53125
1600	0.9714632665452337	0.0765	0.25841450691223145	0.2107618954947942	0.04935326547921968	1.7932156133828997e-05	3.5625
1800	1.092896174863388	0.0375	0.2680838108062744	0.20725892770263696	0.04719748374328527	1.7622366790582406e-05	4.09375
2000	1.2143290831815423	0.0411	0.2577139735221863	0.19937725017028315	0.04564249363867685	1.731257744733581e-05	1.9609375
2200	1.3357619914996963	0.0347	0.26226091384887695	0.19558236839544615	0.04429954763924229	1.7002788104089222e-05	5.03125
2400	1.4571948998178508	0.0325	0.26724469661712646	0.19840420356135058	0.045183064744133444	1.669299876084263e-05	5.09375
2600	1.5786278081360048	0.0306	0.2619116008281708	0.1900360027245305	0.04373409669211196	1.6383209417596036e-05	2.796875
2800	1.7000607164541592	0.0282	0.26362183690071106	0.19217670526418215	0.04311563471868815	1.6073420074349444e-05	2.921875
3000	1.8214936247723132	0.0266	0.26117631793022156	0.18380850442736207	0.042355810008481765	1.576363073110285e-05	2.703125
3200	1.9429265330904675	0.0183	0.26155000925064087	0.19081444001167655	0.04438789934973141	1.5453841387856258e-05	4.0
3400	2.0643594414086217	0.0107	0.28095561265945435	0.22506568064610294	0.08803364433135426	1.5144052044609668e-05	2.65625
3600	2.185792349726776	0.0158	0.27657684683799744	0.18653303493237325	0.04232046932428612	1.4834262701363073e-05	2.734375
3800	2.30722525804493	0.0114	0.2837420701980591	0.1877979955239856	0.043345349165959855	1.4524473358116483e-05	1.6328125
4000	2.4286581663630846	0.0135	0.2861778140068054	0.18945217475917098	0.04308029403449251	1.4214684014869891e-05	6.9375
4200	2.5500910746812386	0.0123	0.29045864939689636	0.18458694171450812	0.04239115069267741	1.3904894671623296e-05	4.53125
4400	2.6715239829993926	0.0073	0.2967182695865631	0.18546268366254742	0.04309796437659033	1.3595105328376705e-05	2.65625
4600	2.792956891317547	0.0073	0.3001556098461151	0.18205702053128345	0.042762227876731694	1.3285315985130111e-05	2.28125
4800	2.9143897996357016	0.0074	0.30166083574295044	0.18030553663520482	0.04210842521911224	1.297552664188352e-05	1.8671875
5000	3.0358227079538556	0.0042	0.30652889609336853	0.18303006714021602	0.04288592027141645	1.2665737298636928e-05	0.578125
5200	3.1572556162720096	0.0034	0.31825488805770874	0.18118127858324415	0.04159598529827537	1.2355947955390335e-05	0.796875
5400	3.278688524590164	0.0027	0.3178451657295227	0.18186241120949692	0.04216143624540571	1.2046158612143743e-05	1.7578125
5600	3.400121432908318	0.0039	0.31455591320991516	0.18234893451396322	0.042355810008481765	1.1736369268897152e-05	0.48828125
5800	3.5215543412264725	0.0024	0.3200189471244812	0.1821543251921767	0.04281523890302516	1.1426579925650558e-05	0.2890625
6000	3.6429872495446265	0.0015	0.31965574622154236	0.17884596672180597	0.04228512864009047	1.1116790582403967e-05	0.4296875
6200	3.764420157862781	0.0021	0.3236559331417084	0.17942979468716552	0.043044953350296865	1.0807001239157374e-05	0.98828125
6400	3.885853066180935	0.002	0.32796546816825867	0.17952709934805877	0.04249717274526435	1.0497211895910782e-05	0.8515625
6600	4.007285974499089	0.0013	0.32159319519996643	0.17826213875644642	0.041560644614079725	1.018742255266419e-05	0.291015625
6800	4.128718882817243	0.0014	0.3293284475803375	0.17894327138269922	0.04143695221939497	9.877633209417597e-06	0.283203125
7000	4.250151791135398	0.0011	0.32988446950912476	0.18001362265252505	0.04148996324568844	9.567843866171006e-06	0.18359375
7200	4.371584699453552	0.0014	0.33286359906196594	0.17826213875644642	0.04101286400904722	9.258054522924412e-06	0.310546875
7400	4.493017607771706	0.0019	0.33369237184524536	0.1790405760435925	0.041118886061634156	8.948265179677819e-06	2.4375
7600	4.61445051608986	0.0013	0.3339177668094635	0.17777561545198015	0.04103053435114504	8.638475836431227e-06	0.4375
7800	4.735883424408015	0.0014	0.33680835366249084	0.17942979468716552	0.04157831495617755	8.328686493184636e-06	0.306640625
8000	4.857316332726169	0.0017	0.3384738266468048	0.18040284129609807	0.04198473282442748	8.018897149938043e-06	0.248046875
8200	4.978749241044323	0.0014	0.3357013761997223	0.17884596672180597	0.04170200735086231	7.709107806691451e-06	0.369140625
8400	5.100182149362477	0.0009	0.33680522441864014	0.1797217086698453	0.04228512864009047	7.399318463444858e-06	0.2353515625
8600	5.221615057680632	0.0009	0.3375456929206848	0.17865135740001947	0.04279756856092734	7.089529120198265e-06	0.138671875
8800	5.343047965998785	0.0012	0.3405899703502655	0.17962440400895202	0.04180802940344925	6.779739776951674e-06	0.158203125
9000	5.46448087431694	0.001	0.34015366435050964	0.17884596672180597	0.042355810008481765	6.469950433705081e-06	0.30078125
9200	5.585913782635094	0.0009	0.3399837613105774	0.17933249002627225	0.04232046932428612	6.160161090458489e-06	0.2060546875
9400	5.707346690953249	0.0013	0.3426145315170288	0.179235185365379	0.04249717274526435	5.850371747211896e-06	0.2314453125
9600	5.828779599271402	0.0012	0.3423478603363037	0.179235185365379	0.04141928187729715	5.540582403965305e-06	0.2275390625
9800	5.950212507589557	0.001	0.34150007367134094	0.17913788070448575	0.041825699745547076	5.2307930607187115e-06	0.1796875
10000	6.071645415907711	0.001	0.34283334016799927	0.17874866206091272	0.04251484308736217	4.921003717472119e-06	0.2197265625
10200	6.193078324225866	0.0012	0.34264615178108215	0.17894327138269922	0.042143765903307887	4.6112143742255274e-06	0.2294921875
10400	6.314511232544019	0.001	0.34246060252189636	0.1776783107910869	0.04080081990387334	4.301425030978935e-06	0.201171875
10600	6.435944140862174	0.001	0.343102365732193	0.17913788070448575	0.04283290924512299	3.9916356877323426e-06	0.1279296875
10800	6.557377049180328	0.0011	0.34253108501434326	0.18050014595699135	0.042762227876731694	3.68184634448575e-06	0.2236328125
11000	6.678809957498482	0.0011	0.34328991174697876	0.1797217086698453	0.042550183771557816	3.372057001239158e-06	0.3203125
11200	6.800242865816636	0.0012	0.34304308891296387	0.17942979468716552	0.042938931297709926	3.0622676579925653e-06	0.26953125
11400	6.9216757741347905	0.0023	0.34302008152008057	0.18030553663520482	0.042638535482046935	2.752478314745973e-06	0.193359375
11600	7.043108682452945	0.001	0.3429974615573883	0.17962440400895202	0.043044953350296865	2.442688971499381e-06	0.2578125
11800	7.164541590771099	0.001	0.3425600230693817	0.18040284129609807	0.04249717274526435	2.1328996282527884e-06	0.220703125
12000	7.285974499089253	0.001	0.3428310751914978	0.18050014595699135	0.04270921685043822	1.823110285006196e-06	0.16796875
12200	7.407407407407407	0.001	0.34253931045532227	0.17845674807823295	0.04226745829799265	1.5133209417596037e-06	0.427734375
12400	7.528840315725562	0.0009	0.34326833486557007	0.17952709934805877	0.04240882103477523	1.2035315985130113e-06	0.2138671875
12600	7.6502732240437155	0.0009	0.34236183762550354	0.17981901333073855	0.04228512864009047	8.93742255266419e-07	0.390625
12800	7.77170613236187	0.001	0.3430650234222412	0.1797217086698453	0.04244416171897088	5.839529120198266e-07	0.2392578125
13000	7.893139040680024	0.0009	0.3429511487483978	0.18079205993967112	0.04313330506078598	2.7416356877323424e-07	0.59765625