andrewbawitlung's picture
Add CER, hyperparameters, and training logs to README
ab5a6e5 verified
|
Raw
History Blame
8.78 kB
metadata
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-E2-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: 19.4954
          - name: Cer
            type: cer
            value: 4.4174

qwen3-asr-0.6b-mizonal3-E2-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: 19.4954
  • Cer: 4.4174

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

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 0.18214936247723132 0.4524 0.7468264102935791 0.5282670039894911 0.1620900480633305 1.9946561338289966e-05 14.5
400 0.36429872495446264 0.2727 0.40020251274108887 0.322467646200253 0.0809301668080294 1.9481877323420074e-05 10.8125
600 0.546448087431694 0.1697 0.3364563286304474 0.27955629074632676 0.06432004523607578 1.9017193308550188e-05 7.5625
800 0.7285974499089253 0.2061 0.29908183217048645 0.24871071324316435 0.059849448685326545 1.85525092936803e-05 5.78125
1000 0.9107468123861566 0.1388 0.2859058380126953 0.24404008952028802 0.05465436810856658 1.808782527881041e-05 7.375
1200 1.092896174863388 0.0883 0.2675563097000122 0.22506568064610294 0.052604608425219114 1.762314126394052e-05 5.65625
1400 1.2750455373406193 0.0722 0.26070064306259155 0.2159190425221368 0.05103194797851286 1.7158457249070632e-05 3.84375
1600 1.4571948998178508 0.0763 0.2592693865299225 0.21309720735623236 0.04882315521628499 1.6693773234200746e-05 4.4375
1800 1.639344262295082 0.0755 0.252381831407547 0.20035029677921573 0.04634930732258977 1.6229089219330857e-05 8.8125
2000 1.8214936247723132 0.0622 0.24645309150218964 0.20229638999708086 0.04539510884930732 1.5764405204460968e-05 4.59375
2200 2.0036429872495445 0.0477 0.24331779778003693 0.19957185949206968 0.045342097823013856 1.529972118959108e-05 3.84375
2400 2.185792349726776 0.0447 0.25787490606307983 0.20180986669261458 0.046773395532937516 1.4835037174721192e-05 5.65625
2600 2.3679417122040074 0.0332 0.25755423307418823 0.19314975187311473 0.045077042691546505 1.4370353159851303e-05 3.4375
2800 2.5500910746812386 0.0305 0.2666151821613312 0.19383088449936753 0.04451159174441617 1.3905669144981413e-05 7.21875
3000 2.73224043715847 0.0297 0.2642105221748352 0.19752846161331128 0.047179813401187445 1.3440985130111526e-05 4.21875
3200 2.9143897996357016 0.0472 0.2631722390651703 0.18964678408095748 0.04343370087644897 1.2976301115241637e-05 4.34375
3400 3.096539162112933 0.0152 0.28036123514175415 0.19305244721222148 0.043857789086796724 1.251161710037175e-05 4.46875
3600 3.278688524590164 0.0114 0.2897799611091614 0.19285783789043495 0.04419352558665536 1.2046933085501859e-05 2.515625
3800 3.4608378870673953 0.0159 0.2936338484287262 0.18770069086309235 0.043769437376307604 1.158224907063197e-05 2.9375
4000 3.6429872495446265 0.009 0.2900753617286682 0.1934416658557945 0.04410517387616624 1.1117565055762082e-05 2.0625
4200 3.8251366120218577 0.0115 0.2965144217014313 0.19110635399435633 0.04484732824427481 1.0652881040892193e-05 4.1875
4400 4.007285974499089 0.0062 0.3001888692378998 0.18828451882845187 0.043363019508057675 1.0188197026022306e-05 5.15625
4600 4.18943533697632 0.0054 0.3130307197570801 0.18604651162790697 0.04290359061351428 9.723513011152417e-06 1.2265625
4800 4.371584699453552 0.0047 0.31496503949165344 0.1905225260289968 0.044158184902459714 9.258828996282528e-06 1.2109375
5000 4.553734061930784 0.0094 0.3163476586341858 0.1927605332295417 0.044564602770709644 8.79414498141264e-06 0.9375
5200 4.735883424408015 0.0037 0.3209853768348694 0.19013330738542375 0.04429954763924229 8.329460966542751e-06 1.0546875
5400 4.918032786885246 0.0041 0.3182981014251709 0.19042522136810353 0.04378710771840543 7.864776951672864e-06 0.9921875
5600 5.100182149362477 0.0022 0.33114519715309143 0.19353897051668775 0.04442324003392706 7.400092936802975e-06 0.4609375
5800 5.2823315118397085 0.0022 0.33354538679122925 0.19169018195971588 0.04440556969182923 6.935408921933086e-06 0.703125
6000 5.46448087431694 0.0023 0.33456122875213623 0.18993869806363725 0.04373409669211196 6.4707249070631975e-06 0.84375
6200 5.646630236794171 0.0027 0.3360964357852936 0.1927605332295417 0.044175855244557534 6.006040892193309e-06 0.59765625
6400 5.828779599271402 0.0025 0.33629176020622253 0.1927605332295417 0.044864998586372634 5.54135687732342e-06 2.15625
6600 6.0109289617486334 0.0019 0.3365257680416107 0.19042522136810353 0.04419352558665536 5.076672862453532e-06 0.45703125
6800 6.193078324225866 0.0022 0.34069618582725525 0.19266322856864843 0.04421119592875318 4.611988847583643e-06 0.349609375
7000 6.375227686703097 0.0018 0.3427211046218872 0.19353897051668775 0.04472363584959005 4.147304832713755e-06 0.380859375
7200 6.557377049180328 0.0018 0.3420465290546417 0.19169018195971588 0.044529262086514 3.6826208178438664e-06 0.61328125
7400 6.739526411657559 0.0034 0.34137850999832153 0.19305244721222148 0.04449392140231835 3.217936802973978e-06 0.4140625
7600 6.9216757741347905 0.0019 0.34198904037475586 0.19149557263792935 0.04382244840260108 2.7532527881040896e-06 0.3515625
7800 7.103825136612022 0.0017 0.34269949793815613 0.19266322856864843 0.044175855244557534 2.288568773234201e-06 0.4296875
8000 7.285974499089253 0.0024 0.34260717034339905 0.19460932178651358 0.0449180096126661 1.8238847583643125e-06 0.51953125
8200 7.468123861566484 0.0016 0.34319037199020386 0.19314975187311473 0.044882668928470454 1.3592007434944238e-06 0.279296875
8400 7.6502732240437155 0.0017 0.3434956669807434 0.19237131458596868 0.04431721798134012 8.945167286245354e-07 0.326171875
8600 7.832422586520947 0.0021 0.342936247587204 0.193636275177581 0.04449392140231835 4.298327137546469e-07 0.73828125