README.md

metadata

library_name: peft
language:
  - si
license: apache-2.0
base_model: openai/whisper-small
tags:
  - base_model:adapter:openai/whisper-small
  - lora
  - transformers
datasets:
  - SPEAK-ASR/openslr-sinhala-asr-norm-noise-rem-preprocessed
metrics:
  - wer
model-index:
  - name: SPEAK-ASR/whisper-si-exp-9
    results:
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: SPEAK-ASR/openslr-sinhala-asr-norm-noise-rem-preprocessed
          type: SPEAK-ASR/openslr-sinhala-asr-norm-noise-rem-preprocessed
          args: 'config: si, split: test'
        metrics:
          - type: wer
            value: 16.6932298129575
            name: Wer

SPEAK-ASR/whisper-si-exp-9

This model is a fine-tuned version of openai/whisper-small on the SPEAK-ASR/openslr-sinhala-asr-norm-noise-rem-preprocessed dataset. It achieves the following results on the evaluation set:

• Loss: 0.1467
• Wer: 16.6932

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: 3e-05
• train_batch_size: 32
• eval_batch_size: 256
• seed: 42
• optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 200
• num_epochs: 15.0

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0	0	1.8019	216.5953
0.2553	0.7941	1500	0.2526	94.6301
0.2093	1.5881	3000	0.2105	88.0346
0.1887	2.3822	4500	0.1933	83.1024
0.1797	3.1763	6000	0.1836	83.8828
0.1733	3.9704	7500	0.1773	78.5470
0.1606	4.7644	9000	0.1737	80.1329
0.1645	5.0	9445	0.1733	19.9827
0.1690	5.5585	10500	0.1751	19.9070
0.1561	6.3526	12000	0.1690	19.4095
0.1470	7.1466	13500	0.1632	18.5933
0.1473	7.9407	15000	0.1596	18.4174
0.1427	8.7348	16500	0.1560	17.8908
0.1359	9.5289	18000	0.1530	17.6062
0.1302	10.3229	19500	0.1518	17.3700
0.1262	11.1170	21000	0.1503	17.1719
0.1266	11.9111	22500	0.1485	16.9562
0.1238	12.7051	24000	0.1480	16.9517
0.1199	13.4992	25500	0.1472	16.7001
0.1206	14.2933	27000	0.1467	16.6932

Framework versions

• PEFT 0.18.1
• Transformers 5.2.0
• Pytorch 2.9.0.dev20250821+rocm7.0.0.git125803b7
• Datasets 4.0.0
• Tokenizers 0.22.2