Whisper Large V3 — Multilingual Luganda + Ugandan English NSS (ml-v1)

Stage 2 multilingual fine-tune on the CDLI Ugandan Nonstandard Speech datasets. Built on top of lg_waxal_tune_whisper_large_7 (Stage 1). Trained jointly on Luganda and Ugandan English NSS data. Evaluated on Luganda NSS test set.

Training Details

• Datasets: cdli/ugandan_luganda_nonstandard_speech_v1.0 + cdli/ugandan_english_nonstandard_speech_v1.0
• Stage 1 checkpoint: lg_waxal_tune_whisper_large_7/best_model
• Encoder frozen: No
• Decoder frozen: No
• Projection frozen: No
• SpecAugment: On
• Data augmentation: Yes (Gaussian noise, gain perturbation, speed perturbation)
• Learning rate: 5e-5 (polynomial decay)
• Max steps: 1000
• Best checkpoint: Step 850

Results on Luganda NSS Test Set (checkpoint-850)

• Overall WER (normalized): 0.745
• Overall CER (normalized): 0.290
• Avg WER (normalized): 0.529
• Avg CER (normalized): 0.182
• Meaning preservation: 85.1%