metadata
language:
- zh
license: apache-2.0
base_model: openai/whisper-tiny
tags:
- generated_from_trainer
datasets:
- formospeech/hat_asr_aligned
model-index:
- name: Whisper Tiny Hakka Condenser
results: []
Whisper Tiny Hakka Condenser
This model is a fine-tuned version of openai/whisper-tiny on the HAT ASR Aligned dataset. It achieves the following results on the evaluation set:
- Loss: 0.1966
- Cer: 12.4315
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: 64
- 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: 976
- training_steps: 9760
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer |
|---|---|---|---|---|
| 1.161 | 0.9980 | 488 | 1.1833 | 47.8084 |
| 0.3742 | 1.9959 | 976 | 0.4698 | 25.2364 |
| 0.1966 | 2.9939 | 1464 | 0.3005 | 18.3346 |
| 0.1169 | 3.9918 | 1952 | 0.2523 | 16.0413 |
| 0.0795 | 4.9898 | 2440 | 0.2257 | 13.6440 |
| 0.0525 | 5.9877 | 2928 | 0.2156 | 15.5570 |
| 0.0336 | 6.9857 | 3416 | 0.2102 | 15.5905 |
| 0.0244 | 7.9836 | 3904 | 0.2036 | 12.6638 |
| 0.0182 | 8.9816 | 4392 | 0.1985 | 12.1437 |
| 0.012 | 9.9796 | 4880 | 0.1972 | 11.6501 |
| 0.0075 | 10.9775 | 5368 | 0.1986 | 12.9840 |
| 0.0056 | 11.9755 | 5856 | 0.1959 | 12.5066 |
| 0.0043 | 12.9734 | 6344 | 0.1970 | 12.8881 |
| 0.0034 | 13.9714 | 6832 | 0.1957 | 12.1622 |
| 0.0028 | 14.9693 | 7320 | 0.1971 | 11.5230 |
| 0.0024 | 15.9673 | 7808 | 0.1958 | 12.4280 |
| 0.0021 | 16.9652 | 8296 | 0.1964 | 12.4812 |
| 0.002 | 17.9632 | 8784 | 0.1976 | 12.8650 |
| 0.0018 | 18.9611 | 9272 | 0.1968 | 12.0385 |
| 0.0017 | 19.9591 | 9760 | 0.1966 | 12.4315 |
Framework versions
- Transformers 4.42.3
- Pytorch 2.3.0+cu121
- Datasets 2.20.0
- Tokenizers 0.19.1