File size: 4,721 Bytes
0501696 6d40efb 0501696 6d40efb 0501696 6d40efb 0501696 6d40efb 0501696 6d40efb 0501696 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 |
---
license: apache-2.0
base_model: Samuael/asr-amharic-phoneme-based-37-6
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: asr-amharic-phoneme-based-37-6
results: []
---
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
# asr-amharic-phoneme-based-37-6
This model is a fine-tuned version of [Samuael/asr-amharic-phoneme-based-37-6](https://huggingface.co/Samuael/asr-amharic-phoneme-based-37-6) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.2467
- Wer: 0.2877
- Phoneme Cer: 0.0539
- Cer: 0.0777
## 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: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- num_epochs: 10
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Phoneme Cer | Cer |
|:-------------:|:-----:|:----:|:---------------:|:------:|:-----------:|:------:|
| 3.2186 | 0.25 | 200 | 3.2626 | 1.0 | 1.0 | 1.0 |
| 3.0459 | 0.5 | 400 | 3.0440 | 1.0 | 1.0 | 1.0 |
| 2.9983 | 0.75 | 600 | 2.9807 | 1.0 | 1.0 | 1.0 |
| 1.8006 | 1.0 | 800 | 1.4020 | 0.8218 | 0.2751 | 0.4270 |
| 0.6145 | 1.25 | 1000 | 0.4420 | 0.3611 | 0.0717 | 0.1040 |
| 0.4288 | 1.5 | 1200 | 0.3304 | 0.3394 | 0.0649 | 0.0941 |
| 0.3335 | 1.75 | 1400 | 0.2996 | 0.3314 | 0.0622 | 0.0909 |
| 0.4583 | 2.01 | 1600 | 0.2815 | 0.3263 | 0.0612 | 0.0890 |
| 0.3966 | 2.26 | 1800 | 0.2743 | 0.3120 | 0.0584 | 0.0847 |
| 0.3781 | 2.51 | 2000 | 0.2669 | 0.3241 | 0.0609 | 0.0881 |
| 0.5792 | 2.76 | 2200 | 0.2597 | 0.3135 | 0.0590 | 0.0857 |
| 0.3322 | 3.01 | 2400 | 0.2548 | 0.3076 | 0.0575 | 0.0837 |
| 0.4129 | 3.26 | 2600 | 0.2577 | 0.3051 | 0.0575 | 0.0832 |
| 0.3453 | 3.51 | 2800 | 0.2520 | 0.3070 | 0.0577 | 0.0828 |
| 0.3279 | 3.76 | 3000 | 0.2516 | 0.2989 | 0.0563 | 0.0818 |
| 0.3312 | 4.01 | 3200 | 0.2487 | 0.2947 | 0.0553 | 0.0801 |
| 0.2818 | 4.26 | 3400 | 0.2530 | 0.2985 | 0.0563 | 0.0816 |
| 0.2244 | 4.51 | 3600 | 0.2529 | 0.3021 | 0.0565 | 0.0820 |
| 0.3562 | 4.76 | 3800 | 0.2477 | 0.2984 | 0.0563 | 0.0814 |
| 0.2922 | 5.01 | 4000 | 0.2530 | 0.2995 | 0.0563 | 0.0813 |
| 0.2535 | 5.26 | 4200 | 0.2516 | 0.2990 | 0.0556 | 0.0805 |
| 0.2043 | 5.51 | 4400 | 0.2532 | 0.2966 | 0.0557 | 0.0800 |
| 0.2653 | 5.76 | 4600 | 0.2471 | 0.2922 | 0.0548 | 0.0792 |
| 0.4389 | 6.02 | 4800 | 0.2493 | 0.2942 | 0.0555 | 0.0805 |
| 0.2757 | 6.27 | 5000 | 0.2533 | 0.2928 | 0.0551 | 0.0799 |
| 0.3063 | 6.52 | 5200 | 0.2459 | 0.2950 | 0.0554 | 0.0800 |
| 0.2069 | 6.77 | 5400 | 0.2499 | 0.2946 | 0.0555 | 0.0800 |
| 0.2018 | 7.02 | 5600 | 0.2503 | 0.2934 | 0.0549 | 0.0792 |
| 0.2068 | 7.27 | 5800 | 0.2469 | 0.2966 | 0.0551 | 0.0798 |
| 0.2535 | 7.52 | 6000 | 0.2513 | 0.2936 | 0.0548 | 0.0792 |
| 0.285 | 7.77 | 6200 | 0.2492 | 0.2941 | 0.0548 | 0.0795 |
| 0.1801 | 8.02 | 6400 | 0.2460 | 0.2896 | 0.0540 | 0.0781 |
| 0.1705 | 8.27 | 6600 | 0.2475 | 0.2933 | 0.0549 | 0.0793 |
| 0.3344 | 8.52 | 6800 | 0.2501 | 0.2890 | 0.0540 | 0.0777 |
| 0.2236 | 8.77 | 7000 | 0.2471 | 0.2871 | 0.0536 | 0.0773 |
| 0.349 | 9.02 | 7200 | 0.2480 | 0.2869 | 0.0541 | 0.0779 |
| 0.2692 | 9.27 | 7400 | 0.2482 | 0.2891 | 0.0540 | 0.0779 |
| 0.2265 | 9.52 | 7600 | 0.2481 | 0.2879 | 0.0539 | 0.0776 |
| 0.1649 | 9.77 | 7800 | 0.2467 | 0.2877 | 0.0539 | 0.0777 |
### Framework versions
- Transformers 4.35.2
- Pytorch 2.1.0+cu121
- Datasets 2.16.1
- Tokenizers 0.15.1
|