File size: 4,267 Bytes
6203e06 233bb82 6203e06 f3cb18b 6203e06 233bb82 6203e06 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 | ---
library_name: transformers
license: apache-2.0
base_model: mouseyy/result_data_2-3
tags:
- generated_from_trainer
datasets:
- common_voice_17_0
metrics:
- wer
model-index:
- name: best_model_2_copy
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: common_voice_17_0
type: common_voice_17_0
config: uk
split: test
args: uk
metrics:
- name: Wer
type: wer
value: 0.2984287348943652
---
<!-- 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. -->
# Ukrainian Wav2Vec2 Model for Transcription with Lexical Stress Marking (`+`)
This model is a fine-tuned version of [mouseyy/result_data_2-3](https://huggingface.co/mouseyy/result_data_2-3) on the common_voice_17_0 dataset.
It achieves the following results on the evaluation set:
- Loss: 0.2132
- Wer: 0.2984
- Cer: 0.1512
## 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: 4e-05
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- distributed_type: multi-GPU
- num_devices: 2
- total_train_batch_size: 32
- total_eval_batch_size: 32
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- num_epochs: 15.0
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|:-------------:|:-------:|:-----:|:---------------:|:------:|:------:|
| 0.1936 | 0.4550 | 500 | 0.2082 | 0.3305 | 0.1590 |
| 0.174 | 0.9099 | 1000 | 0.2082 | 0.3284 | 0.1586 |
| 0.1855 | 1.3649 | 1500 | 0.1981 | 0.3292 | 0.1585 |
| 0.1724 | 1.8198 | 2000 | 0.1998 | 0.3265 | 0.1580 |
| 0.1667 | 2.2748 | 2500 | 0.1994 | 0.3288 | 0.1583 |
| 0.1635 | 2.7298 | 3000 | 0.2050 | 0.3223 | 0.1569 |
| 0.14 | 3.1847 | 3500 | 0.2069 | 0.3205 | 0.1567 |
| 0.1573 | 3.6397 | 4000 | 0.2067 | 0.3207 | 0.1569 |
| 0.1487 | 4.0946 | 4500 | 0.2101 | 0.3224 | 0.1572 |
| 0.1501 | 4.5496 | 5000 | 0.2110 | 0.3176 | 0.1563 |
| 0.1486 | 5.0045 | 5500 | 0.2040 | 0.3159 | 0.1557 |
| 0.1342 | 5.4595 | 6000 | 0.2041 | 0.3144 | 0.1551 |
| 0.1396 | 5.9145 | 6500 | 0.2057 | 0.3143 | 0.1552 |
| 0.136 | 6.3694 | 7000 | 0.2098 | 0.3131 | 0.1545 |
| 0.1266 | 6.8244 | 7500 | 0.2095 | 0.3106 | 0.1542 |
| 0.1283 | 7.2793 | 8000 | 0.2160 | 0.3085 | 0.1538 |
| 0.1229 | 7.7343 | 8500 | 0.2175 | 0.3076 | 0.1538 |
| 0.1267 | 8.1893 | 9000 | 0.2114 | 0.3057 | 0.1531 |
| 0.1127 | 8.6442 | 9500 | 0.2063 | 0.3069 | 0.1528 |
| 0.1165 | 9.0992 | 10000 | 0.2094 | 0.3048 | 0.1532 |
| 0.1222 | 9.5541 | 10500 | 0.2079 | 0.3067 | 0.1532 |
| 0.1127 | 10.0091 | 11000 | 0.2089 | 0.3056 | 0.1531 |
| 0.1084 | 10.4641 | 11500 | 0.2117 | 0.3032 | 0.1526 |
| 0.1155 | 10.9190 | 12000 | 0.2075 | 0.3045 | 0.1527 |
| 0.0955 | 11.3740 | 12500 | 0.2183 | 0.3026 | 0.1523 |
| 0.1146 | 11.8289 | 13000 | 0.2116 | 0.3015 | 0.1521 |
| 0.1094 | 12.2839 | 13500 | 0.2090 | 0.2993 | 0.1516 |
| 0.1072 | 12.7389 | 14000 | 0.2124 | 0.3002 | 0.1517 |
| 0.1125 | 13.1938 | 14500 | 0.2131 | 0.3000 | 0.1517 |
| 0.1058 | 13.6488 | 15000 | 0.2170 | 0.2992 | 0.1515 |
| 0.0951 | 14.1037 | 15500 | 0.2160 | 0.2986 | 0.1513 |
| 0.1035 | 14.5587 | 16000 | 0.2134 | 0.2986 | 0.1513 |
### Framework versions
- Transformers 4.49.0
- Pytorch 2.6.0+cu124
- Datasets 3.3.2
- Tokenizers 0.21.0
|