|
--- |
|
language: |
|
- hy-AM |
|
license: apache-2.0 |
|
tags: |
|
- automatic-speech-recognition |
|
- mozilla-foundation/common_voice_8_0 |
|
- generated_from_trainer |
|
- robust-speech-event |
|
datasets: |
|
- common_voice |
|
model-index: |
|
- name: wav2vec2-xls-r-1b-hy-cv |
|
results: |
|
- task: |
|
type: automatic-speech-recognition |
|
name: Speech Recognition |
|
dataset: |
|
type: mozilla-foundation/common_voice_8_0 |
|
name: Common Voice hy-AM |
|
args: hy-AM |
|
metrics: |
|
- type: wer |
|
value: 10.92896174863388 |
|
name: WER LM |
|
- type: cer |
|
value: 2.3773394031360646 |
|
name: CER LM |
|
- task: |
|
name: Speech Recognition |
|
type: automatic-speech-recognition |
|
dataset: |
|
name: Robust Speech Event - Dev Data |
|
type: speech-recognition-community-v2/dev_data |
|
args: hy |
|
metrics: |
|
- name: Test WER |
|
type: wer |
|
value: 19.942816297355254 |
|
- name: Test CER |
|
type: cer |
|
value: 7.332618465282714 |
|
--- |
|
|
|
|
|
<!-- 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. --> |
|
|
|
# Wav2Vec2-XLS-R-1b-hy |
|
|
|
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-1b](https://huggingface.co/facebook/wav2vec2-xls-r-1b) on the /WORKSPACE/DATA/HY/NOIZY_STUDENT_3/ - NA dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.1827 |
|
- Wer: 0.2389 |
|
- Cer: 0.0427 |
|
- Wer LM: 0.1093 |
|
- Cer LM: 0.0238 |
|
|
|
## 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: 8e-05 |
|
- train_batch_size: 16 |
|
- eval_batch_size: 64 |
|
- seed: 842 |
|
- gradient_accumulation_steps: 8 |
|
- total_train_batch_size: 128 |
|
- optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-08 |
|
- lr_scheduler_type: cosine |
|
- lr_scheduler_warmup_ratio: 0.1 |
|
- training_steps: 3200 |
|
- mixed_precision_training: Native AMP |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer | |
|
|:-------------:|:-----:|:----:|:---------------:|:------:|:------:| |
|
| 4.0311 | 3.51 | 200 | 0.7943 | 0.8981 | 0.2374 | |
|
| 1.4388 | 7.02 | 400 | 0.2546 | 0.3821 | 0.0658 | |
|
| 1.0949 | 10.53 | 600 | 0.2201 | 0.3216 | 0.0573 | |
|
| 1.0279 | 14.04 | 800 | 0.2250 | 0.3271 | 0.0583 | |
|
| 0.9923 | 17.54 | 1000 | 0.2074 | 0.3111 | 0.0543 | |
|
| 0.972 | 21.05 | 1200 | 0.2165 | 0.2955 | 0.0536 | |
|
| 0.9587 | 24.56 | 1400 | 0.2064 | 0.3017 | 0.0535 | |
|
| 0.9421 | 28.07 | 1600 | 0.2062 | 0.2884 | 0.0519 | |
|
| 0.9189 | 31.58 | 1800 | 0.2014 | 0.2822 | 0.0507 | |
|
| 0.8919 | 35.09 | 2000 | 0.1952 | 0.2689 | 0.0488 | |
|
| 0.8615 | 38.6 | 2200 | 0.2020 | 0.2685 | 0.0480 | |
|
| 0.834 | 42.11 | 2400 | 0.2001 | 0.2654 | 0.0467 | |
|
| 0.8056 | 45.61 | 2600 | 0.1935 | 0.2498 | 0.0448 | |
|
| 0.7888 | 49.12 | 2800 | 0.1892 | 0.2451 | 0.0446 | |
|
| 0.761 | 52.63 | 3000 | 0.1884 | 0.2432 | 0.0441 | |
|
| 0.742 | 56.14 | 3200 | 0.1827 | 0.2389 | 0.0427 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.17.0.dev0 |
|
- Pytorch 1.10.2+cu102 |
|
- Datasets 1.18.2.dev0 |
|
- Tokenizers 0.11.0 |
|
|