103 lines
3.3 KiB
Markdown
103 lines
3.3 KiB
Markdown
|
---
|
||
|
|
library_name: transformers
|
||
|
|
language:
|
||
|
|
- en
|
||
|
|
license: mit
|
||
|
|
base_model: openai/whisper-large-v3-turbo
|
||
|
|
tags:
|
||
|
|
- generated_from_trainer
|
||
|
|
datasets:
|
||
|
|
- WillHeld/india_accent_cv
|
||
|
|
metrics:
|
||
|
|
- wer
|
||
|
|
model-index:
|
||
|
|
- name: Whisper Indian English Acccent
|
||
|
|
results:
|
||
|
|
- task:
|
||
|
|
type: automatic-speech-recognition
|
||
|
|
name: Automatic Speech Recognition
|
||
|
|
dataset:
|
||
|
|
name: Indian English Accent
|
||
|
|
type: WillHeld/india_accent_cv
|
||
|
|
args: 'split: train'
|
||
|
|
metrics:
|
||
|
|
- type: wer
|
||
|
|
value: 7.5056000168263415
|
||
|
|
name: Wer
|
||
|
|
---
|
||
|
|
|
||
|
|
<!-- 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. -->
|
||
|
|
|
||
|
|
# Whisper Indian English Acccent
|
||
|
|
|
||
|
|
This model is a fine-tuned version of [openai/whisper-large-v3-turbo](https://huggingface.co/openai/whisper-large-v3-turbo) on the Indian English Accent dataset.
|
||
|
|
It achieves the following results on the evaluation set:
|
||
|
|
- Loss: 0.2065
|
||
|
|
- Wer: 7.5056
|
||
|
|
|
||
|
|
## 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: 16
|
||
|
|
- eval_batch_size: 8
|
||
|
|
- seed: 42
|
||
|
|
- 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
|
||
|
|
- lr_scheduler_warmup_steps: 500
|
||
|
|
- num_epochs: 5
|
||
|
|
- mixed_precision_training: Native AMP
|
||
|
|
|
||
|
|
### Training results
|
||
|
|
|
||
|
|
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
||
|
|
|:-------------:|:------:|:-----:|:---------------:|:-------:|
|
||
|
|
| 0.342 | 0.1943 | 1000 | 0.3226 | 14.1310 |
|
||
|
|
| 0.2741 | 0.3885 | 2000 | 0.3130 | 13.9553 |
|
||
|
|
| 0.2576 | 0.5828 | 3000 | 0.2967 | 12.9931 |
|
||
|
|
| 0.2825 | 0.7770 | 4000 | 0.2692 | 12.3390 |
|
||
|
|
| 0.2295 | 0.9713 | 5000 | 0.2565 | 11.8331 |
|
||
|
|
| 0.1489 | 1.1655 | 6000 | 0.2498 | 11.6933 |
|
||
|
|
| 0.1485 | 1.3598 | 7000 | 0.2452 | 11.1411 |
|
||
|
|
| 0.1385 | 1.5540 | 8000 | 0.2346 | 10.4428 |
|
||
|
|
| 0.1253 | 1.7483 | 9000 | 0.2254 | 10.1852 |
|
||
|
|
| 0.1297 | 1.9425 | 10000 | 0.2144 | 9.7109 |
|
||
|
|
| 0.0594 | 2.1368 | 11000 | 0.2174 | 9.5363 |
|
||
|
|
| 0.0629 | 2.3310 | 12000 | 0.2136 | 9.8276 |
|
||
|
|
| 0.0654 | 2.5253 | 13000 | 0.2102 | 9.4301 |
|
||
|
|
| 0.0625 | 2.7195 | 14000 | 0.2075 | 8.9432 |
|
||
|
|
| 0.0574 | 2.9138 | 15000 | 0.2009 | 8.7802 |
|
||
|
|
| 0.0276 | 3.1080 | 16000 | 0.2050 | 8.4594 |
|
||
|
|
| 0.0251 | 3.3023 | 17000 | 0.2046 | 8.5951 |
|
||
|
|
| 0.0246 | 3.4965 | 18000 | 0.2035 | 8.1187 |
|
||
|
|
| 0.0259 | 3.6908 | 19000 | 0.2002 | 8.0588 |
|
||
|
|
| 0.021 | 3.8850 | 20000 | 0.1951 | 7.9147 |
|
||
|
|
| 0.0072 | 4.0793 | 21000 | 0.2053 | 7.7548 |
|
||
|
|
| 0.0067 | 4.2735 | 22000 | 0.2085 | 7.4972 |
|
||
|
|
| 0.0067 | 4.4678 | 23000 | 0.2094 | 7.6970 |
|
||
|
|
| 0.0062 | 4.6620 | 24000 | 0.2071 | 7.7433 |
|
||
|
|
| 0.0046 | 4.8563 | 25000 | 0.2065 | 7.5056 |
|
||
|
|
|
||
|
|
|
||
|
|
### Framework versions
|
||
|
|
|
||
|
|
- Transformers 4.49.0
|
||
|
|
- Pytorch 2.2.0a0+81ea7a4
|
||
|
|
- Datasets 3.3.2
|
||
|
|
- Tokenizers 0.21.0
|