hindi-asr / README.md
Chanpreet3000's picture
Initial Commit
b24cb59 verified
|
raw
history blame
2.09 kB
---
language:
- hi
license: apache-2.0
tags:
- hf-asr-leaderboard
- generated_from_trainer
datasets:
- mozilla-foundation/common_voice_11_0
metrics:
- wer
base_model: openai/whisper-small
model-index:
- name: Whisper Small Hi - Sanchit Gandhi
results:
- task:
type: automatic-speech-recognition
name: Automatic Speech Recognition
dataset:
name: Common Voice 11.0
type: mozilla-foundation/common_voice_11_0
args: 'Config: hi'
metrics:
- type: wer
value: 32.011343435198505
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 Small Hi - Sanchit Gandhi
This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the Common Voice 11.0 dataset.
It achieves the following results on the evaluation set:
- Loss: 0.4519
- Wer: 32.0113
## 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: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 5000
- mixed_precision_training: Native AMP
### Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|:-------------:|:-----:|:----:|:---------------:|:-------:|
| 0.1011 | 2.44 | 1000 | 0.3075 | 34.6313 |
| 0.0264 | 4.89 | 2000 | 0.3558 | 33.1288 |
| 0.0025 | 7.33 | 3000 | 0.4214 | 32.5912 |
| 0.0006 | 9.78 | 4000 | 0.4519 | 32.0113 |
| 0.0002 | 12.22 | 5000 | 0.4679 | 32.0960 |
### Framework versions
- Transformers 4.25.0.dev0
- Pytorch 1.12.1
- Datasets 2.5.3.dev0
- Tokenizers 0.12.1