|
|
--- |
|
|
library_name: peft |
|
|
license: mit |
|
|
base_model: openai/whisper-large-v3-turbo |
|
|
tags: |
|
|
- generated_from_trainer |
|
|
metrics: |
|
|
- wer |
|
|
model-index: |
|
|
- name: whisper-hf-rslora |
|
|
results: [] |
|
|
datasets: |
|
|
- compulsi0n/heart-failure-audio |
|
|
pipeline_tag: automatic-speech-recognition |
|
|
--- |
|
|
|
|
|
<!-- 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-hf-rslora |
|
|
|
|
|
This model is a fine-tuned version of [openai/whisper-large-v3-turbo](https://huggingface.co/openai/whisper-large-v3-turbo) on compulsion/heart-failure-audio. |
|
|
It achieves the following results on the evaluation set: |
|
|
- Loss: 0.6919 |
|
|
- Wer: 0.2424 |
|
|
|
|
|
## Model description |
|
|
|
|
|
A PEFT rank-stablized LoRA adapter of whisper-large-v3-turbo finetuned on heart failure audio data that is conversational, longitudinal, and focused on chronic illness management and care coordination in a community-based healthcare setting. |
|
|
|
|
|
|
|
|
## Intended uses & limitations |
|
|
|
|
|
To be used in ASR tasks specifically in the heart failure domain. |
|
|
|
|
|
|
|
|
## Benchmark (base whisper-large-v3-turbo vs. finetuned rank-stablized LoRA adapter) |
|
|
|
|
|
Normalized for PHI redactions and throught Transformer's BasicTextNormalizer. |
|
|
|
|
|
| Model | Raw WER (%) | Normalised WER (%) | |
|
|
| :------: | :---------: | :----------------: | |
|
|
| Baseline | 35.00 | 26.71 | |
|
|
| rsLoRA | 26.18 | 20.71 | |
|
|
|
|
|
## Training procedure |
|
|
|
|
|
### Training hyperparameters |
|
|
|
|
|
The following hyperparameters were used during training: |
|
|
- learning_rate: 1e-05 |
|
|
- train_batch_size: 4 |
|
|
- eval_batch_size: 8 |
|
|
- seed: 42 |
|
|
- gradient_accumulation_steps: 4 |
|
|
- total_train_batch_size: 16 |
|
|
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments |
|
|
- lr_scheduler_type: constant_with_warmup |
|
|
- lr_scheduler_warmup_steps: 500 |
|
|
- num_epochs: 8 |
|
|
- mixed_precision_training: Native AMP |
|
|
|
|
|
### Training results |
|
|
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Wer | |
|
|
|:-------------:|:-----:|:----:|:---------------:|:------:| |
|
|
| 2.3062 | 1.0 | 92 | 1.1343 | 0.2388 | |
|
|
| 1.0317 | 2.0 | 184 | 0.7145 | 0.2620 | |
|
|
| 0.6833 | 3.0 | 276 | 0.6606 | 0.2105 | |
|
|
| 0.5934 | 4.0 | 368 | 0.6292 | 0.2122 | |
|
|
| 0.5104 | 5.0 | 460 | 0.6347 | 0.2521 | |
|
|
| 0.4392 | 6.0 | 552 | 0.6444 | 0.2729 | |
|
|
| 0.3653 | 7.0 | 644 | 0.6701 | 0.2198 | |
|
|
| 0.3178 | 8.0 | 736 | 0.6919 | 0.2424 | |
|
|
|
|
|
|
|
|
### Framework versions |
|
|
|
|
|
- PEFT 0.15.2 |
|
|
- Transformers 4.52.4 |
|
|
- Pytorch 2.6.0+cu124 |
|
|
- Datasets 3.6.0 |
|
|
- Tokenizers 0.21.1 |
