|
|
|
--- |
|
pretty_name: "WhisperKit ASR Evaluation Results" |
|
tags: |
|
- whisper |
|
- whisperkit |
|
- coreml |
|
- asr |
|
- quantized |
|
--- |
|
# WhisperKit Evaluation Results |
|
|
|
|
|
|
|
## Dataset: `librispeech` |
|
|
|
### Quality Evaluation |
|
|
|
| | WER | QoI (%) | File Size (MB) | |
|
|:----------------------------------------------------------------------------------------------------------------------------------------------------------------|------:|----------:|-----------------:| |
|
| [WhisperOpenAIAPI/openai_whisper-large-v2](https://huggingface.co/argmaxinc/whisperkit-coreml/tree/main/WhisperOpenAIAPI/openai_whisper-large-v2) | 2.85 | 100 | 3100 | |
|
| [WhisperKit/openai_whisper-large-v2](https://huggingface.co/argmaxinc/whisperkit-coreml/tree/main/WhisperKit/openai_whisper-large-v2) | 3.28 | 96.6 | 3100 | |
|
| [WhisperKit/openai_whisper-large-v2_1050MB](https://huggingface.co/argmaxinc/whisperkit-coreml/tree/main/WhisperKit/openai_whisper-large-v2_1050MB) | 3.32 | 95 | 1050 | |
|
| [WhisperKit/openai_whisper-large-v2_turbo](https://huggingface.co/argmaxinc/whisperkit-coreml/tree/main/WhisperKit/openai_whisper-large-v2_turbo) | 3.24 | 96.6 | 3100 | |
|
| [WhisperKit/openai_whisper-large-v2_turbo_1022MB](https://huggingface.co/argmaxinc/whisperkit-coreml/tree/main/WhisperKit/openai_whisper-large-v2_turbo_1022MB) | 3.33 | 94.9 | 1022 | |
|
| [WhisperKit/openai_whisper-small](https://huggingface.co/argmaxinc/whisperkit-coreml/tree/main/WhisperKit/openai_whisper-small) | 3.98 | 82.9 | 483 | |
|
| [WhisperKit/openai_whisper-base](https://huggingface.co/argmaxinc/whisperkit-coreml/tree/main/WhisperKit/openai_whisper-base) | 6.11 | 67.1 | 145 | |
|
| [WhisperKit/openai_whisper-tiny](https://huggingface.co/argmaxinc/whisperkit-coreml/tree/main/WhisperKit/openai_whisper-tiny) | 8.94 | 52.4 | 66 | |
|
| [WhisperKit/openai_whisper-large-v3](https://huggingface.co/argmaxinc/whisperkit-coreml/tree/main/WhisperKit/openai_whisper-large-v3) | 2.48 | 95.2 | 3100 | |
|
| [WhisperKit/openai_whisper-large-v3_turbo](https://huggingface.co/argmaxinc/whisperkit-coreml/tree/main/WhisperKit/openai_whisper-large-v3_turbo) | 2.44 | 95.4 | 3100 | |
|
| [WhisperKit/openai_whisper-large-v3_turbo_1018MB](https://huggingface.co/argmaxinc/whisperkit-coreml/tree/main/WhisperKit/openai_whisper-large-v3_turbo_1018MB) | 2.49 | 94.8 | 1018 | |
|
|
|
|
|
### Quality-of-Inference (QoI) Certification |
|
We believe that rigorously measuring the quality of inference is necessary for developers and |
|
enterprises to make informed decisions when opting to use optimized or compressed variants of |
|
any machine learning model in production. For WhisperKit, we take the following implementations |
|
and benchmark them using consistent evaluation harnesses: |
|
|
|
- `WhisperOpenAIAPI`: [OpenAI's Whisper API](https://platform.openai.com/docs/guides/speech-to-text)($0.36/hour as of 02/29/24, 25MB max file size) |
|
- `WhisperKit`: Argmax's Core ML implementation [[Eval Harness]](https://github.com/argmaxinc/whisperkittools/blob/main/whisperkit/pipelines.py#L100) [[Repo]](https://github.com/argmaxinc/WhisperKit) |
|
- `whisper.cpp`: A C++ implementation form ggerganov [[Eval Harness]](https://github.com/argmaxinc/whisperkittools/blob/main/whisperkit/pipelines.py#L212) [[Repo]](https://github.com/ggerganov/whisper.cpp) |
|
- `WhisperMLX`: A Python implementation from Apple MLX [[Eval Harness]](https://github.com/argmaxinc/whisperkittools/blob/main/whisperkit/pipelines.py#L338) [[Repo]](https://github.com/ml-explore/mlx-examples/blob/main/whisper/whisper/transcribe.py) |
|
|
|
`WhisperOpenAIAPI` is the reference and we assume that it is using the equivalent of |
|
[openai/whisper-large-v2](https://huggingface.co/openai/whisper-large-v2) in float16 precision. |
|
In all measurements, we care primarily about per-example no-regressions (quantified as `qoi` below) |
|
which is a stricter metric compared to dataset average WER. A 100% `qoi` preserves perfect |
|
backwards-compatibility on the test distribution and avoids "perceived regressions", the phenomenon |
|
where per-example known behavior changes after a code/model update and causes divergence in |
|
downstream code or breaks the user experience itself (even if dataset averages might stay flat |
|
across updates). Pseudocode for `qoi`: |
|
|
|
```python |
|
qoi = [] |
|
for example in dataset: |
|
no_regression = wer(optimized_model(example)) <= wer(reference_model(example)) |
|
qoi.append(no_regression) |
|
qoi = (sum(qoi) / len(qoi)) * 100. |
|
``` |
|
|
|
We use `librispeech/test.clean` (~5 hours of short English audio clips) and `earnings22` (~120 hours of long English audio clips with various accents). |
|
We anticipate developers that use Whisper (or similar models) in production to have their own Quality Assurance test sets and whisperkittools offers |
|
the tooling necessary to run the same measurements on such custom test sets, please see the [Model Evaluation on Custom Dataset](#evaluate-on-custom-dataset) for details. |
|
|
|
### Reproducing Results |
|
Results in this page are generated by our cluster of Apple Silicon Macs. We use them as self-hosted runners on |
|
Github Actions as our CI infrastructure. Due to [security concerns](https://docs.github.com/en/actions/security-guides/security-hardening-for-github-actions#hardening-for-self-hosted-runners), |
|
we are unable to open up the cluster to the public. However, any Apple Silicon Mac (even with 8GB RAM) can be used to |
|
run identical [evaluation jobs](#evaluation) locally. For reference, our M2 Ultra devices complete a `librispeech` + `openai/whisper-large-v3` |
|
evaluation in under 1 hour regardless of the Whisper implementation. Older Apple Silicon Macs should take less than 1 day to complete the same evaluation. |
|
|
|
|
|
|
|
Glossary: |
|
|
|
- `_turbo`: Indicates the presence of additional optimizations (not compression) to unlock streaming transcription |
|
as described in our [Blog Post](https://www.takeargmax.com/blog/whisperkit). |
|
|
|
- `_*MB`: Indicates the presence of model compression. Instead of cluttering the filename with details like |
|
`_AudioEncoder-5.8bits_TextDecoder-6.1bits_QLoRA-rank=16`, we choose to summarize the compression spec as the |
|
resulting total file size since this is what matters to developers in production. |
|
|
|
|
