openai/whisper-large-v3-turbo
openai/whisper-large-v3-turbo
2
1
Overview Code Discussions Pull Requests Wiki Activity

Update README.md

Browse Source
This commit is contained in:
Yoach Lacombe 2024-10-01 07:51:28 +00:00 committed by system
parent 0631025cc0
commit 7a516fab70
No known key found for this signature in database
GPG Key ID: 6A528E38E0733467
1 changed files with 436 additions and 103 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

539
README.md
View File

@ -1,199 +1,532 @@
---
library_name: transformers
tags: []
language:
- en
- zh
- de
- es
- ru
- ko
- fr
- ja
- pt
- tr
- pl
- ca
- nl
- ar
- sv
- it
- id
- hi
- fi
- vi
- he
- uk
- el
- ms
- cs
- ro
- da
- hu
- ta
- no
- th
- ur
- hr
- bg
- lt
- la
- mi
- ml
- cy
- sk
- te
- fa
- lv
- bn
- sr
- az
- sl
- kn
- et
- mk
- br
- eu
- is
- hy
- ne
- mn
- bs
- kk
- sq
- sw
- gl
- mr
- pa
- si
- km
- sn
- yo
- so
- af
- oc
- ka
- be
- tg
- sd
- gu
- am
- yi
- lo
- uz
- fo
- ht
- ps
- tk
- nn
- mt
- sa
- lb
- my
- bo
- tl
- mg
- as
- tt
- haw
- ln
- ha
- ba
- jw
- su
tags:
- audio
- automatic-speech-recognition
- hf-asr-leaderboard
widget:
- example_title: Librispeech sample 1
src: https://cdn-media.huggingface.co/speech_samples/sample1.flac
- example_title: Librispeech sample 2
src: https://cdn-media.huggingface.co/speech_samples/sample2.flac
pipeline_tag: automatic-speech-recognition
license: apache-2.0
---
# Model Card for Model ID
# Whisper
<!-- Provide a quick summary of what the model is/does. -->
Whisper is a state-of-the-art model for automatic speech recognition (ASR) and speech translation, proposed in the paper
[Robust Speech Recognition via Large-Scale Weak Supervision](https://huggingface.co/papers/2212.04356) by Alec Radford
et al. from OpenAI. Trained on >5M hours of labeled data, Whisper demonstrates a strong ability to generalise to many
datasets and domains in a zero-shot setting.
Whisper large-v3-turbo is a distilled version of [Whisper large-v3](https://huggingface.co/openai/whisper-large-v3). In other words, it's the exact same model, except that the number of decoding layers have reduced from 32 to 4.
As a result, the model is way faster, at the expense of a minor quality degradation.
**Disclaimer**: Content for this model card has partly been written by the 🤗 Hugging Face team, and partly copied and
pasted from the original model card.
## Model Details
## Usage
### Model Description
Whisper large-v3-turbo is supported in Hugging Face 🤗 Transformers. To run the model, first install the Transformers
library. For this example, we'll also install 🤗 Datasets to load toy audio dataset from the Hugging Face Hub, and
🤗 Accelerate to reduce the model loading time:
<!-- Provide a longer summary of what this model is. -->
```bash
pip install --upgrade pip
pip install --upgrade transformers datasets[audio] accelerate
```
This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
The model can be used with the [`pipeline`](https://huggingface.co/docs/transformers/main_classes/pipelines#transformers.AutomaticSpeechRecognitionPipeline)
class to transcribe audios of arbitrary length:
- **Developed by:** [More Information Needed]
- **Funded by [optional]:** [More Information Needed]
- **Shared by [optional]:** [More Information Needed]
- **Model type:** [More Information Needed]
- **Language(s) (NLP):** [More Information Needed]
- **License:** [More Information Needed]
- **Finetuned from model [optional]:** [More Information Needed]
```python
import torch
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
from datasets import load_dataset
### Model Sources [optional]
<!-- Provide the basic links for the model. -->
device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
- **Repository:** [More Information Needed]
- **Paper [optional]:** [More Information Needed]
- **Demo [optional]:** [More Information Needed]
model_id = "openai/whisper-large-v3-turbo"
## Uses
model = AutoModelForSpeechSeq2Seq.from_pretrained(
model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True
)
model.to(device)
<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
processor = AutoProcessor.from_pretrained(model_id)
### Direct Use
pipe = pipeline(
"automatic-speech-recognition",
model=model,
tokenizer=processor.tokenizer,
feature_extractor=processor.feature_extractor,
torch_dtype=torch_dtype,
device=device,
)
<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
dataset = load_dataset("distil-whisper/librispeech_long", "clean", split="validation")
sample = dataset[0]["audio"]
[More Information Needed]
result = pipe(sample)
print(result["text"])
```
### Downstream Use [optional]
To transcribe a local audio file, simply pass the path to your audio file when you call the pipeline:
<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
```python
result = pipe("audio.mp3")
```
[More Information Needed]
Multiple audio files can be transcribed in parallel by specifying them as a list and setting the `batch_size` parameter:
### Out-of-Scope Use
```python
result = pipe(["audio_1.mp3", "audio_2.mp3"], batch_size=2)
```
<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
Transformers is compatible with all Whisper decoding strategies, such as temperature fallback and condition on previous
tokens. The following example demonstrates how to enable these heuristics:
[More Information Needed]
```python
generate_kwargs = {
"max_new_tokens": 448,
"num_beams": 1,
"condition_on_prev_tokens": False,
"compression_ratio_threshold": 1.35, # zlib compression ratio threshold (in token space)
"temperature": (0.0, 0.2, 0.4, 0.6, 0.8, 1.0),
"logprob_threshold": -1.0,
"no_speech_threshold": 0.6,
"return_timestamps": True,
}
## Bias, Risks, and Limitations
result = pipe(sample, generate_kwargs=generate_kwargs)
```
<!-- This section is meant to convey both technical and sociotechnical limitations. -->
Whisper predicts the language of the source audio automatically. If the source audio language is known *a-priori*, it
can be passed as an argument to the pipeline:
[More Information Needed]
```python
result = pipe(sample, generate_kwargs={"language": "english"})
```
### Recommendations
By default, Whisper performs the task of *speech transcription*, where the source audio language is the same as the target
text language. To perform *speech translation*, where the target text is in English, set the task to `"translate"`:
<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
```python
result = pipe(sample, generate_kwargs={"task": "translate"})
```
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
Finally, the model can be made to predict timestamps. For sentence-level timestamps, pass the `return_timestamps` argument:
## How to Get Started with the Model
```python
result = pipe(sample, return_timestamps=True)
print(result["chunks"])
```
Use the code below to get started with the model.
And for word-level timestamps:
[More Information Needed]
```python
result = pipe(sample, return_timestamps="word")
print(result["chunks"])
```
## Training Details
The above arguments can be used in isolation or in combination. For example, to perform the task of speech transcription
where the source audio is in French, and we want to return sentence-level timestamps, the following can be used:
### Training Data
```python
result = pipe(sample, return_timestamps=True, generate_kwargs={"language": "french", "task": "translate"})
print(result["chunks"])
```
<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
<details>
[More Information Needed]
<summary> For more control over the generation parameters, use the model + processor API directly: </summary>
### Training Procedure
```python
import torch
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor
from datasets import Audio, load_dataset
<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
#### Preprocessing [optional]
device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
[More Information Needed]
model_id = "openai/whisper-large-v3-turbo"
model = AutoModelForSpeechSeq2Seq.from_pretrained(
model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True
)
model.to(device)
#### Training Hyperparameters
processor = AutoProcessor.from_pretrained(model_id)
- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
dataset = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")
dataset = dataset.cast_column("audio", Audio(processor.feature_extractor.sampling_rate))
sample = dataset[0]["audio"]
#### Speeds, Sizes, Times [optional]
inputs = processor(
sample["array"],
sampling_rate=sample["sampling_rate"],
return_tensors="pt",
truncation=False,
padding="longest",
return_attention_mask=True,
)
inputs = inputs.to(device, dtype=torch_dtype)
<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
gen_kwargs = {
"max_new_tokens": 448,
"num_beams": 1,
"condition_on_prev_tokens": False,
"compression_ratio_threshold": 1.35, # zlib compression ratio threshold (in token space)
"temperature": (0.0, 0.2, 0.4, 0.6, 0.8, 1.0),
"logprob_threshold": -1.0,
"no_speech_threshold": 0.6,
"return_timestamps": True,
}
[More Information Needed]
pred_ids = model.generate(**inputs, **gen_kwargs)
pred_text = processor.batch_decode(pred_ids, skip_special_tokens=True, decode_with_timestamps=False)
## Evaluation
print(pred_text)
```
<!-- This section describes the evaluation protocols and provides the results. -->
</details>
### Testing Data, Factors & Metrics
## Additional Speed & Memory Improvements
#### Testing Data
You can apply additional speed and memory improvements to Whisper to further reduce the inference speed and VRAM
requirements.
<!-- This should link to a Dataset Card if possible. -->
### Chunked Long-Form
[More Information Needed]
Whisper has a receptive field of 30-seconds. To transcribe audios longer than this, one of two long-form algorithms are
required:
1. **Sequential:** uses a "sliding window" for buffered inference, transcribing 30-second slices one after the other
2. **Chunked:** splits long audio files into shorter ones (with a small overlap between segments), transcribes each segment independently, and stitches the resulting transcriptions at the boundaries
#### Factors
The sequential long-form algorithm should be used in either of the following scenarios:
1. Transcription accuracy is the most important factor, and speed is less of a consideration
2. You are transcribing **batches** of long audio files, in which case the latency of sequential is comparable to chunked, while being up to 0.5% WER more accurate
<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
Conversely, the chunked algorithm should be used when:
1. Transcription speed is the most important factor
2. You are transcribing a **single** long audio file
[More Information Needed]
By default, Transformers uses the sequential algorithm. To enable the chunked algorithm, pass the `chunk_length_s`
parameter to the `pipeline`. For large-v3, a chunk length of 30-seconds is optimal. To activate batching over long
audio files, pass the argument `batch_size`:
#### Metrics
```python
import torch
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
from datasets import load_dataset
<!-- These are the evaluation metrics being used, ideally with a description of why. -->
[More Information Needed]
device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
### Results
model_id = "openai/whisper-large-v3-turbo"
[More Information Needed]
model = AutoModelForSpeechSeq2Seq.from_pretrained(
model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True
)
model.to(device)
#### Summary
processor = AutoProcessor.from_pretrained(model_id)
pipe = pipeline(
"automatic-speech-recognition",
model=model,
tokenizer=processor.tokenizer,
feature_extractor=processor.feature_extractor,
chunk_length_s=30,
batch_size=16, # batch size for inference - set based on your device
torch_dtype=torch_dtype,
device=device,
)
dataset = load_dataset("distil-whisper/librispeech_long", "clean", split="validation")
sample = dataset[0]["audio"]
## Model Examination [optional]
result = pipe(sample)
print(result["text"])
```
<!-- Relevant interpretability work for the model goes here -->
#### Torch compile
[More Information Needed]
The Whisper forward pass is compatible with [`torch.compile`](https://pytorch.org/docs/stable/generated/torch.compile.html)
for 4.5x speed-ups.
## Environmental Impact
**Note:** `torch.compile` is currently not compatible with the Chunked long-form algorithm or Flash Attention 2 ⚠️
<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
```python
import torch
from torch.nn.attention import SDPBackend, sdpa_kernel
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
from datasets import load_dataset
from tqdm import tqdm
Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
torch.set_float32_matmul_precision("high")
- **Hardware Type:** [More Information Needed]
- **Hours used:** [More Information Needed]
- **Cloud Provider:** [More Information Needed]
- **Compute Region:** [More Information Needed]
- **Carbon Emitted:** [More Information Needed]
device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
## Technical Specifications [optional]
model_id = "openai/whisper-large-v3-turbo"
### Model Architecture and Objective
model = AutoModelForSpeechSeq2Seq.from_pretrained(
model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True
).to(device)
[More Information Needed]
# Enable static cache and compile the forward pass
model.generation_config.cache_implementation = "static"
model.generation_config.max_new_tokens = 256
model.forward = torch.compile(model.forward, mode="reduce-overhead", fullgraph=True)
### Compute Infrastructure
processor = AutoProcessor.from_pretrained(model_id)
[More Information Needed]
pipe = pipeline(
"automatic-speech-recognition",
model=model,
tokenizer=processor.tokenizer,
feature_extractor=processor.feature_extractor,
torch_dtype=torch_dtype,
device=device,
)
#### Hardware
dataset = load_dataset("distil-whisper/librispeech_long", "clean", split="validation")
sample = dataset[0]["audio"]
[More Information Needed]
# 2 warmup steps
for _ in tqdm(range(2), desc="Warm-up step"):
with sdpa_kernel(SDPBackend.MATH):
result = pipe(sample.copy(), generate_kwargs={"min_new_tokens": 256, "max_new_tokens": 256})
#### Software
# fast run
with sdpa_kernel(SDPBackend.MATH):
result = pipe(sample.copy())
[More Information Needed]
print(result["text"])
```
## Citation [optional]
#### Flash Attention 2
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
We recommend using [Flash-Attention 2](https://huggingface.co/docs/transformers/main/en/perf_infer_gpu_one#flashattention-2) if your GPU supports it and you are not using [torch.compile](#torch-compile).
To do so, first install [Flash Attention](https://github.com/Dao-AILab/flash-attention):
**BibTeX:**
```
pip install flash-attn --no-build-isolation
```
[More Information Needed]
Then pass `attn_implementation="flash_attention_2"` to `from_pretrained`:
**APA:**
```python
model = AutoModelForSpeechSeq2Seq.from_pretrained(model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, attn_implementation="flash_attention_2")
```
[More Information Needed]
#### Torch Scale-Product-Attention (SDPA)
## Glossary [optional]
If your GPU does not support Flash Attention, we recommend making use of PyTorch [scaled dot-product attention (SDPA)](https://pytorch.org/docs/stable/generated/torch.nn.functional.scaled_dot_product_attention.html).
This attention implementation is activated **by default** for PyTorch versions 2.1.1 or greater. To check
whether you have a compatible PyTorch version, run the following Python code snippet:
<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
```python
from transformers.utils import is_torch_sdpa_available
[More Information Needed]
print(is_torch_sdpa_available())
```
## More Information [optional]
If the above returns `True`, you have a valid version of PyTorch installed and SDPA is activated by default. If it
returns `False`, you need to upgrade your PyTorch version according to the [official instructions](https://pytorch.org/get-started/locally/)
[More Information Needed]
Once a valid PyTorch version is installed, SDPA is activated by default. It can also be set explicitly by specifying
`attn_implementation="sdpa"` as follows:
## Model Card Authors [optional]
```python
model = AutoModelForSpeechSeq2Seq.from_pretrained(model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, attn_implementation="sdpa")
```
[More Information Needed]
For more information about how to use the SDPA refer to the [Transformers SDPA documentation](https://huggingface.co/docs/transformers/en/perf_infer_gpu_one#pytorch-scaled-dot-product-attention).
## Model Card Contact
[More Information Needed]
## Model details
Whisper is a Transformer based encoder-decoder model, also referred to as a _sequence-to-sequence_ model. There are two
flavours of Whisper model: English-only and multilingual. The English-only models were trained on the task of English
speech recognition. The multilingual models were trained simultaneously on multilingual speech recognition and speech
translation. For speech recognition, the model predicts transcriptions in the *same* language as the audio. For speech
translation, the model predicts transcriptions to a *different* language to the audio.
Whisper checkpoints come in five configurations of varying model sizes. The smallest four are available as English-only
and multilingual. The largest checkpoints are multilingual only. All ten of the pre-trained checkpoints
are available on the [Hugging Face Hub](https://huggingface.co/models?search=openai/whisper). The
checkpoints are summarised in the following table with links to the models on the Hub:
| Size | Parameters | English-only | Multilingual |
|----------|------------|------------------------------------------------------|-----------------------------------------------------|
| tiny | 39 M | [✓](https://huggingface.co/openai/whisper-tiny.en) | [✓](https://huggingface.co/openai/whisper-tiny) |
| base | 74 M | [✓](https://huggingface.co/openai/whisper-base.en) | [✓](https://huggingface.co/openai/whisper-base) |
| small | 244 M | [✓](https://huggingface.co/openai/whisper-small.en) | [✓](https://huggingface.co/openai/whisper-small) |
| medium | 769 M | [✓](https://huggingface.co/openai/whisper-medium.en) | [✓](https://huggingface.co/openai/whisper-medium) |
| large | 1550 M | x | [✓](https://huggingface.co/openai/whisper-large) |
| large-v2 | 1550 M | x | [✓](https://huggingface.co/openai/whisper-large-v2) |
| large-v3 | 1550 M | x | [✓](https://huggingface.co/openai/whisper-large-v3) |
| large-v3-turbo | 809 M | x | [✓](https://huggingface.co/openai/whisper-large-v3-turbo) |
## Fine-Tuning
The pre-trained Whisper model demonstrates a strong ability to generalise to different datasets and domains. However,
its predictive capabilities can be improved further for certain languages and tasks through *fine-tuning*. The blog
post [Fine-Tune Whisper with 🤗 Transformers](https://huggingface.co/blog/fine-tune-whisper) provides a step-by-step
guide to fine-tuning the Whisper model with as little as 5 hours of labelled data.
### Evaluated Use
The primary intended users of these models are AI researchers studying robustness, generalization, capabilities, biases, and constraints of the current model. However, Whisper is also potentially quite useful as an ASR solution for developers, especially for English speech recognition. We recognize that once models are released, it is impossible to restrict access to only “intended” uses or to draw reasonable guidelines around what is or is not research.
The models are primarily trained and evaluated on ASR and speech translation to English tasks. They show strong ASR results in ~10 languages. They may exhibit additional capabilities, particularly if fine-tuned on certain tasks like voice activity detection, speaker classification, or speaker diarization but have not been robustly evaluated in these areas. We strongly recommend that users perform robust evaluations of the models in a particular context and domain before deploying them.
In particular, we caution against using Whisper models to transcribe recordings of individuals taken without their consent or purporting to use these models for any kind of subjective classification. We recommend against use in high-risk domains like decision-making contexts, where flaws in accuracy can lead to pronounced flaws in outcomes. The models are intended to transcribe and translate speech, use of the model for classification is not only not evaluated but also not appropriate, particularly to infer human attributes.
## Training Data
TODO
The large-v3 checkpoint is trained on 1 million hours of weakly labeled audio and 4 million hours of pseudo-labeled audio collected using Whisper large-v2.
As discussed in [the accompanying paper](https://cdn.openai.com/papers/whisper.pdf), we see that performance on transcription in a given language is directly correlated with the amount of training data we employ in that language.
## Performance and Limitations
Our studies show that, over many existing ASR systems, the models exhibit improved robustness to accents, background noise, technical language, as well as zero shot translation from multiple languages into English; and that accuracy on speech recognition and translation is near the state-of-the-art level.
However, because the models are trained in a weakly supervised manner using large-scale noisy data, the predictions may include texts that are not actually spoken in the audio input (i.e. hallucination). We hypothesize that this happens because, given their general knowledge of language, the models combine trying to predict the next word in audio with trying to transcribe the audio itself.
Our models perform unevenly across languages, and we observe lower accuracy on low-resource and/or low-discoverability languages or languages where we have less training data. The models also exhibit disparate performance on different accents and dialects of particular languages, which may include higher word error rate across speakers of different genders, races, ages, or other demographic criteria. Our full evaluation results are presented in [the paper accompanying this release](https://cdn.openai.com/papers/whisper.pdf).
In addition, the sequence-to-sequence architecture of the model makes it prone to generating repetitive texts, which can be mitigated to some degree by beam search and temperature scheduling but not perfectly. Further analysis on these limitations are provided in [the paper](https://cdn.openai.com/papers/whisper.pdf). It is likely that this behavior and hallucinations may be worse on lower-resource and/or lower-discoverability languages.
## Broader Implications
We anticipate that Whisper models’ transcription capabilities may be used for improving accessibility tools. While Whisper models cannot be used for real-time transcription out of the box – their speed and size suggest that others may be able to build applications on top of them that allow for near-real-time speech recognition and translation. The real value of beneficial applications built on top of Whisper models suggests that the disparate performance of these models may have real economic implications.
There are also potential dual use concerns that come with releasing Whisper. While we hope the technology will be used primarily for beneficial purposes, making ASR technology more accessible could enable more actors to build capable surveillance technologies or scale up existing surveillance efforts, as the speed and accuracy allow for affordable automatic transcription and translation of large volumes of audio communication. Moreover, these models may have some capabilities to recognize specific individuals out of the box, which in turn presents safety concerns related both to dual use and disparate performance. In practice, we expect that the cost of transcription is not the limiting factor of scaling up surveillance projects.
### BibTeX entry and citation info
```bibtex
@misc{radford2022whisper,
doi = {10.48550/ARXIV.2212.04356},
url = {https://arxiv.org/abs/2212.04356},
author = {Radford, Alec and Kim, Jong Wook and Xu, Tao and Brockman, Greg and McLeavey, Christine and Sutskever, Ilya},
title = {Robust Speech Recognition via Large-Scale Weak Supervision},
publisher = {arXiv},
year = {2022},
copyright = {arXiv.org perpetual, non-exclusive license}
}
```