đ Wav2Vec2-Large-100k-VoxPopuli-Portuguese
This model is fine-tuned from facebook/wav2vec2-large-100k-voxpopuli on Portuguese using the Common Voice dataset, aiming to provide high - quality automatic speech recognition for Portuguese.
đ Quick Start
This model is fine-tuned from facebook/wav2vec2-large-100k-voxpopuli on Portuguese using the Common Voice dataset.
⨠Features
- Audio Processing: Specialized for audio and speech processing.
- Language Specific: Fine - tuned for Portuguese, suitable for Portuguese speech recognition tasks.
- Model Type: Based on the Wav2Vec2 architecture.
Property |
Details |
Model Type |
Wav2Vec2 - Large - 100k - VoxPopuli - Portuguese |
Training Data |
Common Voice Portuguese dataset |
đĻ Installation
To use this model, you need to install the necessary Python libraries. You can install them using the following command:
pip install torch torchaudio datasets transformers
If you want to evaluate the model, you also need to install Enelvo:
pip install enelvo
đģ Usage Examples
Basic Usage
The model can be used directly (without a language model) as follows:
import torch
import torchaudio
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
test_dataset = load_dataset("common_voice", "pt", split="test[:2%]")
processor = Wav2Vec2Processor.from_pretrained("joorock12/wav2vec2-large-100k-voxpopuli-pt")
model = Wav2Vec2ForCTC.from_pretrained("joorock12/wav2vec2-large-100k-voxpopuli-pt")
resampler = torchaudio.transforms.Resample(48_000, 16_000)
def speech_file_to_array_fn(batch):
speech_array, sampling_rate = torchaudio.load(batch["path"])
batch["speech"] = resampler(speech_array).squeeze().numpy()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
predicted_ids = torch.argmax(logits, dim=-1)
print("Prediction:", processor.batch_decode(predicted_ids))
print("Reference:", test_dataset["sentence"][:2])
Advanced Usage
The model can be evaluated as follows on the Portuguese test data of Common Voice.
import torch
import torchaudio
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
from enelvo import normaliser
import re
test_dataset = load_dataset("common_voice", "pt", split="test")
wer = load_metric("wer")
processor = Wav2Vec2Processor.from_pretrained("joorock12/wav2vec2-large-100k-voxpopuli-pt")
model = Wav2Vec2ForCTC.from_pretrained("joorock12/wav2vec2-large-100k-voxpopuli-pt")
model.to("cuda")
chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"\â\%\â\â\īŋŊ]'
resampler = torchaudio.transforms.Resample(48_000, 16_000)
norm = normaliser.Normaliser()
def speech_file_to_array_fn(batch):
batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
speech_array, sampling_rate = torchaudio.load(batch["path"])
batch["speech"] = resampler(speech_array).squeeze().numpy()
return batch
test_dataset = test_dataset.map(speech_file_to_array_fn)
def evaluate(batch):
inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
with torch.no_grad():
logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
pred_ids = torch.argmax(logits, dim=-1)
batch["pred_strings"] = [norm.normalise(i) for i in processor.batch_decode(pred_ids)]
return batch
result = test_dataset.map(evaluate, batched=True, batch_size=8)
print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
Test Result (wer): 19.735723%
đ§ Technical Details
The Common Voice train
, validation
datasets were used for training.
đ License
This model is licensed under the Apache - 2.0 license.