---
language: rw
datasets:
- common_voice
metrics:
- wer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 Large Kinyarwanda by Lucio
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice rw
type: common_voice
args: rw
metrics:
- name: Test WER
type: wer
value: ??
---

# Wav2Vec2-Large-XLSR-53-rw

Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Kinyarwanda using the [Common Voice](https://huggingface.co/datasets/common_voice) dataset, using validation data as the train set, and taking 12% of the test data for validation.
When using this model, make sure that your speech input is sampled at 16kHz.

## Usage

The model can be used directly (without a language model) as follows:

```python
import torch
import torchaudio
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor

# WARNING: This will download 40GB of data, and then extract it as another 40GB
test_dataset = load_dataset("common_voice", "rw", split="test[:2%]")

processor = Wav2Vec2Processor.from_pretrained("lucio/wav2vec2-large-xlsr-kinyarwanda")
model = Wav2Vec2ForCTC.from_pretrained("lucio/wav2vec2-large-xlsr-kinyarwanda")

resampler = torchaudio.transforms.Resample(48_000, 16_000)

# Preprocessing the datasets.
# We need to read the audio files as arrays
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[:2]["speech"], 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])
```


## Evaluation

The model can be evaluated as follows on the Kinyarwanda test data of Common Voice, though you will need ~80GB of disk to download and extract the whole corpus.


```python
import torch
import torchaudio
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import re

test_dataset = load_dataset("common_voice", "rw", split="test")
wer = load_metric("wer")

processor = Wav2Vec2Processor.from_pretrained("lucio/wav2vec2-large-xlsr-kinyarwanda")
model = Wav2Vec2ForCTC.from_pretrained("lucio/wav2vec2-large-xlsr-kinyarwanda")
model.to("cuda")

chars_to_ignore_regex = '[\[\] ,?.!;:%"‘’“”(){}‟ˮ´ʺ″«»/…‽�–-]'
resampler = torchaudio.transforms.Resample(48_000, 16_000)

# Preprocessing the datasets.
# We need to read the audio files as arrays
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)

# Preprocessing the datasets.
# We need to read the audio files as arrays
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"] = 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**: ??

## Training

The Common Voice `train`, `validation` and `other` datasets were used for training, with the additional filter applied to remove `other` data that did not have more up votes than down votes (train+validation+other[upvotes > downvotes]).

The script used for training was just the `run_finetuning.py` script provided in OVHcloud's databuzzword/hf-wav2vec image.