README.md

Wav2Vec2-Large-XLSR-53-ml

Fine-tuned facebook/wav2vec2-large-xlsr-53 on ml using the Indic TTS Malayalam Speech Corpus (via Kaggle), Openslr Malayalam Speech Corpus, SMC Malayalam Speech Corpus. 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:

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

test_dataset = <load-test-split-of-combined-dataset> #TODO

processor = Wav2Vec2Processor.from_pretrained("gvs/wav2vec2-large-xlsr-malayalam")
model = Wav2Vec2ForCTC.from_pretrained("gvs/wav2vec2-large-xlsr-malayalam")

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["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"])

Evaluation

The model can be evaluated as follows on the test data of combined custom dataset. For more details on dataset preparation, check the notebooks mentioned at the end of this file.

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

test_dataset = <load-test-split-of-combined-dataset> #TODO

wer = load_metric("wer")

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

chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"\“\%\‘\”\�Utrnle\_]'
unicode_ignore_regex = r'[\u200d\u200c\u200e]'

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"])
    batch["sentence"] = re.sub(unicode_ignore_regex, '', batch["sentence"])
    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: 39.46 %

Training

A combined dataset was created using Indic TTS Malayalam Speech Corpus (via Kaggle), Openslr Malayalam Speech Corpus, SMC Malayalam Speech Corpus. The datasets were downloaded and was converted to HF Dataset format using this notebook

The notebook used for training and evaluation can be found here