Assignment1 Francesco

🚀 S2T-SMALL-LIBRISPEECH-ASR

s2t-small-librispeech-asr is a Speech to Text Transformer (S2T) model trained for automatic speech recognition (ASR). It offers an efficient solution for converting speech to text, leveraging the power of the Transformer architecture.

🚀 Quick Start

The S2T model was proposed in this paper and released in this repository.

✨ Features

• End - to - End Model: S2T is an end-to-end sequence-to-sequence transformer model. It is trained with standard autoregressive cross-entropy loss and generates the transcripts autoregressively.
• Versatile Usage: This model can be used for end-to-end speech recognition (ASR). See the model hub to look for other S2T checkpoints.

📦 Installation

Note: The Speech2TextProcessor object uses torchaudio to extract the filter bank features. Make sure to install the torchaudio package before running this example.

Note: The feature extractor depends on torchaudio and the tokenizer depends on sentencepiece so be sure to install those packages before running the examples.

You could either install those as extra speech dependancies with pip install transformers"[speech, sentencepiece]" or install the packages seperatly with pip install torchaudio sentencepiece.

💻 Usage Examples

Basic Usage

import torch
from transformers import Speech2TextProcessor, Speech2TextForConditionalGeneration
from datasets import load_dataset

model = Speech2TextForConditionalGeneration.from_pretrained("facebook/s2t-small-librispeech-asr")
processor = Speech2TextProcessor.from_pretrained("facebook/s2t-small-librispeech-asr")

ds = load_dataset(
    "patrickvonplaten/librispeech_asr_dummy",
    "clean",
    split="validation"
)

input_features = processor(
    ds[0]["audio"]["array"],
    sampling_rate=16_000,
    return_tensors="pt"
).input_features  # Batch size 1
generated_ids = model.generate(input_ids=input_features)

transcription = processor.batch_decode(generated_ids)

Advanced Usage

The following script shows how to evaluate this model on the LibriSpeech "clean" and "other" test dataset.

from datasets import load_dataset, load_metric
from transformers import Speech2TextForConditionalGeneration, Speech2TextProcessor

librispeech_eval = load_dataset("librispeech_asr", "clean", split="test")  # change to "other" for other test dataset
wer = load_metric("wer")

model = Speech2TextForConditionalGeneration.from_pretrained("facebook/s2t-small-librispeech-asr").to("cuda")
processor = Speech2TextProcessor.from_pretrained("facebook/s2t-small-librispeech-asr", do_upper_case=True)

librispeech_eval = librispeech_eval.map(map_to_array)

def map_to_pred(batch):
    features = processor(batch["audio"]["array"], sampling_rate=16000, padding=True, return_tensors="pt")
    input_features = features.input_features.to("cuda")
    attention_mask = features.attention_mask.to("cuda")

    gen_tokens = model.generate(input_ids=input_features, attention_mask=attention_mask)
    batch["transcription"] = processor.batch_decode(gen_tokens, skip_special_tokens=True)
    return batch

result = librispeech_eval.map(map_to_pred, batched=True, batch_size=8, remove_columns=["speech"])

print("WER:", wer(predictions=result["transcription"], references=result["text"]))

Result (WER):

📚 Documentation

Training data

The S2T-SMALL-LIBRISPEECH-ASR is trained on LibriSpeech ASR Corpus, a dataset consisting of approximately 1000 hours of 16kHz read English speech.

Training procedure

Preprocessing

The speech data is pre-processed by extracting Kaldi-compliant 80-channel log mel-filter bank features automatically from WAV/FLAC audio files via PyKaldi or torchaudio. Further utterance-level CMVN (cepstral mean and variance normalization) is applied to each example.

The texts are lowercased and tokenized using SentencePiece and a vocabulary size of 10,000.

Training

The model is trained with standard autoregressive cross-entropy loss and using SpecAugment. The encoder receives speech features, and the decoder generates the transcripts autoregressively.

BibTeX entry and citation info

@inproceedings{wang2020fairseqs2t,
  title = {fairseq S2T: Fast Speech-to-Text Modeling with fairseq},
  author = {Changhan Wang and Yun Tang and Xutai Ma and Anne Wu and Dmytro Okhonko and Juan Pino},
  booktitle = {Proceedings of the 2020 Conference of the Asian Chapter of the Association for Computational Linguistics (AACL): System Demonstrations},
  year = {2020},
}

🔧 Technical Details

The model is an end-to-end sequence-to-sequence transformer. It uses standard autoregressive cross - entropy loss for training. The speech data is pre - processed to extract 80 - channel log mel - filter bank features and CMVN is applied. Texts are lowercased and tokenized with SentencePiece. During training, SpecAugment is used to improve the model's performance.

📄 License

This project is licensed under the MIT license.

⚠️ Important Note

The Speech2TextProcessor object uses torchaudio to extract the filter bank features. Make sure to install the torchaudio package before running this example.

💡 Usage Tip

The feature extractor depends on torchaudio and the tokenizer depends on sentencepiece so be sure to install those packages before running the examples. You could either install those as extra speech dependancies with pip install transformers"[speech, sentencepiece]" or install the packages seperatly with pip install torchaudio sentencepiece.