Wav2vec2 Conformer Rel Pos Large 960h Ft

🚀 Wav2Vec2-Conformer-Large-960h with Relative Position Embeddings

Wav2Vec2-Conformer with relative position embeddings, pretrained and fine - tuned on 960 hours of Librispeech for 16kHz sampled speech audio.

Key Information

Model Results

The model's performance on different datasets is as follows:

Paper and Authors

• Paper: fairseq S2T: Fast Speech-to-Text Modeling with fairseq
• Authors: Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Sravya Popuri, Dmytro Okhonko, Juan Pino

The results of Wav2Vec2-Conformer can be found in Table 3 and Table 4 of the official paper. The original model can be found under https://github.com/pytorch/fairseq/tree/master/examples/wav2vec#wav2vec-20.

🚀 Quick Start

When using the model, make sure that your speech input is also sampled at 16Khz.

💻 Usage Examples

Basic Usage

from transformers import Wav2Vec2Processor, Wav2Vec2ConformerForCTC
from datasets import load_dataset
import torch

# load model and processor
processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-conformer-rel-pos-large-960h-ft")
model = Wav2Vec2ConformerForCTC.from_pretrained("facebook/wav2vec2-conformer-rel-pos-large-960h-ft")

# load dummy dataset and read soundfiles
ds = load_dataset("patrickvonplaten/librispeech_asr_dummy", "clean", split="validation")

# tokenize
input_values = processor(ds[0]["audio"]["array"], return_tensors="pt", padding="longest").input_values

# retrieve logits
logits = model(input_values).logits

# take argmax and decode
predicted_ids = torch.argmax(logits, dim=-1)
transcription = processor.batch_decode(predicted_ids)

Advanced Usage

This code snippet shows how to evaluate facebook/wav2vec2-conformer-rel-pos-large-960h-ft on LibriSpeech's "clean" and "other" test data.

from datasets import load_dataset
from transformers import Wav2Vec2ConformerForCTC, Wav2Vec2Processor
import torch
from jiwer import wer

librispeech_eval = load_dataset("librispeech_asr", "clean", split="test")

model = Wav2Vec2ConformerForCTC.from_pretrained("facebook/wav2vec2-large-960h-lv60-self").to("cuda")
processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-large-960h-lv60-self")

def map_to_pred(batch):
    inputs = processor(batch["audio"]["array"], return_tensors="pt", padding="longest")
    input_values = inputs.input_values.to("cuda")
    attention_mask = inputs.attention_mask.to("cuda")

    with torch.no_grad():
        logits = model(input_values, attention_mask=attention_mask).logits

    predicted_ids = torch.argmax(logits, dim=-1)
    transcription = processor.batch_decode(predicted_ids)
    batch["transcription"] = transcription
    return batch

result = librispeech_eval.map(map_to_pred, remove_columns=["audio"])

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

Result (WER):