anuragshas/wav2vec2-xls-r-300m-mt-cv8-with-lm

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

XLS-R-300M - Maltese

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA-FOUNDATION/COMMONVOICE8_0 - MT dataset. It achieves the following results on the evaluation set:

• —Loss: 0.1895
• —Wer: 0.1984

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• —learning_rate: 7.5e-05
• —trainbatchsize: 32
• —evalbatchsize: 16
• —seed: 42
• —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• —lrschedulertype: linear
• —lrschedulerwarmup_steps: 1000
• —num_epochs: 60.0
• —mixedprecisiontraining: Native AMP

Training results

Framework versions

• —Transformers 4.17.0.dev0
• —Pytorch 1.10.2+cu102
• —Datasets 1.18.2.dev0
• —Tokenizers 0.11.0

Evaluation Commands

1. 1.To evaluate on mozilla-foundation/common_voice_8_0 with split test

python eval.py --model_id anuragshas/wav2vec2-xls-r-300m-mt-cv8-with-lm --dataset mozilla-foundation/common_voice_8_0 --config mt --split test

Inference With LM

import torch
from datasets import load_dataset
from transformers import AutoModelForCTC, AutoProcessor
import torchaudio.functional as F
model_id = "anuragshas/wav2vec2-xls-r-300m-mt-cv8-with-lm"
sample_iter = iter(load_dataset("mozilla-foundation/common_voice_8_0", "mt", split="test", streaming=True, use_auth_token=True))
sample = next(sample_iter)
resampled_audio = F.resample(torch.tensor(sample["audio"]["array"]), 48_000, 16_000).numpy()
model = AutoModelForCTC.from_pretrained(model_id)
processor = AutoProcessor.from_pretrained(model_id)
input_values = processor(resampled_audio, return_tensors="pt").input_values
with torch.no_grad():
    logits = model(input_values).logits
transcription = processor.batch_decode(logits.numpy()).text
# => "għadu jilagħbu ċirku tant bilfondi"

Eval results on Common Voice 8 "test" (WER):