Updated WER

00ac229 about 3 years ago

No virus

4.73 kB

	---
	language: el
	datasets:
	- common_voice
	metrics:
	- wer
	tags:
	- audio
	- automatic-speech-recognition
	- speech
	- xlsr-fine-tuning-week
	license: apache-2.0
	model-index:
	- name: Greek XLSR Wav2Vec2 Large 53
	results:
	- task:
	name: Speech Recognition
	type: automatic-speech-recognition
	dataset:
	name: Common Voice el
	type: common_voice
	args: el
	metrics:
	- name: Test WER
	type: wer
	value: 45.048955
	---

	# Wav2Vec2-Large-XLSR-53-Greek

	Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Greek using the [Common Voice](https://huggingface.co/datasets/common_voice),
	The Greek CV data has a majority of male voices. To balance it synthesised female voices were created using the approach discussed here [slack](https://huggingface.slack.com/archives/C01QZ90Q83Z/p1616741140114800)
	The text from the common-voice dataset was used to synthesize vocies of female speakers using [Googe's TTS Standard Voice model](https://cloud.google.com/text-to-speech)

	Fine-tuned on facebook/wav2vec2-large-xlsr-53 using Greek CommonVoice :: 5 epochs >> 56.25% WER
	Resuming from checkpoints trained for another 15 epochs >> 34.00%
	Added synthesised female voices trained for 12 epochs >> 34.00% (no change)

	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
	test_dataset = load_dataset("common_voice", "el", split="test[:2%]")

	processor = Wav2Vec2Processor.from_pretrained("skylord/greek_lsr_1")
	model = Wav2Vec2ForCTC.from_pretrained("skylord/greek_lsr_1")

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

	# Preprocessing the datasets.
	# We need to read the aduio 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"][:2])
	```


	## Evaluation

	The model can be evaluated as follows on the Greek test data of Common Voice.


	```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", "el", split="test")
	wer = load_metric("wer")

	processor = Wav2Vec2Processor.from_pretrained("skylord/greek_lsr_1")
	model = Wav2Vec2ForCTC.from_pretrained("skylord/greek_lsr_1")
	model.to("cuda")

	chars_to_ignore_regex = '[\\\\\\\\,\\\\\\\\?\\\\\\\\.\\\\\\\\!\\\\\\\\-\\\\\\\\;\\\\\\\\:\\\\\\\\"\\\\\\\\“]'
	resampler = torchaudio.transforms.Resample(48_000, 16_000)

	# Preprocessing the datasets.
	# We need to read the aduio 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 aduio 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: 45.048955 %


	## Training

	The Common Voice `train`, `validation`, datasets were used for training as well as

	The script used for training can be found [here](...) # TODO: fill in a link to your training script here. If you trained your model in a colab, simply fill in the link here. If you trained the model locally, it would be great if you could upload the training script on github and paste the link here.