anas/wav2vec2-large-xlsr-arabic
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

初始化项目，由ModelHub XC社区提供模型

Browse Source 
Model: anas/wav2vec2-large-xlsr-arabic
Source: Original Platform
This commit is contained in:
ModelHub XC
2026-05-28 11:02:18 +08:00
commit 9515e23928
14 changed files with 409 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
.gitattributes vendored Normal file
View File

@@ -0,0 +1,17 @@
*.bin.* filter=lfs diff=lfs merge=lfs -text
*.lfs.* filter=lfs diff=lfs merge=lfs -text
*.bin filter=lfs diff=lfs merge=lfs -text
*.h5 filter=lfs diff=lfs merge=lfs -text
*.tflite filter=lfs diff=lfs merge=lfs -text
*.tar.gz filter=lfs diff=lfs merge=lfs -text
*.ot filter=lfs diff=lfs merge=lfs -text
*.onnx filter=lfs diff=lfs merge=lfs -text
*.arrow filter=lfs diff=lfs merge=lfs -text
*.ftz filter=lfs diff=lfs merge=lfs -text
*.joblib filter=lfs diff=lfs merge=lfs -text
*.model filter=lfs diff=lfs merge=lfs -text
*.msgpack filter=lfs diff=lfs merge=lfs -text
*.pb filter=lfs diff=lfs merge=lfs -text
*.pt filter=lfs diff=lfs merge=lfs -text
*.pth filter=lfs diff=lfs merge=lfs -text
*.msgpack filter=lfs diff=lfs merge=lfs -text

144
.ipynb_checkpoints/README-checkpoint.md Normal file
View File

@@ -0,0 +1,144 @@
---
language: ar
datasets:
- common_voice: Common Voice Corpus 4
metrics:
- wer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: Hasni XLSR Wav2Vec2 Large 53
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice ar
type: common_voice
args: ar
metrics:
- name: Test WER
type: wer
value: 52.18
---
# Wav2Vec2-Large-XLSR-53-Arabic
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Arabic using the [Common Voice Corpus 4](https://commonvoice.mozilla.org/en/datasets) dataset.
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", "ar", split="test[:2%]")
processor = Wav2Vec2Processor.from_pretrained("anas/wav2vec2-large-xlsr-arabic")
model = Wav2Vec2ForCTC.from_pretrained("anas/wav2vec2-large-xlsr-arabic")
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 Arabic 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", "ar", split="test")
processor = Wav2Vec2Processor.from_pretrained("anas/wav2vec2-large-xlsr-arabic")
model = Wav2Vec2ForCTC.from_pretrained("anas/wav2vec2-large-xlsr-arabic/")
model.to("cuda")
chars_to_ignore_regex = '[\\\\,\\\\؟\\\\.\\\\!\\\\-\\\\;\\\\\\\\:\\\\'\\\\"\\\\☭\\\\«\\\\»\\\\؛\\\\—\\\\ـ\\\\_\\\\،\\\\“\\\\%\\\\‘\\\\”\\\\<EFBFBD>]'
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()
batch["sentence"] = re.sub('[a-z]','',batch["sentence"])
batch["sentence"] = re.sub("[إأٱآا]", "ا", batch["sentence"])
noise = re.compile(""" ّ | # Tashdid
َ | # Fatha
ً | # Tanwin Fath
ُ | # Damma
ٌ | # Tanwin Damm
ِ | # Kasra
ٍ | # Tanwin Kasr
ْ | # Sukun
ـ # Tatwil/Kashida
""", re.VERBOSE)
batch["sentence"] = re.sub(noise, '', 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 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**: 52.18 %
## Training
The Common Voice Corpus 4 `train`, `validation`, datasets were used for training
The script used for training can be found [here](...)
Twitter: [here](https://twitter.com/hasnii_anas)
Email: anashasni146@gmail.com

8
.ipynb_checkpoints/preprocessor_config-checkpoint.json Normal file
View File

@@ -0,0 +1,8 @@
{
"do_normalize": true,
"feature_size": 1,
"padding_side": "right",
"padding_value": 0.0,
"return_attention_mask": true,
"sampling_rate": 16000
}

1
.ipynb_checkpoints/special_tokens_map-checkpoint.json Normal file
View File

@@ -0,0 +1 @@
{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "[UNK]", "pad_token": "[PAD]"}

1
.ipynb_checkpoints/tokenizer_config-checkpoint.json Normal file
View File

@@ -0,0 +1 @@
{"unk_token": "[UNK]", "bos_token": "<s>", "eos_token": "</s>", "pad_token": "[PAD]", "do_lower_case": false, "word_delimiter_token": "|"}

1
.ipynb_checkpoints/vocab-checkpoint.json Normal file
View File

@@ -0,0 +1 @@
{"خ": 0, "ة": 1, "د": 2, "ا": 4, "ض": 5, "م": 6, "و": 7, "ك": 8, "ث": 9, "ش": 10, "ع": 11, "ز": 12, "ء": 13, "ی": 14, "ن": 15, "ه": 16, "ق": 17, "ت": 18, "ب": 19, "ف": 20, "ظ": 21, "ح": 22, "ص": 23, "ئ": 24, "ذ": 25, "ى": 26, "غ": 27, "س": 28, "ر": 29, "ط": 30, "ي": 31, "ل": 32, "ؤ": 33, "ج": 34, "|": 3, "[UNK]": 35, "[PAD]": 36}

144
README.md Normal file
View File

@@ -0,0 +1,144 @@
---
language: ar
datasets:
- common_voice: Common Voice Corpus 4
metrics:
- wer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: Hasni XLSR Wav2Vec2 Large 53
results:
- task:
name: Speech Recognition
type: automatic-speech-recognition
dataset:
name: Common Voice ar
type: common_voice
args: ar
metrics:
- name: Test WER
type: wer
value: 52.18
---
# Wav2Vec2-Large-XLSR-53-Arabic
Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Arabic using the [Common Voice Corpus 4](https://commonvoice.mozilla.org/en/datasets) dataset.
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", "ar", split="test[:2%]")
processor = Wav2Vec2Processor.from_pretrained("anas/wav2vec2-large-xlsr-arabic")
model = Wav2Vec2ForCTC.from_pretrained("anas/wav2vec2-large-xlsr-arabic")
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 Arabic 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", "ar", split="test")
processor = Wav2Vec2Processor.from_pretrained("anas/wav2vec2-large-xlsr-arabic")
model = Wav2Vec2ForCTC.from_pretrained("anas/wav2vec2-large-xlsr-arabic/")
model.to("cuda")
chars_to_ignore_regex = '[\,\؟\.\!\-\;\\:\'\"\☭\«\»\؛\—\ـ\_\،\“\%\‘\”\<5C>]'
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()
batch["sentence"] = re.sub('[a-z]','',batch["sentence"])
batch["sentence"] = re.sub("[إأٱآا]", "ا", batch["sentence"])
noise = re.compile(""" ّ | # Tashdid
َ | # Fatha
ً | # Tanwin Fath
ُ | # Damma
ٌ | # Tanwin Damm
ِ | # Kasra
ٍ | # Tanwin Kasr
ْ | # Sukun
ـ # Tatwil/Kashida
""", re.VERBOSE)
batch["sentence"] = re.sub(noise, '', 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 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**: 52.18 %
## Training
The Common Voice Corpus 4 `train`, `validation`, datasets were used for training
The script used for training can be found [here](https://github.com/anashas/Fine-Tuning-of-XLSR-Wav2Vec2-on-Arabic)
Twitter: [here](https://twitter.com/hasnii_anas)
Email: anashasni146@gmail.com

76
config.json Normal file
View File

@@ -0,0 +1,76 @@
{
"_name_or_path": "facebook/wav2vec2-large-xlsr-53",
"activation_dropout": 0.0,
"apply_spec_augment": true,
"architectures": [
"Wav2Vec2ForCTC"
],
"attention_dropout": 0.1,
"bos_token_id": 1,
"conv_bias": true,
"conv_dim": [
512,
512,
512,
512,
512,
512,
512
],
"conv_kernel": [
10,
3,
3,
3,
3,
2,
2
],
"conv_stride": [
5,
2,
2,
2,
2,
2,
2
],
"ctc_loss_reduction": "mean",
"ctc_zero_infinity": false,
"do_stable_layer_norm": true,
"eos_token_id": 2,
"feat_extract_activation": "gelu",
"feat_extract_dropout": 0.0,
"feat_extract_norm": "layer",
"feat_proj_dropout": 0.0,
"final_dropout": 0.0,
"gradient_checkpointing": true,
"hidden_act": "gelu",
"hidden_dropout": 0.05,
"hidden_size": 1024,
"initializer_range": 0.02,
"intermediate_size": 4096,
"layer_norm_eps": 1e-05,
"layerdrop": 0.1,
"mask_channel_length": 10,
"mask_channel_min_space": 1,
"mask_channel_other": 0.0,
"mask_channel_prob": 0.0,
"mask_channel_selection": "static",
"mask_feature_length": 10,
"mask_feature_prob": 0.0,
"mask_time_length": 10,
"mask_time_min_space": 1,
"mask_time_other": 0.0,
"mask_time_prob": 0.05,
"mask_time_selection": "static",
"model_type": "wav2vec2",
"num_attention_heads": 16,
"num_conv_pos_embedding_groups": 16,
"num_conv_pos_embeddings": 128,
"num_feat_extract_layers": 7,
"num_hidden_layers": 24,
"pad_token_id": 36,
"transformers_version": "4.5.0.dev0",
"vocab_size": 37
}

3
flax_model.msgpack Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:28991f821fbe8a965410fed20f36cc750d7c9b874ca371781dea99a5130f6841
size 1261921972

8
preprocessor_config.json Normal file
View File

@@ -0,0 +1,8 @@
{
"do_normalize": true,
"feature_size": 1,
"padding_side": "right",
"padding_value": 0.0,
"return_attention_mask": true,
"sampling_rate": 16000
}

3
pytorch_model.bin Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:c9d92c7e4e59488cb3de5cb0336893c24517ea0da99be9f9cd6f77ada2ecbe0b
size 1262085527

1
special_tokens_map.json Normal file
View File

@@ -0,0 +1 @@
{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "[UNK]", "pad_token": "[PAD]"}

1
tokenizer_config.json Normal file
View File

@@ -0,0 +1 @@
{"unk_token": "[UNK]", "bos_token": "<s>", "eos_token": "</s>", "pad_token": "[PAD]", "do_lower_case": false, "word_delimiter_token": "|"}

1
vocab.json Normal file
View File

@@ -0,0 +1 @@
{"خ": 0, "ة": 1, "د": 2, "ا": 4, "ض": 5, "م": 6, "و": 7, "ك": 8, "ث": 9, "ش": 10, "ع": 11, "ز": 12, "ء": 13, "ی": 14, "ن": 15, "ه": 16, "ق": 17, "ت": 18, "ب": 19, "ف": 20, "ظ": 21, "ح": 22, "ص": 23, "ئ": 24, "ذ": 25, "ى": 26, "غ": 27, "س": 28, "ر": 29, "ط": 30, "ي": 31, "ل": 32, "ؤ": 33, "ج": 34, "|": 3, "[UNK]": 35, "[PAD]": 36}