初始化项目，由ModelHub XC社区提供模型

Model: arbml/wav2vec2-large-xlsr-53-arabic-egyptian
Source: Original Platform

.gitattributes

@@ -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

README.md

@@ -0,0 +1,125 @@
+---
+language: ???
+datasets:
+- common_voice
+tags:
+- audio
+- automatic-speech-recognition
+- speech
+- xlsr-fine-tuning-week
+license: apache-2.0
+model-index:
+- name: XLSR Wav2Vec2 Arabic Egyptian by Zaid
+  results:
+  - task: 
+      name: Speech Recognition
+      type: automatic-speech-recognition
+    dataset:
+      name: Common Voice ???
+      type: common_voice
+      args: ???
+    metrics:
+       - name: Test WER
+         type: wer
+         value: ???
+---
+
+# Wav2Vec2-Large-XLSR-53-Tamil
+
+Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) in Tamil using the [Common Voice](https://huggingface.co/datasets/common_voice)
+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", "???", split="test[:2%]").
+
+processor = Wav2Vec2Processor.from_pretrained("Zaid/wav2vec2-large-xlsr-53-arabic-egyptian")
+model = Wav2Vec2ForCTC.from_pretrained("Zaid/wav2vec2-large-xlsr-53-arabic-egyptian")
+
+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 {language} 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", "???", split="test")
+wer = load_metric("wer")
+
+processor = Wav2Vec2Processor.from_pretrained("Zaid/wav2vec2-large-xlsr-53-arabic-egyptian")
+model = Wav2Vec2ForCTC.from_pretrained("Zaid/wav2vec2-large-xlsr-53-arabic-egyptian")
+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**: ??? %
+
+
+## Training
+
+The Common Voice `train`, `validation` datasets were used for training.
+
+The script used for training can be found ???

config.json

@@ -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": 0.1,
+  "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": 44,
+  "transformers_version": "4.4.0",
+  "vocab_size": 45
+}

dialects_speech_corpus.py

@@ -0,0 +1,108 @@
+"""Arabic Speech Corpus"""
+
+from __future__ import absolute_import, division, print_function
+
+import os
+
+import datasets
+
+
+_CITATION = """
+"""
+
+_DESCRIPTION = """\
+
+
+```python
+import soundfile as sf
+
+def map_to_array(batch):
+    speech_array, _ = sf.read(batch["file"])
+    batch["speech"] = speech_array
+    return batch
+
+dataset = dataset.map(map_to_array, remove_columns=["file"])
+```
+"""
+
+_URL = "mgb3.zip"
+corrupt_files = ['familyKids_02_first_12min.wav','sports_04_first_12min.wav',
+'cooking_05_first_12min.wav', 'moviesDrama_07_first_12min.wav','science_06_first_12min.wav',
+'comedy_09_first_12min.wav','cultural_08_first_12min.wav','familyKids_11_first_12min.wav',
+'science_10_first_12min.wav']
+import soundfile as sf
+
+class EgyptianSpeechCorpusConfig(datasets.BuilderConfig):
+    """BuilderConfig for EgyptianSpeechCorpus."""
+
+    def __init__(self, **kwargs):
+        """
+        Args:
+          data_dir: `string`, the path to the folder containing the files in the
+            downloaded .tar
+          citation: `string`, citation for the data set
+          url: `string`, url for information about the data set
+          **kwargs: keyword arguments forwarded to super.
+        """
+        super(EgyptianSpeechCorpusConfig, self).__init__(version=datasets.Version("2.1.0", ""), **kwargs)
+
+
+def map_to_array(batch):
+    start, stop = batch['segment'].split('_')
+    speech_array, _ = sf.read(batch["file"], start = start, stop = stop)
+    batch["speech"] = speech_array
+    return batch
+
+class EgyptionSpeechCorpus(datasets.GeneratorBasedBuilder):
+    """EgyptianSpeechCorpus dataset."""
+
+    BUILDER_CONFIGS = [
+        EgyptianSpeechCorpusConfig(name="clean", description="'Clean' speech."),
+    ]
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features(
+                {
+                    "file": datasets.Value("string"),
+                    "text": datasets.Value("string"),
+                    "segment": datasets.Value("string")
+                }
+            ),
+            supervised_keys=("file", "text"),
+            homepage=_URL,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        self.archive_path = '/content/mgb3'
+        return [
+            datasets.SplitGenerator(name="train", gen_kwargs={"archive_path": os.path.join(self.archive_path, "adapt")}),
+            datasets.SplitGenerator(name="dev", gen_kwargs={"archive_path": os.path.join(self.archive_path, "dev")}),
+            datasets.SplitGenerator(name="test", gen_kwargs={"archive_path": os.path.join(self.archive_path, "test")}),
+        ]
+
+    def _generate_examples(self, archive_path):
+        """Generate examples from a Librispeech archive_path."""
+        text_dir = os.path.join(archive_path, "Alaa")
+        wav_dir = os.path.join(self.archive_path, "wav")
+        
+        segments_file = os.path.join(text_dir, "text_noverlap")
+
+        with open(segments_file, "r", encoding="utf-8") as f:
+            for _id, line in enumerate(f):
+                segment = line.split(' ')[0]
+                text = ' '.join(line.split(' ')[1:])
+                wav_file = '_'.join(segment.split('_')[:4]) +'.wav'
+                start, stop = segment.split('_')[4:6]
+                wav_path = os.path.join(wav_dir, wav_file) 
+                if (wav_file in corrupt_files) or (wav_file not in os.listdir(wav_dir)):
+                    continue
+                example = {
+                    "file": wav_path,
+                    "text": text,
+                    "segment":('_').join([start, stop])
+                }
+                yield str(_id), example
+

flax_model.msgpack

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95832f40a311c8ab065115bd0cd977d2537b69a66e798c400a0c1247e49232d9
+size 1261954772

optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35567b797d2ad024959817f25246b0d06ececc4dd794d69da7e9334950b455aa
+size 2490446215

preprocessor_config.json

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

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a20efe1c5480b033037f9ee3e21e24d9a51cf765fd5b96f0375b02012664b7b0
+size 1262118359

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0818666afb82b7b432824bd91f7b1cbfd826c0bbd672aea23cda693e38746b70
+size 623

special_tokens_map.json

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

tokenizer_config.json

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

trainer_state.json

@@ -0,0 +1,100 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 18.46153846153846,
+  "global_step": 1200,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.00011999999999999999,
+      "loss": 5.7855,
+      "step": 200
+    },
+    {
+      "epoch": 3.08,
+      "eval_loss": 3.094594717025757,
+      "eval_runtime": 32.9209,
+      "eval_samples_per_second": 7.017,
+      "eval_wer": 1.0,
+      "step": 200
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.00023999999999999998,
+      "loss": 3.0293,
+      "step": 400
+    },
+    {
+      "epoch": 6.15,
+      "eval_loss": 3.005324125289917,
+      "eval_runtime": 32.0726,
+      "eval_samples_per_second": 7.202,
+      "eval_wer": 1.0,
+      "step": 400
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.0002625,
+      "loss": 1.9629,
+      "step": 600
+    },
+    {
+      "epoch": 9.23,
+      "eval_loss": 1.0649160146713257,
+      "eval_runtime": 32.4082,
+      "eval_samples_per_second": 7.128,
+      "eval_wer": 0.7503426222019187,
+      "step": 600
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.00018749999999999998,
+      "loss": 0.9013,
+      "step": 800
+    },
+    {
+      "epoch": 12.31,
+      "eval_loss": 0.8863689303398132,
+      "eval_runtime": 32.197,
+      "eval_samples_per_second": 7.175,
+      "eval_wer": 0.6196893558702604,
+      "step": 800
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 0.0001125,
+      "loss": 0.6283,
+      "step": 1000
+    },
+    {
+      "epoch": 15.38,
+      "eval_loss": 0.8295900821685791,
+      "eval_runtime": 32.1739,
+      "eval_samples_per_second": 7.18,
+      "eval_wer": 0.5616719963453631,
+      "step": 1000
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 3.75e-05,
+      "loss": 0.4995,
+      "step": 1200
+    },
+    {
+      "epoch": 18.46,
+      "eval_loss": 0.8448638319969177,
+      "eval_runtime": 32.5186,
+      "eval_samples_per_second": 7.104,
+      "eval_wer": 0.5415714938328003,
+      "step": 1200
+    }
+  ],
+  "max_steps": 1300,
+  "num_train_epochs": 20,
+  "total_flos": 7.999595747693107e+18,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12f8d81898a4d31d043925ba819cf0ce4cfa60b145af57af0c27995159ac9a6b
+size 2287

vocab.json

@@ -0,0 +1 @@
+{"ذ": 0, "آ": 1, "ز": 2, "ى": 3, ")": 4, "خ": 5, "ط": 6, "ض": 7, "ج": 8, "م": 9, "ك": 10, "ء": 11, "ح": 12, "ؤ": 13, "ا": 14, "ع": 16, "ه": 17, "ق": 18, "ب": 19, "غ": 20, "ر": 21, "د": 22, "ن": 23, "#": 24, "(": 25, "ة": 26, "س": 27, "ظ": 28, "ي": 29, "ث": 30, "ئ": 31, "ش": 32, "ف": 33, "ت": 34, "ص": 35, "و": 36, "إ": 37, "،": 38, "ّ": 39, "أ": 40, "ل": 41, "؟": 42, "|": 15, "[UNK]": 43, "[PAD]": 44}