first commit

.gitattributes

@@ -31,5 +31,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.wasm filter=lfs diff=lfs merge=lfs -text
 *.xz filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
+*.arpa filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,13 +1,84 @@
----
-title: Tunisian Speech Recognition
-emoji: 🏢
-colorFrom: red
-colorTo: blue
-sdk: gradio
-sdk_version: 3.44.4
-app_file: app.py
-pinned: false
-license: cc-by-4.0
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# Overview
+
+This project aims to create an Automatic Speech Recognition (ASR) model dedicated for the Tunisian Arabic dialect. The goal is to improve speech recognition technology for underrepresented linguistic communities by transcribing Tunisian dialect speech into written text.
+
+## Dataset
+Part of the audio and text data (The ones we collected) used to train and test the model has been provided to encourage and support research within the community. Please find the dataset [here](https://zenodo.org/record/8370566). This Zenodo record contains labeled and unlabeled Tunisian Arabic audio data, along with textual data for language modelling. 
+The folder also contains a 4-gram language model trained with KenLM on data released within the Zenodo record. The .arpa file is called "outdomain.arpa". 
+
+## Performance
+
+The following table summarizes the performance of the model on various considered test sets :
+
+| Dataset | CER     | WER                |
+| :-------- | :------- | :------------------------- |
+| `TARIC` | `6.22%` | `10.55%` |
+| `IWSLT` | `21.18%` | `39.53%` |
+| `TunSwitch TO` | `9.67%` | `25.54%` |
+
+More details about the test sets, and the conditions leading to this performance in the paper.
+
+
+
+## Team
+
+Here are the team members who have contributed to  this project
+
+* [Salah Zaiem](https://fr.linkedin.com/in/salah-zaiem)
+* [Ahmed Amine Ben Aballah](https://www.linkedin.com/in/aabenz/)
+* [Ata Kaboudi](https://www.linkedin.com/in/ata-kaboudi-63365b1a8)
+* [Amir Kanoun](https://tn.linkedin.com/in/ahmed-amir-kanoun)
+
+## Paper
+ More in-depth details and insights are available in a released preprint. Please find the paper [here](https://arxiv.org/abs/2309.11327).
+If you use or refer to this model, please cite : 
+
+```
+@misc{abdallah2023leveraging,
+      title={Leveraging Data Collection and Unsupervised Learning for Code-switched Tunisian Arabic Automatic Speech Recognition}, 
+      author={Ahmed Amine Ben Abdallah and Ata Kabboudi and Amir Kanoun and Salah Zaiem},
+      year={2023},
+      eprint={2309.11327},
+      archivePrefix={arXiv},
+      primaryClass={eess.AS}
+}
+```
+
+
+## Datasets
+This ASR model was trained on :
+* TARIC : The corpus, named TARIC (Tunisian Arabic Railway Interaction Corpus) has a collection of audio recordings and transcriptions from dialogues in the Tunisian Railway Transport Network. - [Taric Corpus](https://aclanthology.org/L14-1385/) -
+* IWSLT : A Tunisian conversational speech - [IWSLT Corpus](https://iwslt.org/2022/dialect)-
+* TunSwitch : Our crowd-collected dataset described in the paper presented above. 
+
+## Demo
+Here is a working live demo :  [LINK](https://huggingface.co/spaces/SalahZa/Code-Switched-Tunisian-SpeechToText)
+
+
+## Inference 
+
+### 1. Create a CSV test file
+First, you have to create a csv file that follows SpeechBrain's format which contain 4 columns:
+* ID: contain ID to identify each audio sample in the dataset
+* wav: contain the path to the audio file
+* wrd: contain the text transcription of the spoken content in the audio file if you have it and use your set for evaluation. Put anything if you don't have transcriptions. An example is provided in this folder, the file is called : taric_test.csv
+* duration: the duration of the audio in seconds
+
+
+### 2. Adjust the hyperparams.yaml file
+
+Adjust the path of **test_csv** parameter to your csv file path
+
+
+To run this recipe, do the following:
+```
+> python train_with_wavlm.py semi_wavlm_large_tunisian_ctc/1234/hyperparams.yaml --test_csv = path_to_csv
+```
+
+If you want to infer on single files, the space demo offers proper easy-to-use inference code.
+
+
+## Contact : 
+If you have questions, you can send an email to : [email protected] 
+
+

app.py

@@ -0,0 +1,430 @@
+#!/usr/bin/env python3
+import sys
+import torch
+import logging
+import gradio as gr
+import speechbrain as sb
+from pathlib import Path
+import os
+import torchaudio
+from hyperpyyaml import load_hyperpyyaml
+from speechbrain.tokenizers.SentencePiece import SentencePiece
+from speechbrain.utils.data_utils import undo_padding
+from speechbrain.utils.distributed import run_on_main
+
+"""Recipe for training a sequence-to-sequence ASR system with CommonVoice.
+The system employs a wav2vec2 encoder and a CTC decoder.
+Decoding is performed with greedy decoding (will be extended to beam search).
+
+To run this recipe, do the following:
+> python train_with_wav2vec2.py hparams/train_with_wav2vec2.yaml
+
+With the default hyperparameters, the system employs a pretrained wav2vec2 encoder.
+The wav2vec2 model is pretrained following the model given in the hprams file.
+It may be dependent on the language.
+
+The neural network is trained with CTC on sub-word units estimated with
+Byte Pairwise Encoding (BPE).
+
+The experiment file is flexible enough to support a large variety of
+different systems. By properly changing the parameter files, you can try
+different encoders, decoders, tokens (e.g, characters instead of BPE),
+training languages (all CommonVoice languages), and many
+other possible variations.
+
+Authors
+ * Titouan Parcollet 2021
+"""
+
+logger = logging.getLogger(__name__)
+
+
+# Define training procedure
+class ASR(sb.core.Brain):
+    def compute_forward(self, batch, stage):
+        """Forward computations from the waveform batches to the output probabilities."""
+
+        batch = batch.to(self.device)
+        wavs, wav_lens = batch.sig
+        wavs, wav_lens = wavs.to(self.device), wav_lens.to(self.device)
+        if stage == sb.Stage.TRAIN:
+            if hasattr(self.hparams, "augmentation"):
+                wavs = self.hparams.augmentation(wavs, wav_lens)
+
+        # Forward pass
+        feats = self.modules.wav2vec2(wavs, wav_lens)
+        x = self.modules.enc(feats)
+        logits = self.modules.ctc_lin(x)
+        p_ctc = self.hparams.log_softmax(logits)
+
+        return p_ctc, wav_lens
+    
+    def treat_wav(self,sig):
+        feats = self.modules.wav2vec2(sig.to("cpu"), torch.tensor([1]).to("cpu"))
+        feats = self.modules.enc(feats)
+        logits = self.modules.ctc_lin(feats)
+        p_ctc = self.hparams.log_softmax(logits)
+        predicted_words =[]
+        for logs in p_ctc:
+            text = decoder.decode(logs.detach().cpu().numpy())
+            predicted_words.append(text.split(" "))
+        return " ".join(predicted_words[0])
+ 
+    def compute_objectives(self, predictions, batch, stage):
+        """Computes the loss (CTC) given predictions and targets."""
+
+        p_ctc, wav_lens = predictions
+
+        ids = batch.id
+        tokens, tokens_lens = batch.tokens
+
+        loss = self.hparams.ctc_cost(p_ctc, tokens, wav_lens, tokens_lens)
+
+        if stage != sb.Stage.TRAIN:
+            predicted_tokens = sb.decoders.ctc_greedy_decode(
+                p_ctc, wav_lens, blank_id=self.hparams.blank_index
+            )
+            # Decode token terms to words
+            if self.hparams.use_language_modelling:
+                predicted_words = []
+                for logs in p_ctc:
+                    text = decoder.decode(logs.detach().cpu().numpy())
+                    predicted_words.append(text.split(" "))
+            else:
+                predicted_words = [
+                    "".join(self.tokenizer.decode_ndim(utt_seq)).split(" ")
+                    for utt_seq in predicted_tokens
+                ]
+            # Convert indices to words
+            target_words = [wrd.split(" ") for wrd in batch.wrd]
+
+            self.wer_metric.append(ids, predicted_words, target_words)
+            self.cer_metric.append(ids, predicted_words, target_words)
+
+        return loss
+
+    def fit_batch(self, batch):
+        """Train the parameters given a single batch in input"""
+        should_step = self.step % self.grad_accumulation_factor == 0
+        # Managing automatic mixed precision
+        # TOFIX: CTC fine-tuning currently is unstable
+        # This is certainly due to CTC being done in fp16 instead of fp32
+        if self.auto_mix_prec:
+            with torch.cuda.amp.autocast():
+                with self.no_sync():
+                    outputs = self.compute_forward(batch, sb.Stage.TRAIN)
+                loss = self.compute_objectives(outputs, batch, sb.Stage.TRAIN)
+            with self.no_sync(not should_step):
+                self.scaler.scale(
+                    loss / self.grad_accumulation_factor
+                ).backward()
+            if should_step:
+
+                if not self.hparams.wav2vec2.freeze:
+                    self.scaler.unscale_(self.wav2vec_optimizer)
+                self.scaler.unscale_(self.model_optimizer)
+                if self.check_gradients(loss):
+                    if not self.hparams.wav2vec2.freeze:
+                        if self.optimizer_step >= self.hparams.warmup_steps:
+                            self.scaler.step(self.wav2vec_optimizer)
+                    self.scaler.step(self.model_optimizer)
+                self.scaler.update()
+                self.zero_grad()
+                self.optimizer_step += 1
+        else:
+            # This is mandatory because HF models have a weird behavior with DDP
+            # on the forward pass
+            with self.no_sync():
+                outputs = self.compute_forward(batch, sb.Stage.TRAIN)
+
+            loss = self.compute_objectives(outputs, batch, sb.Stage.TRAIN)
+
+            with self.no_sync(not should_step):
+                (loss / self.grad_accumulation_factor).backward()
+            if should_step:
+                if self.check_gradients(loss):
+                    if not self.hparams.wav2vec2.freeze:
+                        if self.optimizer_step >= self.hparams.warmup_steps:
+                            self.wav2vec_optimizer.step()
+                    self.model_optimizer.step()
+                self.zero_grad()
+                self.optimizer_step += 1
+
+        self.on_fit_batch_end(batch, outputs, loss, should_step)
+        return loss.detach().cpu()
+
+    def evaluate_batch(self, batch, stage):
+        """Computations needed for validation/test batches"""
+        predictions = self.compute_forward(batch, stage=stage)
+        with torch.no_grad():
+            loss = self.compute_objectives(predictions, batch, stage=stage)
+        return loss.detach()
+
+    def on_stage_start(self, stage, epoch):
+        """Gets called at the beginning of each epoch"""
+        if stage != sb.Stage.TRAIN:
+            self.cer_metric = self.hparams.cer_computer()
+            self.wer_metric = self.hparams.error_rate_computer()
+
+    def on_stage_end(self, stage, stage_loss, epoch):
+        """Gets called at the end of an epoch."""
+        # Compute/store important stats
+        stage_stats = {"loss": stage_loss}
+        if stage == sb.Stage.TRAIN:
+            self.train_stats = stage_stats
+        else:
+            stage_stats["CER"] = self.cer_metric.summarize("error_rate")
+            stage_stats["WER"] = self.wer_metric.summarize("error_rate")
+
+        # Perform end-of-iteration things, like annealing, logging, etc.
+        if stage == sb.Stage.VALID:
+            old_lr_model, new_lr_model = self.hparams.lr_annealing_model(
+                stage_stats["loss"]
+            )
+            old_lr_wav2vec, new_lr_wav2vec = self.hparams.lr_annealing_wav2vec(
+                stage_stats["loss"]
+            )
+            sb.nnet.schedulers.update_learning_rate(
+                self.model_optimizer, new_lr_model
+            )
+            if not self.hparams.wav2vec2.freeze:
+                sb.nnet.schedulers.update_learning_rate(
+                    self.wav2vec_optimizer, new_lr_wav2vec
+                )
+            self.hparams.train_logger.log_stats(
+                stats_meta={
+                    "epoch": epoch,
+                    "lr_model": old_lr_model,
+                    "lr_wav2vec": old_lr_wav2vec,
+                },
+                train_stats=self.train_stats,
+                valid_stats=stage_stats,
+            )
+            self.checkpointer.save_and_keep_only(
+                meta={"WER": stage_stats["WER"]}, min_keys=["WER"],
+            )
+        elif stage == sb.Stage.TEST:
+            self.hparams.train_logger.log_stats(
+                stats_meta={"Epoch loaded": self.hparams.epoch_counter.current},
+                test_stats=stage_stats,
+            )
+            with open(self.hparams.wer_file, "w") as w:
+                self.wer_metric.write_stats(w)
+
+    def init_optimizers(self):
+        "Initializes the wav2vec2 optimizer and model optimizer"
+
+        # If the wav2vec encoder is unfrozen, we create the optimizer
+        if not self.hparams.wav2vec2.freeze:
+            self.wav2vec_optimizer = self.hparams.wav2vec_opt_class(
+                self.modules.wav2vec2.parameters()
+            )
+            if self.checkpointer is not None:
+                self.checkpointer.add_recoverable(
+                    "wav2vec_opt", self.wav2vec_optimizer
+                )
+
+        self.model_optimizer = self.hparams.model_opt_class(
+            self.hparams.model.parameters()
+        )
+
+        if self.checkpointer is not None:
+            self.checkpointer.add_recoverable("modelopt", self.model_optimizer)
+
+    def zero_grad(self, set_to_none=False):
+        if not self.hparams.wav2vec2.freeze:
+            self.wav2vec_optimizer.zero_grad(set_to_none)
+        self.model_optimizer.zero_grad(set_to_none)
+
+
+# Define custom data procedure
+def dataio_prepare(hparams):
+    """This function prepares the datasets to be used in the brain class.
+    It also defines the data processing pipeline through user-defined functions."""
+
+    # 1. Define datasets
+    data_folder = hparams["data_folder"]
+
+    train_data = sb.dataio.dataset.DynamicItemDataset.from_csv(
+        csv_path=hparams["train_csv"], replacements={"data_root": data_folder},
+    )
+
+    if hparams["sorting"] == "ascending":
+        # we sort training data to speed up training and get better results.
+        train_data = train_data.filtered_sorted(
+            sort_key="duration",
+            key_max_value={"duration": hparams["avoid_if_longer_than"]},
+        )
+        # when sorting do not shuffle in dataloader ! otherwise is pointless
+        hparams["dataloader_options"]["shuffle"] = False
+
+    elif hparams["sorting"] == "descending":
+        train_data = train_data.filtered_sorted(
+            sort_key="duration",
+            reverse=True,
+            key_max_value={"duration": hparams["avoid_if_longer_than"]},
+        )
+        # when sorting do not shuffle in dataloader ! otherwise is pointless
+        hparams["dataloader_options"]["shuffle"] = False
+
+    elif hparams["sorting"] == "random":
+        pass
+
+    else:
+        raise NotImplementedError(
+            "sorting must be random, ascending or descending"
+        )
+
+    valid_data = sb.dataio.dataset.DynamicItemDataset.from_csv(
+        csv_path=hparams["valid_csv"], replacements={"data_root": data_folder},
+    )
+    # We also sort the validation data so it is faster to validate
+    valid_data = valid_data.filtered_sorted(sort_key="duration")
+    test_datasets = {}
+    for csv_file in hparams["test_csv"]:
+        name = Path(csv_file).stem
+        test_datasets[name] = sb.dataio.dataset.DynamicItemDataset.from_csv(
+            csv_path=csv_file, replacements={"data_root": data_folder}
+        )
+        test_datasets[name] = test_datasets[name].filtered_sorted(
+            sort_key="duration"
+        )
+
+    datasets = [train_data, valid_data] + [i for k, i in test_datasets.items()]
+
+
+    # 2. Define audio pipeline:
+    @sb.utils.data_pipeline.takes("wav")
+    @sb.utils.data_pipeline.provides("sig")
+    def audio_pipeline(wav):
+        info = torchaudio.info(wav)
+        sig = sb.dataio.dataio.read_audio(wav)
+        resampled = torchaudio.transforms.Resample(
+            info.sample_rate, hparams["sample_rate"],
+        )(sig)
+        return resampled
+
+    sb.dataio.dataset.add_dynamic_item(datasets, audio_pipeline)
+    label_encoder = sb.dataio.encoder.CTCTextEncoder()
+
+    # 3. Define text pipeline:
+    @sb.utils.data_pipeline.takes("wrd")
+    @sb.utils.data_pipeline.provides(
+        "wrd", "char_list", "tokens_list", "tokens"
+    )
+    def text_pipeline(wrd):
+        yield wrd
+        char_list = list(wrd)
+        yield char_list
+        tokens_list = label_encoder.encode_sequence(char_list)
+        yield tokens_list
+        tokens = torch.LongTensor(tokens_list)
+        yield tokens
+
+    sb.dataio.dataset.add_dynamic_item(datasets, text_pipeline)
+    lab_enc_file = os.path.join(hparams["save_folder"], "label_encoder.txt")
+    special_labels = {
+        "blank_label": hparams["blank_index"],
+        "unk_label": hparams["unk_index"]
+    }
+    label_encoder.load_or_create(
+        path=lab_enc_file,
+        from_didatasets=[train_data],
+        output_key="char_list",
+        special_labels=special_labels,
+        sequence_input=True,
+    )
+
+    # 4. Set output:
+    sb.dataio.dataset.set_output_keys(
+        datasets, ["id", "sig", "wrd", "char_list", "tokens"],
+    )
+    return train_data, valid_data,test_datasets, label_encoder
+
+
+
+# Load hyperparameters file with command-line overrides
+hparams_file, run_opts, overrides = sb.parse_arguments(["train_semi.yaml"])
+with open(hparams_file) as fin:
+    hparams = load_hyperpyyaml(fin, overrides)
+
+# If --distributed_launch then
+# create ddp_group with the right communication protocol
+sb.utils.distributed.ddp_init_group(run_opts)
+
+
+# Create experiment directory
+sb.create_experiment_directory(
+    experiment_directory=hparams["output_folder"],
+    hyperparams_to_save=hparams_file,
+    overrides=overrides,
+)
+
+# Due to DDP, we do the preparation ONLY on the main python process
+# Defining tokenizer and loading it
+# Create the datasets objects as well as tokenization and encoding :-D
+label_encoder = sb.dataio.encoder.CTCTextEncoder()
+
+lab_enc_file = os.path.join(hparams["save_folder"], "label_encoder.txt")
+special_labels = {
+    "blank_label": hparams["blank_index"],
+    "unk_label": hparams["unk_index"]
+}
+label_encoder.load_or_create(
+    path=lab_enc_file,
+    from_didatasets=[[]],
+    output_key="char_list",
+    special_labels=special_labels,
+    sequence_input=True,
+)
+
+from pyctcdecode import build_ctcdecoder
+ind2lab = label_encoder.ind2lab
+print(ind2lab)
+labels = [ind2lab[x] for x in range(len(ind2lab))]
+labels = [""] + labels[1:-1] + ["1"] 
+# Replace the <blank> token with a blank character, needed for PyCTCdecode
+print(labels)
+decoder = build_ctcdecoder(
+    labels,
+    kenlm_model_path=hparams["ngram_lm_path"],  # .arpa or .bin
+    alpha=0.5,  # Default by KenLM
+    beta=1.0,  # Default by KenLM
+)
+# Trainer initialization
+run_opts["device"] = "cpu"
+asr_brain = ASR(
+    modules=hparams["modules"],
+    hparams=hparams,
+    run_opts=run_opts,
+    checkpointer=hparams["checkpointer"],
+)
+
+# Adding objects to trainer.
+asr_brain.tokenizer = label_encoder
+asr_brain.checkpointer.recover_if_possible(device="cpu")
+asr_brain.modules.eval()
+
+def treat_wav_file(file_mic,file_upload ,asr=asr_brain, device="cpu") :
+    if (file_mic is not None) and (file_upload is not None):
+        warn_output = "WARNING: You've uploaded an audio file and used the microphone. The recorded file from the microphone will be used and the uploaded audio will be discarded.\n"
+        wav = file_mic
+    elif (file_mic is None) and (file_upload is None):
+        return "ERROR: You have to either use the microphone or upload an audio file"
+    elif file_mic is not None:
+        wav = file_mic
+    else:
+        wav = file_upload
+    sig, sr = torchaudio.load(wav)
+    tensor_wav = sig.to(device)
+    resampled = torchaudio.functional.resample( tensor_wav, sr, 16000)
+    sentence = asr.treat_wav(resampled)
+    return sentence
+
+gr.Interface(
+    fn=treat_wav_file,
+    inputs=[gr.Audio(source="microphone", type='filepath', label = "record", optional = True),
+            gr.Audio(source="upload", type='filepath', label="filein", optional=True)]
+    ,outputs="text").launch()
+
+

outdomain.arpa

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24654c1d236bb1bd367125131c847c4a734e69914eda71a6786964c20440d8fe
+size 324243244

requirements.txt

@@ -0,0 +1,5 @@
+https://github.com/kpu/kenlm/archive/master.zip
+transformers
+speechbrain
+pyctcdecode
+

semi_wavlm_large_tunisian_ctc/1234/app.py

@@ -0,0 +1,430 @@
+#!/usr/bin/env python3
+import sys
+import torch
+import logging
+import gradio as gr
+import speechbrain as sb
+from pathlib import Path
+import os
+import torchaudio
+from hyperpyyaml import load_hyperpyyaml
+from speechbrain.tokenizers.SentencePiece import SentencePiece
+from speechbrain.utils.data_utils import undo_padding
+from speechbrain.utils.distributed import run_on_main
+
+"""Recipe for training a sequence-to-sequence ASR system with CommonVoice.
+The system employs a wav2vec2 encoder and a CTC decoder.
+Decoding is performed with greedy decoding (will be extended to beam search).
+
+To run this recipe, do the following:
+> python train_with_wav2vec2.py hparams/train_with_wav2vec2.yaml
+
+With the default hyperparameters, the system employs a pretrained wav2vec2 encoder.
+The wav2vec2 model is pretrained following the model given in the hprams file.
+It may be dependent on the language.
+
+The neural network is trained with CTC on sub-word units estimated with
+Byte Pairwise Encoding (BPE).
+
+The experiment file is flexible enough to support a large variety of
+different systems. By properly changing the parameter files, you can try
+different encoders, decoders, tokens (e.g, characters instead of BPE),
+training languages (all CommonVoice languages), and many
+other possible variations.
+
+Authors
+ * Titouan Parcollet 2021
+"""
+
+logger = logging.getLogger(__name__)
+
+
+# Define training procedure
+class ASR(sb.core.Brain):
+    def compute_forward(self, batch, stage):
+        """Forward computations from the waveform batches to the output probabilities."""
+
+        batch = batch.to(self.device)
+        wavs, wav_lens = batch.sig
+        wavs, wav_lens = wavs.to(self.device), wav_lens.to(self.device)
+        if stage == sb.Stage.TRAIN:
+            if hasattr(self.hparams, "augmentation"):
+                wavs = self.hparams.augmentation(wavs, wav_lens)
+
+        # Forward pass
+        feats = self.modules.wav2vec2(wavs, wav_lens)
+        x = self.modules.enc(feats)
+        logits = self.modules.ctc_lin(x)
+        p_ctc = self.hparams.log_softmax(logits)
+
+        return p_ctc, wav_lens
+    
+    def treat_wav(self,sig):
+        feats = self.modules.wav2vec2(sig.to("cpu"), torch.tensor([1]).to("cpu"))
+        feats = self.modules.enc(feats)
+        logits = self.modules.ctc_lin(feats)
+        p_ctc = self.hparams.log_softmax(logits)
+        predicted_words =[]
+        for logs in p_ctc:
+            text = decoder.decode(logs.detach().cpu().numpy())
+            predicted_words.append(text.split(" "))
+        return " ".join(predicted_words[0])
+ 
+    def compute_objectives(self, predictions, batch, stage):
+        """Computes the loss (CTC) given predictions and targets."""
+
+        p_ctc, wav_lens = predictions
+
+        ids = batch.id
+        tokens, tokens_lens = batch.tokens
+
+        loss = self.hparams.ctc_cost(p_ctc, tokens, wav_lens, tokens_lens)
+
+        if stage != sb.Stage.TRAIN:
+            predicted_tokens = sb.decoders.ctc_greedy_decode(
+                p_ctc, wav_lens, blank_id=self.hparams.blank_index
+            )
+            # Decode token terms to words
+            if self.hparams.use_language_modelling:
+                predicted_words = []
+                for logs in p_ctc:
+                    text = decoder.decode(logs.detach().cpu().numpy())
+                    predicted_words.append(text.split(" "))
+            else:
+                predicted_words = [
+                    "".join(self.tokenizer.decode_ndim(utt_seq)).split(" ")
+                    for utt_seq in predicted_tokens
+                ]
+            # Convert indices to words
+            target_words = [wrd.split(" ") for wrd in batch.wrd]
+
+            self.wer_metric.append(ids, predicted_words, target_words)
+            self.cer_metric.append(ids, predicted_words, target_words)
+
+        return loss
+
+    def fit_batch(self, batch):
+        """Train the parameters given a single batch in input"""
+        should_step = self.step % self.grad_accumulation_factor == 0
+        # Managing automatic mixed precision
+        # TOFIX: CTC fine-tuning currently is unstable
+        # This is certainly due to CTC being done in fp16 instead of fp32
+        if self.auto_mix_prec:
+            with torch.cuda.amp.autocast():
+                with self.no_sync():
+                    outputs = self.compute_forward(batch, sb.Stage.TRAIN)
+                loss = self.compute_objectives(outputs, batch, sb.Stage.TRAIN)
+            with self.no_sync(not should_step):
+                self.scaler.scale(
+                    loss / self.grad_accumulation_factor
+                ).backward()
+            if should_step:
+
+                if not self.hparams.wav2vec2.freeze:
+                    self.scaler.unscale_(self.wav2vec_optimizer)
+                self.scaler.unscale_(self.model_optimizer)
+                if self.check_gradients(loss):
+                    if not self.hparams.wav2vec2.freeze:
+                        if self.optimizer_step >= self.hparams.warmup_steps:
+                            self.scaler.step(self.wav2vec_optimizer)
+                    self.scaler.step(self.model_optimizer)
+                self.scaler.update()
+                self.zero_grad()
+                self.optimizer_step += 1
+        else:
+            # This is mandatory because HF models have a weird behavior with DDP
+            # on the forward pass
+            with self.no_sync():
+                outputs = self.compute_forward(batch, sb.Stage.TRAIN)
+
+            loss = self.compute_objectives(outputs, batch, sb.Stage.TRAIN)
+
+            with self.no_sync(not should_step):
+                (loss / self.grad_accumulation_factor).backward()
+            if should_step:
+                if self.check_gradients(loss):
+                    if not self.hparams.wav2vec2.freeze:
+                        if self.optimizer_step >= self.hparams.warmup_steps:
+                            self.wav2vec_optimizer.step()
+                    self.model_optimizer.step()
+                self.zero_grad()
+                self.optimizer_step += 1
+
+        self.on_fit_batch_end(batch, outputs, loss, should_step)
+        return loss.detach().cpu()
+
+    def evaluate_batch(self, batch, stage):
+        """Computations needed for validation/test batches"""
+        predictions = self.compute_forward(batch, stage=stage)
+        with torch.no_grad():
+            loss = self.compute_objectives(predictions, batch, stage=stage)
+        return loss.detach()
+
+    def on_stage_start(self, stage, epoch):
+        """Gets called at the beginning of each epoch"""
+        if stage != sb.Stage.TRAIN:
+            self.cer_metric = self.hparams.cer_computer()
+            self.wer_metric = self.hparams.error_rate_computer()
+
+    def on_stage_end(self, stage, stage_loss, epoch):
+        """Gets called at the end of an epoch."""
+        # Compute/store important stats
+        stage_stats = {"loss": stage_loss}
+        if stage == sb.Stage.TRAIN:
+            self.train_stats = stage_stats
+        else:
+            stage_stats["CER"] = self.cer_metric.summarize("error_rate")
+            stage_stats["WER"] = self.wer_metric.summarize("error_rate")
+
+        # Perform end-of-iteration things, like annealing, logging, etc.
+        if stage == sb.Stage.VALID:
+            old_lr_model, new_lr_model = self.hparams.lr_annealing_model(
+                stage_stats["loss"]
+            )
+            old_lr_wav2vec, new_lr_wav2vec = self.hparams.lr_annealing_wav2vec(
+                stage_stats["loss"]
+            )
+            sb.nnet.schedulers.update_learning_rate(
+                self.model_optimizer, new_lr_model
+            )
+            if not self.hparams.wav2vec2.freeze:
+                sb.nnet.schedulers.update_learning_rate(
+                    self.wav2vec_optimizer, new_lr_wav2vec
+                )
+            self.hparams.train_logger.log_stats(
+                stats_meta={
+                    "epoch": epoch,
+                    "lr_model": old_lr_model,
+                    "lr_wav2vec": old_lr_wav2vec,
+                },
+                train_stats=self.train_stats,
+                valid_stats=stage_stats,
+            )
+            self.checkpointer.save_and_keep_only(
+                meta={"WER": stage_stats["WER"]}, min_keys=["WER"],
+            )
+        elif stage == sb.Stage.TEST:
+            self.hparams.train_logger.log_stats(
+                stats_meta={"Epoch loaded": self.hparams.epoch_counter.current},
+                test_stats=stage_stats,
+            )
+            with open(self.hparams.wer_file, "w") as w:
+                self.wer_metric.write_stats(w)
+
+    def init_optimizers(self):
+        "Initializes the wav2vec2 optimizer and model optimizer"
+
+        # If the wav2vec encoder is unfrozen, we create the optimizer
+        if not self.hparams.wav2vec2.freeze:
+            self.wav2vec_optimizer = self.hparams.wav2vec_opt_class(
+                self.modules.wav2vec2.parameters()
+            )
+            if self.checkpointer is not None:
+                self.checkpointer.add_recoverable(
+                    "wav2vec_opt", self.wav2vec_optimizer
+                )
+
+        self.model_optimizer = self.hparams.model_opt_class(
+            self.hparams.model.parameters()
+        )
+
+        if self.checkpointer is not None:
+            self.checkpointer.add_recoverable("modelopt", self.model_optimizer)
+
+    def zero_grad(self, set_to_none=False):
+        if not self.hparams.wav2vec2.freeze:
+            self.wav2vec_optimizer.zero_grad(set_to_none)
+        self.model_optimizer.zero_grad(set_to_none)
+
+
+# Define custom data procedure
+def dataio_prepare(hparams):
+    """This function prepares the datasets to be used in the brain class.
+    It also defines the data processing pipeline through user-defined functions."""
+
+    # 1. Define datasets
+    data_folder = hparams["data_folder"]
+
+    train_data = sb.dataio.dataset.DynamicItemDataset.from_csv(
+        csv_path=hparams["train_csv"], replacements={"data_root": data_folder},
+    )
+
+    if hparams["sorting"] == "ascending":
+        # we sort training data to speed up training and get better results.
+        train_data = train_data.filtered_sorted(
+            sort_key="duration",
+            key_max_value={"duration": hparams["avoid_if_longer_than"]},
+        )
+        # when sorting do not shuffle in dataloader ! otherwise is pointless
+        hparams["dataloader_options"]["shuffle"] = False
+
+    elif hparams["sorting"] == "descending":
+        train_data = train_data.filtered_sorted(
+            sort_key="duration",
+            reverse=True,
+            key_max_value={"duration": hparams["avoid_if_longer_than"]},
+        )
+        # when sorting do not shuffle in dataloader ! otherwise is pointless
+        hparams["dataloader_options"]["shuffle"] = False
+
+    elif hparams["sorting"] == "random":
+        pass
+
+    else:
+        raise NotImplementedError(
+            "sorting must be random, ascending or descending"
+        )
+
+    valid_data = sb.dataio.dataset.DynamicItemDataset.from_csv(
+        csv_path=hparams["valid_csv"], replacements={"data_root": data_folder},
+    )
+    # We also sort the validation data so it is faster to validate
+    valid_data = valid_data.filtered_sorted(sort_key="duration")
+    test_datasets = {}
+    for csv_file in hparams["test_csv"]:
+        name = Path(csv_file).stem
+        test_datasets[name] = sb.dataio.dataset.DynamicItemDataset.from_csv(
+            csv_path=csv_file, replacements={"data_root": data_folder}
+        )
+        test_datasets[name] = test_datasets[name].filtered_sorted(
+            sort_key="duration"
+        )
+
+    datasets = [train_data, valid_data] + [i for k, i in test_datasets.items()]
+
+
+    # 2. Define audio pipeline:
+    @sb.utils.data_pipeline.takes("wav")
+    @sb.utils.data_pipeline.provides("sig")
+    def audio_pipeline(wav):
+        info = torchaudio.info(wav)
+        sig = sb.dataio.dataio.read_audio(wav)
+        resampled = torchaudio.transforms.Resample(
+            info.sample_rate, hparams["sample_rate"],
+        )(sig)
+        return resampled
+
+    sb.dataio.dataset.add_dynamic_item(datasets, audio_pipeline)
+    label_encoder = sb.dataio.encoder.CTCTextEncoder()
+
+    # 3. Define text pipeline:
+    @sb.utils.data_pipeline.takes("wrd")
+    @sb.utils.data_pipeline.provides(
+        "wrd", "char_list", "tokens_list", "tokens"
+    )
+    def text_pipeline(wrd):
+        yield wrd
+        char_list = list(wrd)
+        yield char_list
+        tokens_list = label_encoder.encode_sequence(char_list)
+        yield tokens_list
+        tokens = torch.LongTensor(tokens_list)
+        yield tokens
+
+    sb.dataio.dataset.add_dynamic_item(datasets, text_pipeline)
+    lab_enc_file = os.path.join(hparams["save_folder"], "label_encoder.txt")
+    special_labels = {
+        "blank_label": hparams["blank_index"],
+        "unk_label": hparams["unk_index"]
+    }
+    label_encoder.load_or_create(
+        path=lab_enc_file,
+        from_didatasets=[train_data],
+        output_key="char_list",
+        special_labels=special_labels,
+        sequence_input=True,
+    )
+
+    # 4. Set output:
+    sb.dataio.dataset.set_output_keys(
+        datasets, ["id", "sig", "wrd", "char_list", "tokens"],
+    )
+    return train_data, valid_data,test_datasets, label_encoder
+
+
+
+# Load hyperparameters file with command-line overrides
+hparams_file, run_opts, overrides = sb.parse_arguments(["train_semi.yaml"])
+with open(hparams_file) as fin:
+    hparams = load_hyperpyyaml(fin, overrides)
+
+# If --distributed_launch then
+# create ddp_group with the right communication protocol
+sb.utils.distributed.ddp_init_group(run_opts)
+
+
+# Create experiment directory
+sb.create_experiment_directory(
+    experiment_directory=hparams["output_folder"],
+    hyperparams_to_save=hparams_file,
+    overrides=overrides,
+)
+
+# Due to DDP, we do the preparation ONLY on the main python process
+# Defining tokenizer and loading it
+# Create the datasets objects as well as tokenization and encoding :-D
+label_encoder = sb.dataio.encoder.CTCTextEncoder()
+
+lab_enc_file = os.path.join(hparams["save_folder"], "label_encoder.txt")
+special_labels = {
+    "blank_label": hparams["blank_index"],
+    "unk_label": hparams["unk_index"]
+}
+label_encoder.load_or_create(
+    path=lab_enc_file,
+    from_didatasets=[[]],
+    output_key="char_list",
+    special_labels=special_labels,
+    sequence_input=True,
+)
+
+from pyctcdecode import build_ctcdecoder
+ind2lab = label_encoder.ind2lab
+print(ind2lab)
+labels = [ind2lab[x] for x in range(len(ind2lab))]
+labels = [""] + labels[1:-1] + ["1"] 
+# Replace the <blank> token with a blank character, needed for PyCTCdecode
+print(labels)
+decoder = build_ctcdecoder(
+    labels,
+    kenlm_model_path=hparams["ngram_lm_path"],  # .arpa or .bin
+    alpha=0.5,  # Default by KenLM
+    beta=1.0,  # Default by KenLM
+)
+# Trainer initialization
+run_opts["device"] = "cpu"
+asr_brain = ASR(
+    modules=hparams["modules"],
+    hparams=hparams,
+    run_opts=run_opts,
+    checkpointer=hparams["checkpointer"],
+)
+
+# Adding objects to trainer.
+asr_brain.tokenizer = label_encoder
+asr_brain.checkpointer.recover_if_possible(device="cpu")
+asr_brain.modules.eval()
+
+def treat_wav_file(file_mic,file_upload ,asr=asr_brain, device="cpu") :
+    if (file_mic is not None) and (file_upload is not None):
+        warn_output = "WARNING: You've uploaded an audio file and used the microphone. The recorded file from the microphone will be used and the uploaded audio will be discarded.\n"
+        wav = file_mic
+    elif (file_mic is None) and (file_upload is None):
+        return "ERROR: You have to either use the microphone or upload an audio file"
+    elif file_mic is not None:
+        wav = file_mic
+    else:
+        wav = file_upload
+    sig, sr = torchaudio.load(wav)
+    tensor_wav = sig.to(device)
+    resampled = torchaudio.functional.resample( tensor_wav, sr, 16000)
+    sentence = asr.treat_wav(resampled)
+    return sentence
+
+gr.Interface(
+    fn=treat_wav_file,
+    inputs=[gr.Audio(source="microphone", type='filepath', label = "record", optional = True),
+            gr.Audio(source="upload", type='filepath', label="filein", optional=True)]
+    ,outputs="text").launch()
+
+

semi_wavlm_large_tunisian_ctc/1234/env.log

@@ -0,0 +1,479 @@
+SpeechBrain system description
+==============================
+Python version:
+3.8.5 (default, Sep  4 2020, 07:30:14) 
+[GCC 7.3.0]
+==============================
+Installed Python packages:
+abkhazia==1.0
+absl-py==0.11.0
+aiofiles==23.2.1
+aiohttp==3.8.0
+aiosignal==1.2.0
+alabaster==0.7.12
+alembic==1.7.4
+altair==4.2.0
+altgraph==0.17
+antlr4-python3-runtime==4.9.3
+anyio==3.6.2
+appdirs==1.4.4
+argcomplete==1.12.2
+argon2-cffi==20.1.0
+arrow==1.2.3
+asgiref==3.6.0
+asteroid-filterbanks==0.4.0
+astunparse==1.6.3
+async-generator==1.10
+async-timeout==4.0.0
+attrdict==2.0.1
+attrs==20.3.0
+audeer==1.16.0
+audformat==0.11.5
+audinterface==0.7.0
+audiofile==1.0.0
+audiomentations==0.25.0
+audioread==2.1.9
+audobject==0.4.14
+audresample==0.1.6
+-e git+https://github.com/facebookresearch/WavAugment.git@54afcdb00ccc852c2f030f239f8532c9562b550e#egg=augment
+autopage==0.4.0
+Babel==2.9.0
+backcall==0.2.0
+backports.cached-property==1.0.2
+beautifulsoup4==4.10.0
+black==19.10b0
+bleach==3.3.0
+blessed==1.20.0
+boto3==1.20.2
+botocore==1.23.2
+bpemb==0.3.4
+braceexpand==0.1.7
+cachetools==4.2.0
+certifi @ file:///croot/certifi_1671487769961/work/certifi
+cffi==1.14.3
+cfgv==3.2.0
+chardet==3.0.4
+charset-normalizer==2.0.7
+click==7.1.2
+cliff==3.9.0
+clldutils==3.5.4
+cloudpickle==2.2.1
+cmaes==0.8.2
+cmake==3.18.4.post1
+cmd2==2.2.0
+colorama==0.4.4
+colorlog==4.6.2
+configparser==5.1.0
+conllu==4.5.3
+croniter==1.3.15
+cryptography==38.0.4
+csrgraph==0.1.28
+csvw==1.8.1
+cycler==0.10.0
+Cython==0.29.21
+dataclasses==0.6
+dateutils==0.6.12
+decorator==4.4.2
+deepdiff==6.3.0
+deepspeech==0.9.1
+defusedxml==0.7.1
+Deprecated==1.2.14
+dill==0.3.3
+Distance==0.1.3
+distlib==0.3.1
+Django==3.2.16
+django-auditlog==2.2.1
+django-filter==22.1
+django-js-asset==1.2.2
+django-mptt==0.14.0
+djangorestframework==3.14.0
+docker-pycreds==0.4.0
+docopt==0.6.2
+docutils==0.16
+drf-excel==2.2.0
+drf-flex-fields==1.0.0
+drf-renderer-xlsx==0.4.1
+easyocr==1.2.1
+editdistance==0.6.0
+einops==0.3.2
+emoji==2.2.0
+entrypoints==0.3
+et-xmlfile==1.1.0
+exceptiongroup==1.1.0
+farasapy==0.0.14
+fastapi==0.98.0
+fastjsonschema==2.17.1
+fasttext==0.9.2
+ffmpeg-python==0.2.0
+ffmpy==0.3.0
+filelock==3.0.12
+flair==0.12.2
+flake8==3.7.9
+flatbuffers==1.12
+frozendict==2.0.7
+frozenlist==1.2.0
+fsspec==2021.11.0
+ftfy==6.1.1
+future==0.18.2
+g2p-en==2.1.0
+gast==0.3.3
+gdown==4.4.0
+gdrive==0.1.5
+gensim==4.0.1
+gitdb==4.0.9
+GitPython==3.1.24
+google-api-core==2.11.1
+google-api-python-client==2.43.0
+google-auth==1.24.0
+google-auth-httplib2==0.1.0
+google-auth-oauthlib==0.5.3
+google-pasta==0.2.0
+googleapis-common-protos==1.59.1
+gradio==3.44.4
+gradio-client==0.5.1
+greenlet==1.1.2
+grpcio==1.32.0
+h11==0.14.0
+h5features==1.3.2
+h5py==2.10.0
+hierarchy==0.4.0
+hmmlearn==0.2.8
+htk-io==0.5
+httpcore==0.16.3
+httplib2==0.22.0
+httpx==0.23.3
+huggingface-hub==0.15.1
+hydra-colorlog==0.1.4
+hydra-core==1.3.2
+hyperopt==0.2.7
+HyperPyYAML==1.1.0
+hypothesis==6.61.2
+identify==1.5.10
+idna==2.10
+imageio==2.9.0
+imagesize==1.2.0
+importlib-metadata==4.8.1
+importlib-resources==5.2.2
+inflect==5.3.0
+inquirer==3.1.3
+ipadic==1.0.0
+ipyevents==2.0.1
+ipykernel==5.3.4
+ipython==7.19.0
+ipython-genutils==0.2.0
+ipywebrtc==0.6.0
+ipywidgets==7.6.3
+iso-639==0.4.5
+isodate==0.6.0
+isort==4.3.21
+itsdangerous==2.1.2
+Janome==0.5.0
+jedi==0.17.2
+jeepney==0.8.0
+jieba==0.42.1
+Jinja2==3.0.3
+jiwer==2.2.0
+jmespath==0.10.0
+joblib==0.17.0
+jsonschema==3.2.0
+julius==0.2.7
+jupyter-client==6.1.7
+jupyter-core==4.7.0
+jupyterlab-pygments==0.1.2
+jupyterlab-widgets==1.0.0
+kaitaistruct==0.9
+kaldi-io==0.9.4
+kaldi-python-io==1.2.2
+kaldiio==2.17.2
+kenlm @ https://github.com/kpu/kenlm/archive/master.zip
+Keras-Preprocessing==1.1.2
+kiwisolver==1.3.1
+lang-trans==0.6.0
+langdetect==1.0.9
+latexcodec==2.0.1
+ldap3==2.9.1
+librosa==0.9.0
+lightning-cloud==0.5.37
+lightning-utilities==0.8.0
+linkify-it-py==1.0.3
+lit==16.0.6
+llvmlite==0.35.0
+lxml==4.9.0
+Mako==1.1.5
+Markdown==3.3.3
+markdown-it-py==3.0.0
+MarkupSafe==2.1.3
+marshmallow==3.14.0
+matplotlib==3.3.3
+mccabe==0.6.1
+mcd==0.4
+mdit-py-plugins==0.3.3
+mdurl==0.1.2
+mecab-python3==1.0.3
+megatron-lm==2.2.0
+metrics==0.3.3
+mido==1.2.10
+mistune==0.8.4
+more-itertools==8.6.0
+mpld3==0.3
+mpmath==1.2.1
+multidict==5.2.0
+multiprocess==0.70.11.1
+nbclient==0.5.3
+nbconvert==5.6.1
+nbformat==5.9.0
+NEMO==4.3.2
+nemo-toolkit==1.4.0
+nest-asyncio==1.5.1
+networkx==2.8.8
+nltk==3.2.4
+nodeenv==1.5.0
+normalize==2.0.2
+notebook==6.3.0
+numba==0.52.0
+numpy==1.19.4
+nvidia-cublas-cu11==11.10.3.66
+nvidia-cuda-cupti-cu11==11.7.101
+nvidia-cuda-nvrtc-cu11==11.7.99
+nvidia-cuda-runtime-cu11==11.7.99
+nvidia-cudnn-cu11==8.5.0.96
+nvidia-cufft-cu11==10.9.0.58
+nvidia-curand-cu11==10.2.10.91
+nvidia-cusolver-cu11==11.4.0.1
+nvidia-cusparse-cu11==11.7.4.91
+nvidia-nccl-cu11==2.14.3
+nvidia-nvtx-cu11==11.7.91
+oauthlib==3.1.0
+omegaconf==2.3.0
+onnx==1.10.2
+OpenCC==1.1.2
+opencv-python==4.4.0.46
+openpyxl==3.0.9
+opensmile==2.2.0
+opt-einsum==3.3.0
+optuna==2.10.0
+ordered-set==4.1.0
+orjson==3.8.4
+oyaml==1.0
+packaging==22.0
+pandas==1.2.5
+pandocfilters==1.4.3
+pangu==4.0.6.1
+parameterized==0.8.1
+parso==0.7.1
+pathlib2==2.3.7.post1
+pathspec==0.5.5
+pathtools==0.1.2
+pbr==5.6.0
+pefile==2019.4.18
+pescador==2.1.0
+pesq==0.0.3
+pexpect==4.8.0
+phonemizer==2.2.1
+pickleshare==0.7.5
+Pillow==9.3.0
+pip-api==0.0.23
+pipreqs==0.4.11
+pluggy==0.13.1
+pooch==1.3.0
+portalocker==2.3.2
+pptree==3.1
+pre-commit==2.9.0
+preprocessing==0.1.13
+pretty-midi==0.2.9
+prettytable==2.2.1
+primePy==1.3
+progressbar2==3.53.1
+prometheus-client==0.10.1
+promise==2.3
+prompt-toolkit==3.0.8
+protobuf==3.20.3
+psutil==5.6.6
+ptyprocess==0.6.0
+py==1.9.0
+py-espeak-ng==0.1.8
+py4j==0.10.9.7
+pyannote.audio==2.1.1
+pyannote.core==4.5
+pyannote.database==4.1.3
+pyannote.metrics==3.2.1
+pyannote.pipeline==2.3
+pyannotebook==0.1.0.dev0
+PyArabic==0.6.15
+pyarrow==3.0.0
+pyasn1==0.4.8
+pyasn1-modules==0.2.8
+pybind11==2.8.1
+pybtex==0.24.0
+pybtex-docutils==1.0.1
+pycodestyle==2.5.0
+pycparser==2.20
+pycryptodome==3.16.0
+pyctcdecode==0.4.0
+pydantic==1.10.4
+pyDeprecate==0.3.1
+pydub==0.25.1
+pyflakes==2.1.1
+Pygments==2.15.1
+pygtrie==2.5.0
+PyJWT==2.7.0
+pymodbus==2.5.3
+pyparsing==2.4.7
+pyperclip==1.8.2
+pypinyin==0.43.0
+pyrsistent==0.17.3
+pyserial==3.5
+PySocks==1.7.1
+pystoi==0.3.3
+pytest==5.4.1
+pytest-runner==5.3.1
+python-bidi==0.4.2
+python-crfsuite==0.9.7
+python-dateutil==2.8.2
+python-editor==1.0.4
+python-Levenshtein==0.12.2
+python-multipart==0.0.5
+python-utils==2.4.0
+pytorch-lightning==1.6.5
+pytorch-metric-learning==1.7.3
+pytorch-revgrad==0.2.0
+pytube==11.0.1
+pytz==2022.6
+PyWavelets==1.1.1
+PyYAML==6.0
+pyzmq==20.0.0
+rapidfuzz==1.8.2
+readchar==4.0.5
+regex==2020.11.13
+requests==2.28.1
+requests-oauthlib==1.3.0
+resampy==0.2.2
+rfc3986==1.4.0
+rich==13.4.2
+richenum==1.3.1
+rsa==4.7
+ruamel.yaml==0.17.21
+ruamel.yaml.clib==0.2.7
+s3m==1.1.0
+s3transfer==0.5.0
+sacrebleu==2.0.0
+sacremoses==0.0.44
+safetensors==0.3.1
+scikit-image==0.18.1
+scikit-learn==0.23.2
+scipy==1.5.4
+-e git+https://github.com/sanghack81/SDCIT@00d060dde733fde9345154a494f81e97fb395ca7#egg=SDCIT
+seaborn==0.11.1
+SecretStorage==3.3.3
+segments==2.1.3
+segtok==1.5.11
+semantic-version==2.10.0
+semver==2.13.0
+Send2Trash==1.5.0
+sentencepiece==0.1.99
+sentry-sdk==1.4.3
+shellingham==1.4.0
+shortuuid==1.0.7
+SIDEKIT==1.3.8.5.2
+simplejson==3.17.5
+singledispatchmethod==1.0
+six==1.15.0
+smart-open==5.0.0
+smmap==5.0.0
+sniffio==1.3.0
+snowballstemmer==2.0.0
+sortedcollections==2.1.0
+sortedcontainers==2.4.0
+sounddevice==0.4.5
+SoundFile==0.10.3.post1
+soupsieve==2.3
+sox==1.4.1
+sparsemax==0.1.9
+speechbrain==0.5.14
+sphfile==1.0.3
+Sphinx==3.3.1
+sphinx-rtd-theme==0.2.4
+sphinxcontrib-applehelp==1.0.2
+sphinxcontrib-bibtex==2.4.1
+sphinxcontrib-devhelp==1.0.2
+sphinxcontrib-htmlhelp==1.0.3
+sphinxcontrib-jsmath==1.0.1
+sphinxcontrib-qthelp==1.0.3
+sphinxcontrib-serializinghtml==1.1.4
+SQLAlchemy==1.4.25
+sqlitedict==2.1.0
+sqlparse==0.4.2
+stanza==1.4.2
+starlette==0.27.0
+starsessions==1.3.0
+stevedore==3.4.0
+subprocess32==3.5.4
+sympy==1.9
+tabulate==0.8.9
+tensorboard==2.4.0
+tensorboard-plugin-wit==1.7.0
+tensorboardX==2.6.1
+tensorflow==2.4.0
+tensorflow-estimator==2.4.0
+termcolor==1.1.0
+terminado==0.9.4
+testpath==0.4.4
+threadpoolctl==2.1.0
+tifffile==2020.12.8
+tikzplotlib==0.9.8
+tinycss2==1.2.1
+tkseem==0.0.3
+tokenizers==0.13.3
+toml==0.10.2
+toolz==0.12.0
+torch==1.13.1
+torch-audiomentations==0.11.0
+torch-pitch-shift==1.2.4
+torch-stft==0.1.4
+torchaudio==0.13.1
+torchmetrics==0.11.4
+torchvision==0.14.1
+tornado==6.1
+tqdm==4.61.1
+trackrip==1.2.1
+traitlets==5.9.0
+transformer-smaller-training-vocab==0.3.1
+transformers==4.30.2
+triton==2.0.0
+typed-ast==1.4.1
+typer==0.4.0
+typing-extensions==4.4.0
+uc-micro-py==1.0.1
+Unidecode==1.3.2
+uritemplate==3.0.1
+urllib3==1.26.2
+uvicorn==0.20.0
+versioneer==0.28
+virtualenv==20.2.1
+wandb==0.12.6
+wcwidth==0.2.5
+webdataset==0.1.62
+webencodings==0.5.1
+websocket-client==1.6.1
+websockets==10.4
+Werkzeug==1.0.1
+wget==3.2
+widgetsnbextension==3.5.1
+Wikipedia-API==0.6.0
+wordninja==2.0.0
+wrapt==1.12.1
+xmltodict==0.13.0
+xxhash==2.0.0
+yamllint==1.23.0
+yarg==0.1.9
+yarl==1.7.2
+yaspin==2.1.0
+youtokentome==1.0.6
+youtube-dl==2021.6.6
+zipp==3.6.0
+==============================
+Git revision:
+082323a
+==============================
+CUDA version:
+11.7

semi_wavlm_large_tunisian_ctc/1234/hyperparams.yaml

@@ -0,0 +1,177 @@
+# Generated 2023-09-25 from:
+# /home/salah/Tunisian_Automatic_Speech_Recognition/train_semi.yaml
+# yamllint disable
+# ################################
+# Model: wav2vec2 + DNN + CTC
+# Augmentation: SpecAugment
+# Authors: Titouan Parcollet 2021
+# ################################
+
+# Seed needs to be set at top of yaml, before objects with parameters are made
+seed: 1234
+__set_seed: !!python/object/apply:torch.manual_seed [1234]
+output_folder: semi_wavlm_large_tunisian_ctc/1234
+wer_file: semi_wavlm_large_tunisian_ctc/1234/wer.txt
+save_folder: semi_wavlm_large_tunisian_ctc/1234/save
+train_log: semi_wavlm_large_tunisian_ctc/1234/train_log.txt
+
+# URL for the biggest LeBenchmark wav2vec french.
+wav2vec2_folder: semi_wavlm_large_tunisian_ctc/1234/save/wav2vec2_checkpoint
+
+# Data files
+data_folder: /path/to/data  # e.g, /localscratch/cv-corpus-5.1-2020-06-22/fr
+train_tsv_file: /path/to/data/train.tsv       # Standard CommonVoice .tsv files
+dev_tsv_file: /path/to/data/dev.tsv       # Standard CommonVoice .tsv files
+test_tsv_file: /path/to/data/test.tsv       # Standard CommonVoice .tsv files
+accented_letters: true
+language: fr # use 'it' for Italian, 'rw' for Kinyarwanda, 'en' for english
+test_csv:
+- /path/to/test_data
+
+skip_prep: true # Skip data preparation
+
+use_language_modelling: true
+ngram_lm_path: outdomain.arpa
+
+# We remove utterance slonger than 10s in the train/dev/test sets as
+# longer sentences certainly correspond to "open microphones".
+avoid_if_longer_than: 10.0
+avoid_if_shorter_than: 1.2
+
+
+# Training parameters
+number_of_epochs: 12
+lr: 1.0
+lr_wav2vec: 0.0001
+sorting: ascending
+auto_mix_prec: false
+sample_rate: 16000
+ckpt_interval_minutes: 30 # save checkpoint every N min
+
+# With data_parallel batch_size is split into N jobs
+# With DDP batch_size is multiplied by N jobs
+# Must be 6 per GPU to fit 16GB of VRAM
+batch_size: 10
+test_batch_size: 4
+
+dataloader_options:
+  batch_size: 10
+  num_workers: 6
+test_dataloader_options:
+  batch_size: 4
+  num_workers: 6
+
+# BPE parameters
+token_type: char  # ["unigram", "bpe", "char"]
+character_coverage: 1.0
+
+# Model parameters
+# activation: !name:torch.nn.LeakyReLU
+wav2vec_output_dim: 1024
+dnn_neurons: 1024
+freeze_wav2vec: false
+freeze_feature_extractor: true
+dropout: 0.15
+warmup_steps: 500 # The wav2vec 2 model isn't updated for this amount of steps
+
+# Outputs
+output_neurons: 40  # BPE size, index(blank/eos/bos) = 0
+
+# Decoding parameters
+# Be sure that the bos and eos index match with the BPEs ones
+blank_index: 0
+unk_index: 1
+
+#
+# Functions and classes
+#
+epoch_counter: &id007 !new:speechbrain.utils.epoch_loop.EpochCounter
+
+  limit: 12
+
+augmentation: !new:speechbrain.lobes.augment.TimeDomainSpecAugment
+  sample_rate: 16000
+  speeds: [95, 100, 105]
+
+enc: &id002 !new:speechbrain.nnet.containers.Sequential
+  input_shape: [null, null, 1024]
+  linear1: !name:speechbrain.nnet.linear.Linear
+    n_neurons: 1024
+    bias: true
+  bn1: !name:speechbrain.nnet.normalization.BatchNorm1d
+  activation: !new:torch.nn.LeakyReLU
+  drop: !new:torch.nn.Dropout
+    p: 0.15
+  linear2: !name:speechbrain.nnet.linear.Linear
+    n_neurons: 1024
+    bias: true
+  bn2: !name:speechbrain.nnet.normalization.BatchNorm1d
+  activation2: !new:torch.nn.LeakyReLU
+  drop2: !new:torch.nn.Dropout
+    p: 0.15
+  linear3: !name:speechbrain.nnet.linear.Linear
+    n_neurons: 1024
+    bias: true
+  bn3: !name:speechbrain.nnet.normalization.BatchNorm1d
+  activation3: !new:torch.nn.LeakyReLU
+
+wav2vec2: &id001 !new:speechbrain.lobes.models.huggingface_wav2vec.HuggingFaceWav2Vec2
+  source: wavlm-large/
+  output_norm: false
+  freeze: false
+  freeze_feature_extractor: true
+  save_path: semi_wavlm_large_tunisian_ctc/1234/save/wav2vec2_checkpoint
+
+
+ctc_lin: &id003 !new:speechbrain.nnet.linear.Linear
+
+  input_size: 1024
+  n_neurons: 40
+
+log_softmax: !new:speechbrain.nnet.activations.Softmax
+  apply_log: true
+
+ctc_cost: !name:speechbrain.nnet.losses.ctc_loss
+  blank_index: 0
+
+modules:
+  wav2vec2: *id001
+  enc: *id002
+  ctc_lin: *id003
+model: &id004 !new:torch.nn.ModuleList
+- [*id002, *id003]
+model_opt_class: !name:torch.optim.Adadelta
+  lr: 1.0
+  rho: 0.95
+  eps: 1.e-8
+
+wav2vec_opt_class: !name:torch.optim.Adam
+  lr: 0.0001
+
+lr_annealing_model: &id005 !new:speechbrain.nnet.schedulers.NewBobScheduler
+  initial_value: 1.0
+  improvement_threshold: 0.0025
+  annealing_factor: 0.8
+  patient: 0
+
+lr_annealing_wav2vec: &id006 !new:speechbrain.nnet.schedulers.NewBobScheduler
+  initial_value: 0.0001
+  improvement_threshold: 0.0025
+  annealing_factor: 0.9
+  patient: 0
+
+checkpointer: !new:speechbrain.utils.checkpoints.Checkpointer
+  checkpoints_dir: semi_wavlm_large_tunisian_ctc/1234/save
+  recoverables:
+    wav2vec2: *id001
+    model: *id004
+    scheduler_model: *id005
+    scheduler_wav2vec: *id006
+    counter: *id007
+train_logger: !new:speechbrain.utils.train_logger.FileTrainLogger
+  save_file: semi_wavlm_large_tunisian_ctc/1234/train_log.txt
+
+error_rate_computer: !name:speechbrain.utils.metric_stats.ErrorRateStats
+
+cer_computer: !name:speechbrain.utils.metric_stats.ErrorRateStats
+  split_tokens: true

semi_wavlm_large_tunisian_ctc/1234/log.txt

@@ -0,0 +1,2270 @@
+2023-09-25 11:52:05,502 - speechbrain.core - INFO - Beginning experiment!
+2023-09-25 11:52:05,502 - speechbrain.core - INFO - Experiment folder: semi_wavlm_large_tunisian_ctc/1234
+2023-09-25 11:52:06,047 - speechbrain.utils.superpowers - DEBUG - abkhazia==1.0
+absl-py==0.11.0
+aiofiles==23.2.1
+aiohttp==3.8.0
+aiosignal==1.2.0
+alabaster==0.7.12
+alembic==1.7.4
+altair==4.2.0
+altgraph==0.17
+antlr4-python3-runtime==4.9.3
+anyio==3.6.2
+appdirs==1.4.4
+argcomplete==1.12.2
+argon2-cffi==20.1.0
+arrow==1.2.3
+asgiref==3.6.0
+asteroid-filterbanks==0.4.0
+astunparse==1.6.3
+async-generator==1.10
+async-timeout==4.0.0
+attrdict==2.0.1
+attrs==20.3.0
+audeer==1.16.0
+audformat==0.11.5
+audinterface==0.7.0
+audiofile==1.0.0
+audiomentations==0.25.0
+audioread==2.1.9
+audobject==0.4.14
+audresample==0.1.6
+-e git+https://github.com/facebookresearch/WavAugment.git@54afcdb00ccc852c2f030f239f8532c9562b550e#egg=augment
+autopage==0.4.0
+Babel==2.9.0
+backcall==0.2.0
+backports.cached-property==1.0.2
+beautifulsoup4==4.10.0
+black==19.10b0
+bleach==3.3.0
+blessed==1.20.0
+boto3==1.20.2
+botocore==1.23.2
+bpemb==0.3.4
+braceexpand==0.1.7
+cachetools==4.2.0
+certifi @ file:///croot/certifi_1671487769961/work/certifi
+cffi==1.14.3
+cfgv==3.2.0
+chardet==3.0.4
+charset-normalizer==2.0.7
+click==7.1.2
+cliff==3.9.0
+clldutils==3.5.4
+cloudpickle==2.2.1
+cmaes==0.8.2
+cmake==3.18.4.post1
+cmd2==2.2.0
+colorama==0.4.4
+colorlog==4.6.2
+configparser==5.1.0
+conllu==4.5.3
+croniter==1.3.15
+cryptography==38.0.4
+csrgraph==0.1.28
+csvw==1.8.1
+cycler==0.10.0
+Cython==0.29.21
+dataclasses==0.6
+dateutils==0.6.12
+decorator==4.4.2
+deepdiff==6.3.0
+deepspeech==0.9.1
+defusedxml==0.7.1
+Deprecated==1.2.14
+dill==0.3.3
+Distance==0.1.3
+distlib==0.3.1
+Django==3.2.16
+django-auditlog==2.2.1
+django-filter==22.1
+django-js-asset==1.2.2
+django-mptt==0.14.0
+djangorestframework==3.14.0
+docker-pycreds==0.4.0
+docopt==0.6.2
+docutils==0.16
+drf-excel==2.2.0
+drf-flex-fields==1.0.0
+drf-renderer-xlsx==0.4.1
+easyocr==1.2.1
+editdistance==0.6.0
+einops==0.3.2
+emoji==2.2.0
+entrypoints==0.3
+et-xmlfile==1.1.0
+exceptiongroup==1.1.0
+farasapy==0.0.14
+fastapi==0.98.0
+fastjsonschema==2.17.1
+fasttext==0.9.2
+ffmpeg-python==0.2.0
+ffmpy==0.3.0
+filelock==3.0.12
+flair==0.12.2
+flake8==3.7.9
+flatbuffers==1.12
+frozendict==2.0.7
+frozenlist==1.2.0
+fsspec==2021.11.0
+ftfy==6.1.1
+future==0.18.2
+g2p-en==2.1.0
+gast==0.3.3
+gdown==4.4.0
+gdrive==0.1.5
+gensim==4.0.1
+gitdb==4.0.9
+GitPython==3.1.24
+google-api-core==2.11.1
+google-api-python-client==2.43.0
+google-auth==1.24.0
+google-auth-httplib2==0.1.0
+google-auth-oauthlib==0.5.3
+google-pasta==0.2.0
+googleapis-common-protos==1.59.1
+gradio==3.44.4
+gradio-client==0.5.1
+greenlet==1.1.2
+grpcio==1.32.0
+h11==0.14.0
+h5features==1.3.2
+h5py==2.10.0
+hierarchy==0.4.0
+hmmlearn==0.2.8
+htk-io==0.5
+httpcore==0.16.3
+httplib2==0.22.0
+httpx==0.23.3
+huggingface-hub==0.15.1
+hydra-colorlog==0.1.4
+hydra-core==1.3.2
+hyperopt==0.2.7
+HyperPyYAML==1.1.0
+hypothesis==6.61.2
+identify==1.5.10
+idna==2.10
+imageio==2.9.0
+imagesize==1.2.0
+importlib-metadata==4.8.1
+importlib-resources==5.2.2
+inflect==5.3.0
+inquirer==3.1.3
+ipadic==1.0.0
+ipyevents==2.0.1
+ipykernel==5.3.4
+ipython==7.19.0
+ipython-genutils==0.2.0
+ipywebrtc==0.6.0
+ipywidgets==7.6.3
+iso-639==0.4.5
+isodate==0.6.0
+isort==4.3.21
+itsdangerous==2.1.2
+Janome==0.5.0
+jedi==0.17.2
+jeepney==0.8.0
+jieba==0.42.1
+Jinja2==3.0.3
+jiwer==2.2.0
+jmespath==0.10.0
+joblib==0.17.0
+jsonschema==3.2.0
+julius==0.2.7
+jupyter-client==6.1.7
+jupyter-core==4.7.0
+jupyterlab-pygments==0.1.2
+jupyterlab-widgets==1.0.0
+kaitaistruct==0.9
+kaldi-io==0.9.4
+kaldi-python-io==1.2.2
+kaldiio==2.17.2
+kenlm @ https://github.com/kpu/kenlm/archive/master.zip
+Keras-Preprocessing==1.1.2
+kiwisolver==1.3.1
+lang-trans==0.6.0
+langdetect==1.0.9
+latexcodec==2.0.1
+ldap3==2.9.1
+librosa==0.9.0
+lightning-cloud==0.5.37
+lightning-utilities==0.8.0
+linkify-it-py==1.0.3
+lit==16.0.6
+llvmlite==0.35.0
+lxml==4.9.0
+Mako==1.1.5
+Markdown==3.3.3
+markdown-it-py==3.0.0
+MarkupSafe==2.1.3
+marshmallow==3.14.0
+matplotlib==3.3.3
+mccabe==0.6.1
+mcd==0.4
+mdit-py-plugins==0.3.3
+mdurl==0.1.2
+mecab-python3==1.0.3
+megatron-lm==2.2.0
+metrics==0.3.3
+mido==1.2.10
+mistune==0.8.4
+more-itertools==8.6.0
+mpld3==0.3
+mpmath==1.2.1
+multidict==5.2.0
+multiprocess==0.70.11.1
+nbclient==0.5.3
+nbconvert==5.6.1
+nbformat==5.9.0
+NEMO==4.3.2
+nemo-toolkit==1.4.0
+nest-asyncio==1.5.1
+networkx==2.8.8
+nltk==3.2.4
+nodeenv==1.5.0
+normalize==2.0.2
+notebook==6.3.0
+numba==0.52.0
+numpy==1.19.4
+nvidia-cublas-cu11==11.10.3.66
+nvidia-cuda-cupti-cu11==11.7.101
+nvidia-cuda-nvrtc-cu11==11.7.99
+nvidia-cuda-runtime-cu11==11.7.99
+nvidia-cudnn-cu11==8.5.0.96
+nvidia-cufft-cu11==10.9.0.58
+nvidia-curand-cu11==10.2.10.91
+nvidia-cusolver-cu11==11.4.0.1
+nvidia-cusparse-cu11==11.7.4.91
+nvidia-nccl-cu11==2.14.3
+nvidia-nvtx-cu11==11.7.91
+oauthlib==3.1.0
+omegaconf==2.3.0
+onnx==1.10.2
+OpenCC==1.1.2
+opencv-python==4.4.0.46
+openpyxl==3.0.9
+opensmile==2.2.0
+opt-einsum==3.3.0
+optuna==2.10.0
+ordered-set==4.1.0
+orjson==3.8.4
+oyaml==1.0
+packaging==22.0
+pandas==1.2.5
+pandocfilters==1.4.3
+pangu==4.0.6.1
+parameterized==0.8.1
+parso==0.7.1
+pathlib2==2.3.7.post1
+pathspec==0.5.5
+pathtools==0.1.2
+pbr==5.6.0
+pefile==2019.4.18
+pescador==2.1.0
+pesq==0.0.3
+pexpect==4.8.0
+phonemizer==2.2.1
+pickleshare==0.7.5
+Pillow==9.3.0
+pip-api==0.0.23
+pipreqs==0.4.11
+pluggy==0.13.1
+pooch==1.3.0
+portalocker==2.3.2
+pptree==3.1
+pre-commit==2.9.0
+preprocessing==0.1.13
+pretty-midi==0.2.9
+prettytable==2.2.1
+primePy==1.3
+progressbar2==3.53.1
+prometheus-client==0.10.1
+promise==2.3
+prompt-toolkit==3.0.8
+protobuf==3.20.3
+psutil==5.6.6
+ptyprocess==0.6.0
+py==1.9.0
+py-espeak-ng==0.1.8
+py4j==0.10.9.7
+pyannote.audio==2.1.1
+pyannote.core==4.5
+pyannote.database==4.1.3
+pyannote.metrics==3.2.1
+pyannote.pipeline==2.3
+pyannotebook==0.1.0.dev0
+PyArabic==0.6.15
+pyarrow==3.0.0
+pyasn1==0.4.8
+pyasn1-modules==0.2.8
+pybind11==2.8.1
+pybtex==0.24.0
+pybtex-docutils==1.0.1
+pycodestyle==2.5.0
+pycparser==2.20
+pycryptodome==3.16.0
+pyctcdecode==0.4.0
+pydantic==1.10.4
+pyDeprecate==0.3.1
+pydub==0.25.1
+pyflakes==2.1.1
+Pygments==2.15.1
+pygtrie==2.5.0
+PyJWT==2.7.0
+pymodbus==2.5.3
+pyparsing==2.4.7
+pyperclip==1.8.2
+pypinyin==0.43.0
+pyrsistent==0.17.3
+pyserial==3.5
+PySocks==1.7.1
+pystoi==0.3.3
+pytest==5.4.1
+pytest-runner==5.3.1
+python-bidi==0.4.2
+python-crfsuite==0.9.7
+python-dateutil==2.8.2
+python-editor==1.0.4
+python-Levenshtein==0.12.2
+python-multipart==0.0.5
+python-utils==2.4.0
+pytorch-lightning==1.6.5
+pytorch-metric-learning==1.7.3
+pytorch-revgrad==0.2.0
+pytube==11.0.1
+pytz==2022.6
+PyWavelets==1.1.1
+PyYAML==6.0
+pyzmq==20.0.0
+rapidfuzz==1.8.2
+readchar==4.0.5
+regex==2020.11.13
+requests==2.28.1
+requests-oauthlib==1.3.0
+resampy==0.2.2
+rfc3986==1.4.0
+rich==13.4.2
+richenum==1.3.1
+rsa==4.7
+ruamel.yaml==0.17.21
+ruamel.yaml.clib==0.2.7
+s3m==1.1.0
+s3transfer==0.5.0
+sacrebleu==2.0.0
+sacremoses==0.0.44
+safetensors==0.3.1
+scikit-image==0.18.1
+scikit-learn==0.23.2
+scipy==1.5.4
+-e git+https://github.com/sanghack81/SDCIT@00d060dde733fde9345154a494f81e97fb395ca7#egg=SDCIT
+seaborn==0.11.1
+SecretStorage==3.3.3
+segments==2.1.3
+segtok==1.5.11
+semantic-version==2.10.0
+semver==2.13.0
+Send2Trash==1.5.0
+sentencepiece==0.1.99
+sentry-sdk==1.4.3
+shellingham==1.4.0
+shortuuid==1.0.7
+SIDEKIT==1.3.8.5.2
+simplejson==3.17.5
+singledispatchmethod==1.0
+six==1.15.0
+smart-open==5.0.0
+smmap==5.0.0
+sniffio==1.3.0
+snowballstemmer==2.0.0
+sortedcollections==2.1.0
+sortedcontainers==2.4.0
+sounddevice==0.4.5
+SoundFile==0.10.3.post1
+soupsieve==2.3
+sox==1.4.1
+sparsemax==0.1.9
+speechbrain==0.5.14
+sphfile==1.0.3
+Sphinx==3.3.1
+sphinx-rtd-theme==0.2.4
+sphinxcontrib-applehelp==1.0.2
+sphinxcontrib-bibtex==2.4.1
+sphinxcontrib-devhelp==1.0.2
+sphinxcontrib-htmlhelp==1.0.3
+sphinxcontrib-jsmath==1.0.1
+sphinxcontrib-qthelp==1.0.3
+sphinxcontrib-serializinghtml==1.1.4
+SQLAlchemy==1.4.25
+sqlitedict==2.1.0
+sqlparse==0.4.2
+stanza==1.4.2
+starlette==0.27.0
+starsessions==1.3.0
+stevedore==3.4.0
+subprocess32==3.5.4
+sympy==1.9
+tabulate==0.8.9
+tensorboard==2.4.0
+tensorboard-plugin-wit==1.7.0
+tensorboardX==2.6.1
+tensorflow==2.4.0
+tensorflow-estimator==2.4.0
+termcolor==1.1.0
+terminado==0.9.4
+testpath==0.4.4
+threadpoolctl==2.1.0
+tifffile==2020.12.8
+tikzplotlib==0.9.8
+tinycss2==1.2.1
+tkseem==0.0.3
+tokenizers==0.13.3
+toml==0.10.2
+toolz==0.12.0
+torch==1.13.1
+torch-audiomentations==0.11.0
+torch-pitch-shift==1.2.4
+torch-stft==0.1.4
+torchaudio==0.13.1
+torchmetrics==0.11.4
+torchvision==0.14.1
+tornado==6.1
+tqdm==4.61.1
+trackrip==1.2.1
+traitlets==5.9.0
+transformer-smaller-training-vocab==0.3.1
+transformers==4.30.2
+triton==2.0.0
+typed-ast==1.4.1
+typer==0.4.0
+typing-extensions==4.4.0
+uc-micro-py==1.0.1
+Unidecode==1.3.2
+uritemplate==3.0.1
+urllib3==1.26.2
+uvicorn==0.20.0
+versioneer==0.28
+virtualenv==20.2.1
+wandb==0.12.6
+wcwidth==0.2.5
+webdataset==0.1.62
+webencodings==0.5.1
+websocket-client==1.6.1
+websockets==10.4
+Werkzeug==1.0.1
+wget==3.2
+widgetsnbextension==3.5.1
+Wikipedia-API==0.6.0
+wordninja==2.0.0
+wrapt==1.12.1
+xmltodict==0.13.0
+xxhash==2.0.0
+yamllint==1.23.0
+yarg==0.1.9
+yarl==1.7.2
+yaspin==2.1.0
+youtokentome==1.0.6
+youtube-dl==2021.6.6
+zipp==3.6.0
+
+
+2023-09-25 11:52:06,077 - speechbrain.utils.superpowers - DEBUG - 082323a
+
+
+2023-09-25 11:52:06,124 - speechbrain.core - ERROR - Exception:
+Traceback (most recent call last):
+  File "app.py", line 365, in <module>
+    train_data, valid_data, test_datasets, label_encoder = dataio_prepare(hparams)
+  File "app.py", line 248, in dataio_prepare
+    csv_path=hparams["train_csv"], replacements={"data_root": data_folder},
+KeyError: 'train_csv'
+2023-09-25 11:53:48,838 - speechbrain.core - INFO - Beginning experiment!
+2023-09-25 11:53:48,838 - speechbrain.core - INFO - Experiment folder: semi_wavlm_large_tunisian_ctc/1234
+2023-09-25 11:53:49,330 - speechbrain.utils.superpowers - DEBUG - abkhazia==1.0
+absl-py==0.11.0
+aiofiles==23.2.1
+aiohttp==3.8.0
+aiosignal==1.2.0
+alabaster==0.7.12
+alembic==1.7.4
+altair==4.2.0
+altgraph==0.17
+antlr4-python3-runtime==4.9.3
+anyio==3.6.2
+appdirs==1.4.4
+argcomplete==1.12.2
+argon2-cffi==20.1.0
+arrow==1.2.3
+asgiref==3.6.0
+asteroid-filterbanks==0.4.0
+astunparse==1.6.3
+async-generator==1.10
+async-timeout==4.0.0
+attrdict==2.0.1
+attrs==20.3.0
+audeer==1.16.0
+audformat==0.11.5
+audinterface==0.7.0
+audiofile==1.0.0
+audiomentations==0.25.0
+audioread==2.1.9
+audobject==0.4.14
+audresample==0.1.6
+-e git+https://github.com/facebookresearch/WavAugment.git@54afcdb00ccc852c2f030f239f8532c9562b550e#egg=augment
+autopage==0.4.0
+Babel==2.9.0
+backcall==0.2.0
+backports.cached-property==1.0.2
+beautifulsoup4==4.10.0
+black==19.10b0
+bleach==3.3.0
+blessed==1.20.0
+boto3==1.20.2
+botocore==1.23.2
+bpemb==0.3.4
+braceexpand==0.1.7
+cachetools==4.2.0
+certifi @ file:///croot/certifi_1671487769961/work/certifi
+cffi==1.14.3
+cfgv==3.2.0
+chardet==3.0.4
+charset-normalizer==2.0.7
+click==7.1.2
+cliff==3.9.0
+clldutils==3.5.4
+cloudpickle==2.2.1
+cmaes==0.8.2
+cmake==3.18.4.post1
+cmd2==2.2.0
+colorama==0.4.4
+colorlog==4.6.2
+configparser==5.1.0
+conllu==4.5.3
+croniter==1.3.15
+cryptography==38.0.4
+csrgraph==0.1.28
+csvw==1.8.1
+cycler==0.10.0
+Cython==0.29.21
+dataclasses==0.6
+dateutils==0.6.12
+decorator==4.4.2
+deepdiff==6.3.0
+deepspeech==0.9.1
+defusedxml==0.7.1
+Deprecated==1.2.14
+dill==0.3.3
+Distance==0.1.3
+distlib==0.3.1
+Django==3.2.16
+django-auditlog==2.2.1
+django-filter==22.1
+django-js-asset==1.2.2
+django-mptt==0.14.0
+djangorestframework==3.14.0
+docker-pycreds==0.4.0
+docopt==0.6.2
+docutils==0.16
+drf-excel==2.2.0
+drf-flex-fields==1.0.0
+drf-renderer-xlsx==0.4.1
+easyocr==1.2.1
+editdistance==0.6.0
+einops==0.3.2
+emoji==2.2.0
+entrypoints==0.3
+et-xmlfile==1.1.0
+exceptiongroup==1.1.0
+farasapy==0.0.14
+fastapi==0.98.0
+fastjsonschema==2.17.1
+fasttext==0.9.2
+ffmpeg-python==0.2.0
+ffmpy==0.3.0
+filelock==3.0.12
+flair==0.12.2
+flake8==3.7.9
+flatbuffers==1.12
+frozendict==2.0.7
+frozenlist==1.2.0
+fsspec==2021.11.0
+ftfy==6.1.1
+future==0.18.2
+g2p-en==2.1.0
+gast==0.3.3
+gdown==4.4.0
+gdrive==0.1.5
+gensim==4.0.1
+gitdb==4.0.9
+GitPython==3.1.24
+google-api-core==2.11.1
+google-api-python-client==2.43.0
+google-auth==1.24.0
+google-auth-httplib2==0.1.0
+google-auth-oauthlib==0.5.3
+google-pasta==0.2.0
+googleapis-common-protos==1.59.1
+gradio==3.44.4
+gradio-client==0.5.1
+greenlet==1.1.2
+grpcio==1.32.0
+h11==0.14.0
+h5features==1.3.2
+h5py==2.10.0
+hierarchy==0.4.0
+hmmlearn==0.2.8
+htk-io==0.5
+httpcore==0.16.3
+httplib2==0.22.0
+httpx==0.23.3
+huggingface-hub==0.15.1
+hydra-colorlog==0.1.4
+hydra-core==1.3.2
+hyperopt==0.2.7
+HyperPyYAML==1.1.0
+hypothesis==6.61.2
+identify==1.5.10
+idna==2.10
+imageio==2.9.0
+imagesize==1.2.0
+importlib-metadata==4.8.1
+importlib-resources==5.2.2
+inflect==5.3.0
+inquirer==3.1.3
+ipadic==1.0.0
+ipyevents==2.0.1
+ipykernel==5.3.4
+ipython==7.19.0
+ipython-genutils==0.2.0
+ipywebrtc==0.6.0
+ipywidgets==7.6.3
+iso-639==0.4.5
+isodate==0.6.0
+isort==4.3.21
+itsdangerous==2.1.2
+Janome==0.5.0
+jedi==0.17.2
+jeepney==0.8.0
+jieba==0.42.1
+Jinja2==3.0.3
+jiwer==2.2.0
+jmespath==0.10.0
+joblib==0.17.0
+jsonschema==3.2.0
+julius==0.2.7
+jupyter-client==6.1.7
+jupyter-core==4.7.0
+jupyterlab-pygments==0.1.2
+jupyterlab-widgets==1.0.0
+kaitaistruct==0.9
+kaldi-io==0.9.4
+kaldi-python-io==1.2.2
+kaldiio==2.17.2
+kenlm @ https://github.com/kpu/kenlm/archive/master.zip
+Keras-Preprocessing==1.1.2
+kiwisolver==1.3.1
+lang-trans==0.6.0
+langdetect==1.0.9
+latexcodec==2.0.1
+ldap3==2.9.1
+librosa==0.9.0
+lightning-cloud==0.5.37
+lightning-utilities==0.8.0
+linkify-it-py==1.0.3
+lit==16.0.6
+llvmlite==0.35.0
+lxml==4.9.0
+Mako==1.1.5
+Markdown==3.3.3
+markdown-it-py==3.0.0
+MarkupSafe==2.1.3
+marshmallow==3.14.0
+matplotlib==3.3.3
+mccabe==0.6.1
+mcd==0.4
+mdit-py-plugins==0.3.3
+mdurl==0.1.2
+mecab-python3==1.0.3
+megatron-lm==2.2.0
+metrics==0.3.3
+mido==1.2.10
+mistune==0.8.4
+more-itertools==8.6.0
+mpld3==0.3
+mpmath==1.2.1
+multidict==5.2.0
+multiprocess==0.70.11.1
+nbclient==0.5.3
+nbconvert==5.6.1
+nbformat==5.9.0
+NEMO==4.3.2
+nemo-toolkit==1.4.0
+nest-asyncio==1.5.1
+networkx==2.8.8
+nltk==3.2.4
+nodeenv==1.5.0
+normalize==2.0.2
+notebook==6.3.0
+numba==0.52.0
+numpy==1.19.4
+nvidia-cublas-cu11==11.10.3.66
+nvidia-cuda-cupti-cu11==11.7.101
+nvidia-cuda-nvrtc-cu11==11.7.99
+nvidia-cuda-runtime-cu11==11.7.99
+nvidia-cudnn-cu11==8.5.0.96
+nvidia-cufft-cu11==10.9.0.58
+nvidia-curand-cu11==10.2.10.91
+nvidia-cusolver-cu11==11.4.0.1
+nvidia-cusparse-cu11==11.7.4.91
+nvidia-nccl-cu11==2.14.3
+nvidia-nvtx-cu11==11.7.91
+oauthlib==3.1.0
+omegaconf==2.3.0
+onnx==1.10.2
+OpenCC==1.1.2
+opencv-python==4.4.0.46
+openpyxl==3.0.9
+opensmile==2.2.0
+opt-einsum==3.3.0
+optuna==2.10.0
+ordered-set==4.1.0
+orjson==3.8.4
+oyaml==1.0
+packaging==22.0
+pandas==1.2.5
+pandocfilters==1.4.3
+pangu==4.0.6.1
+parameterized==0.8.1
+parso==0.7.1
+pathlib2==2.3.7.post1
+pathspec==0.5.5
+pathtools==0.1.2
+pbr==5.6.0
+pefile==2019.4.18
+pescador==2.1.0
+pesq==0.0.3
+pexpect==4.8.0
+phonemizer==2.2.1
+pickleshare==0.7.5
+Pillow==9.3.0
+pip-api==0.0.23
+pipreqs==0.4.11
+pluggy==0.13.1
+pooch==1.3.0
+portalocker==2.3.2
+pptree==3.1
+pre-commit==2.9.0
+preprocessing==0.1.13
+pretty-midi==0.2.9
+prettytable==2.2.1
+primePy==1.3
+progressbar2==3.53.1
+prometheus-client==0.10.1
+promise==2.3
+prompt-toolkit==3.0.8
+protobuf==3.20.3
+psutil==5.6.6
+ptyprocess==0.6.0
+py==1.9.0
+py-espeak-ng==0.1.8
+py4j==0.10.9.7
+pyannote.audio==2.1.1
+pyannote.core==4.5
+pyannote.database==4.1.3
+pyannote.metrics==3.2.1
+pyannote.pipeline==2.3
+pyannotebook==0.1.0.dev0
+PyArabic==0.6.15
+pyarrow==3.0.0
+pyasn1==0.4.8
+pyasn1-modules==0.2.8
+pybind11==2.8.1
+pybtex==0.24.0
+pybtex-docutils==1.0.1
+pycodestyle==2.5.0
+pycparser==2.20
+pycryptodome==3.16.0
+pyctcdecode==0.4.0
+pydantic==1.10.4
+pyDeprecate==0.3.1
+pydub==0.25.1
+pyflakes==2.1.1
+Pygments==2.15.1
+pygtrie==2.5.0
+PyJWT==2.7.0
+pymodbus==2.5.3
+pyparsing==2.4.7
+pyperclip==1.8.2
+pypinyin==0.43.0
+pyrsistent==0.17.3
+pyserial==3.5
+PySocks==1.7.1
+pystoi==0.3.3
+pytest==5.4.1
+pytest-runner==5.3.1
+python-bidi==0.4.2
+python-crfsuite==0.9.7
+python-dateutil==2.8.2
+python-editor==1.0.4
+python-Levenshtein==0.12.2
+python-multipart==0.0.5
+python-utils==2.4.0
+pytorch-lightning==1.6.5
+pytorch-metric-learning==1.7.3
+pytorch-revgrad==0.2.0
+pytube==11.0.1
+pytz==2022.6
+PyWavelets==1.1.1
+PyYAML==6.0
+pyzmq==20.0.0
+rapidfuzz==1.8.2
+readchar==4.0.5
+regex==2020.11.13
+requests==2.28.1
+requests-oauthlib==1.3.0
+resampy==0.2.2
+rfc3986==1.4.0
+rich==13.4.2
+richenum==1.3.1
+rsa==4.7
+ruamel.yaml==0.17.21
+ruamel.yaml.clib==0.2.7
+s3m==1.1.0
+s3transfer==0.5.0
+sacrebleu==2.0.0
+sacremoses==0.0.44
+safetensors==0.3.1
+scikit-image==0.18.1
+scikit-learn==0.23.2
+scipy==1.5.4
+-e git+https://github.com/sanghack81/SDCIT@00d060dde733fde9345154a494f81e97fb395ca7#egg=SDCIT
+seaborn==0.11.1
+SecretStorage==3.3.3
+segments==2.1.3
+segtok==1.5.11
+semantic-version==2.10.0
+semver==2.13.0
+Send2Trash==1.5.0
+sentencepiece==0.1.99
+sentry-sdk==1.4.3
+shellingham==1.4.0
+shortuuid==1.0.7
+SIDEKIT==1.3.8.5.2
+simplejson==3.17.5
+singledispatchmethod==1.0
+six==1.15.0
+smart-open==5.0.0
+smmap==5.0.0
+sniffio==1.3.0
+snowballstemmer==2.0.0
+sortedcollections==2.1.0
+sortedcontainers==2.4.0
+sounddevice==0.4.5
+SoundFile==0.10.3.post1
+soupsieve==2.3
+sox==1.4.1
+sparsemax==0.1.9
+speechbrain==0.5.14
+sphfile==1.0.3
+Sphinx==3.3.1
+sphinx-rtd-theme==0.2.4
+sphinxcontrib-applehelp==1.0.2
+sphinxcontrib-bibtex==2.4.1
+sphinxcontrib-devhelp==1.0.2
+sphinxcontrib-htmlhelp==1.0.3
+sphinxcontrib-jsmath==1.0.1
+sphinxcontrib-qthelp==1.0.3
+sphinxcontrib-serializinghtml==1.1.4
+SQLAlchemy==1.4.25
+sqlitedict==2.1.0
+sqlparse==0.4.2
+stanza==1.4.2
+starlette==0.27.0
+starsessions==1.3.0
+stevedore==3.4.0
+subprocess32==3.5.4
+sympy==1.9
+tabulate==0.8.9
+tensorboard==2.4.0
+tensorboard-plugin-wit==1.7.0
+tensorboardX==2.6.1
+tensorflow==2.4.0
+tensorflow-estimator==2.4.0
+termcolor==1.1.0
+terminado==0.9.4
+testpath==0.4.4
+threadpoolctl==2.1.0
+tifffile==2020.12.8
+tikzplotlib==0.9.8
+tinycss2==1.2.1
+tkseem==0.0.3
+tokenizers==0.13.3
+toml==0.10.2
+toolz==0.12.0
+torch==1.13.1
+torch-audiomentations==0.11.0
+torch-pitch-shift==1.2.4
+torch-stft==0.1.4
+torchaudio==0.13.1
+torchmetrics==0.11.4
+torchvision==0.14.1
+tornado==6.1
+tqdm==4.61.1
+trackrip==1.2.1
+traitlets==5.9.0
+transformer-smaller-training-vocab==0.3.1
+transformers==4.30.2
+triton==2.0.0
+typed-ast==1.4.1
+typer==0.4.0
+typing-extensions==4.4.0
+uc-micro-py==1.0.1
+Unidecode==1.3.2
+uritemplate==3.0.1
+urllib3==1.26.2
+uvicorn==0.20.0
+versioneer==0.28
+virtualenv==20.2.1
+wandb==0.12.6
+wcwidth==0.2.5
+webdataset==0.1.62
+webencodings==0.5.1
+websocket-client==1.6.1
+websockets==10.4
+Werkzeug==1.0.1
+wget==3.2
+widgetsnbextension==3.5.1
+Wikipedia-API==0.6.0
+wordninja==2.0.0
+wrapt==1.12.1
+xmltodict==0.13.0
+xxhash==2.0.0
+yamllint==1.23.0
+yarg==0.1.9
+yarl==1.7.2
+yaspin==2.1.0
+youtokentome==1.0.6
+youtube-dl==2021.6.6
+zipp==3.6.0
+
+
+2023-09-25 11:53:49,358 - speechbrain.utils.superpowers - DEBUG - 082323a
+
+
+2023-09-25 11:53:49,390 - speechbrain.dataio.encoder - DEBUG - Loaded categorical encoding from semi_wavlm_large_tunisian_ctc/1234/save/label_encoder.txt
+2023-09-25 11:53:49,390 - speechbrain.dataio.encoder - INFO - Load called, but CTCTextEncoder is not empty. Loaded data will overwrite everything. This is normal if there is e.g. an unk label defined at init.
+2023-09-25 11:53:49,391 - speechbrain.dataio.encoder - DEBUG - Loaded categorical encoding from semi_wavlm_large_tunisian_ctc/1234/save/label_encoder.txt
+2023-09-25 11:53:51,607 - pyctcdecode.decoder - INFO - Using arpa instead of binary LM file, decoder instantiation might be slow.
+2023-09-25 11:53:51,713 - pyctcdecode.alphabet - INFO - Alphabet determined to be of regular style.
+2023-09-25 11:53:51,774 - pyctcdecode.alphabet - WARNING - Unigrams and labels don't seem to agree.
+2023-09-25 11:53:52,870 - speechbrain.core - INFO - Info: auto_mix_prec arg from hparam file is used
+2023-09-25 11:53:52,871 - speechbrain.core - INFO - Info: ckpt_interval_minutes arg from hparam file is used
+2023-09-25 11:53:53,969 - speechbrain.core - INFO - 314.4M trainable parameters in ASR
+2023-09-25 11:53:53,970 - speechbrain.utils.checkpoints - INFO - Loading a checkpoint from semi_wavlm_large_tunisian_ctc/1234/save/CKPT+2023-09-05+01-09-23+00
+2023-09-25 11:53:54,516 - speechbrain.core - ERROR - Exception:
+Traceback (most recent call last):
+  File "app.py", line 422, in <module>
+    gr.Interface(
+NameError: name 'gr' is not defined
+2023-09-25 11:54:18,422 - speechbrain.core - INFO - Beginning experiment!
+2023-09-25 11:54:18,422 - speechbrain.core - INFO - Experiment folder: semi_wavlm_large_tunisian_ctc/1234
+2023-09-25 11:54:18,903 - speechbrain.utils.superpowers - DEBUG - abkhazia==1.0
+absl-py==0.11.0
+aiofiles==23.2.1
+aiohttp==3.8.0
+aiosignal==1.2.0
+alabaster==0.7.12
+alembic==1.7.4
+altair==4.2.0
+altgraph==0.17
+antlr4-python3-runtime==4.9.3
+anyio==3.6.2
+appdirs==1.4.4
+argcomplete==1.12.2
+argon2-cffi==20.1.0
+arrow==1.2.3
+asgiref==3.6.0
+asteroid-filterbanks==0.4.0
+astunparse==1.6.3
+async-generator==1.10
+async-timeout==4.0.0
+attrdict==2.0.1
+attrs==20.3.0
+audeer==1.16.0
+audformat==0.11.5
+audinterface==0.7.0
+audiofile==1.0.0
+audiomentations==0.25.0
+audioread==2.1.9
+audobject==0.4.14
+audresample==0.1.6
+-e git+https://github.com/facebookresearch/WavAugment.git@54afcdb00ccc852c2f030f239f8532c9562b550e#egg=augment
+autopage==0.4.0
+Babel==2.9.0
+backcall==0.2.0
+backports.cached-property==1.0.2
+beautifulsoup4==4.10.0
+black==19.10b0
+bleach==3.3.0
+blessed==1.20.0
+boto3==1.20.2
+botocore==1.23.2
+bpemb==0.3.4
+braceexpand==0.1.7
+cachetools==4.2.0
+certifi @ file:///croot/certifi_1671487769961/work/certifi
+cffi==1.14.3
+cfgv==3.2.0
+chardet==3.0.4
+charset-normalizer==2.0.7
+click==7.1.2
+cliff==3.9.0
+clldutils==3.5.4
+cloudpickle==2.2.1
+cmaes==0.8.2
+cmake==3.18.4.post1
+cmd2==2.2.0
+colorama==0.4.4
+colorlog==4.6.2
+configparser==5.1.0
+conllu==4.5.3
+croniter==1.3.15
+cryptography==38.0.4
+csrgraph==0.1.28
+csvw==1.8.1
+cycler==0.10.0
+Cython==0.29.21
+dataclasses==0.6
+dateutils==0.6.12
+decorator==4.4.2
+deepdiff==6.3.0
+deepspeech==0.9.1
+defusedxml==0.7.1
+Deprecated==1.2.14
+dill==0.3.3
+Distance==0.1.3
+distlib==0.3.1
+Django==3.2.16
+django-auditlog==2.2.1
+django-filter==22.1
+django-js-asset==1.2.2
+django-mptt==0.14.0
+djangorestframework==3.14.0
+docker-pycreds==0.4.0
+docopt==0.6.2
+docutils==0.16
+drf-excel==2.2.0
+drf-flex-fields==1.0.0
+drf-renderer-xlsx==0.4.1
+easyocr==1.2.1
+editdistance==0.6.0
+einops==0.3.2
+emoji==2.2.0
+entrypoints==0.3
+et-xmlfile==1.1.0
+exceptiongroup==1.1.0
+farasapy==0.0.14
+fastapi==0.98.0
+fastjsonschema==2.17.1
+fasttext==0.9.2
+ffmpeg-python==0.2.0
+ffmpy==0.3.0
+filelock==3.0.12
+flair==0.12.2
+flake8==3.7.9
+flatbuffers==1.12
+frozendict==2.0.7
+frozenlist==1.2.0
+fsspec==2021.11.0
+ftfy==6.1.1
+future==0.18.2
+g2p-en==2.1.0
+gast==0.3.3
+gdown==4.4.0
+gdrive==0.1.5
+gensim==4.0.1
+gitdb==4.0.9
+GitPython==3.1.24
+google-api-core==2.11.1
+google-api-python-client==2.43.0
+google-auth==1.24.0
+google-auth-httplib2==0.1.0
+google-auth-oauthlib==0.5.3
+google-pasta==0.2.0
+googleapis-common-protos==1.59.1
+gradio==3.44.4
+gradio-client==0.5.1
+greenlet==1.1.2
+grpcio==1.32.0
+h11==0.14.0
+h5features==1.3.2
+h5py==2.10.0
+hierarchy==0.4.0
+hmmlearn==0.2.8
+htk-io==0.5
+httpcore==0.16.3
+httplib2==0.22.0
+httpx==0.23.3
+huggingface-hub==0.15.1
+hydra-colorlog==0.1.4
+hydra-core==1.3.2
+hyperopt==0.2.7
+HyperPyYAML==1.1.0
+hypothesis==6.61.2
+identify==1.5.10
+idna==2.10
+imageio==2.9.0
+imagesize==1.2.0
+importlib-metadata==4.8.1
+importlib-resources==5.2.2
+inflect==5.3.0
+inquirer==3.1.3
+ipadic==1.0.0
+ipyevents==2.0.1
+ipykernel==5.3.4
+ipython==7.19.0
+ipython-genutils==0.2.0
+ipywebrtc==0.6.0
+ipywidgets==7.6.3
+iso-639==0.4.5
+isodate==0.6.0
+isort==4.3.21
+itsdangerous==2.1.2
+Janome==0.5.0
+jedi==0.17.2
+jeepney==0.8.0
+jieba==0.42.1
+Jinja2==3.0.3
+jiwer==2.2.0
+jmespath==0.10.0
+joblib==0.17.0
+jsonschema==3.2.0
+julius==0.2.7
+jupyter-client==6.1.7
+jupyter-core==4.7.0
+jupyterlab-pygments==0.1.2
+jupyterlab-widgets==1.0.0
+kaitaistruct==0.9
+kaldi-io==0.9.4
+kaldi-python-io==1.2.2
+kaldiio==2.17.2
+kenlm @ https://github.com/kpu/kenlm/archive/master.zip
+Keras-Preprocessing==1.1.2
+kiwisolver==1.3.1
+lang-trans==0.6.0
+langdetect==1.0.9
+latexcodec==2.0.1
+ldap3==2.9.1
+librosa==0.9.0
+lightning-cloud==0.5.37
+lightning-utilities==0.8.0
+linkify-it-py==1.0.3
+lit==16.0.6
+llvmlite==0.35.0
+lxml==4.9.0
+Mako==1.1.5
+Markdown==3.3.3
+markdown-it-py==3.0.0
+MarkupSafe==2.1.3
+marshmallow==3.14.0
+matplotlib==3.3.3
+mccabe==0.6.1
+mcd==0.4
+mdit-py-plugins==0.3.3
+mdurl==0.1.2
+mecab-python3==1.0.3
+megatron-lm==2.2.0
+metrics==0.3.3
+mido==1.2.10
+mistune==0.8.4
+more-itertools==8.6.0
+mpld3==0.3
+mpmath==1.2.1
+multidict==5.2.0
+multiprocess==0.70.11.1
+nbclient==0.5.3
+nbconvert==5.6.1
+nbformat==5.9.0
+NEMO==4.3.2
+nemo-toolkit==1.4.0
+nest-asyncio==1.5.1
+networkx==2.8.8
+nltk==3.2.4
+nodeenv==1.5.0
+normalize==2.0.2
+notebook==6.3.0
+numba==0.52.0
+numpy==1.19.4
+nvidia-cublas-cu11==11.10.3.66
+nvidia-cuda-cupti-cu11==11.7.101
+nvidia-cuda-nvrtc-cu11==11.7.99
+nvidia-cuda-runtime-cu11==11.7.99
+nvidia-cudnn-cu11==8.5.0.96
+nvidia-cufft-cu11==10.9.0.58
+nvidia-curand-cu11==10.2.10.91
+nvidia-cusolver-cu11==11.4.0.1
+nvidia-cusparse-cu11==11.7.4.91
+nvidia-nccl-cu11==2.14.3
+nvidia-nvtx-cu11==11.7.91
+oauthlib==3.1.0
+omegaconf==2.3.0
+onnx==1.10.2
+OpenCC==1.1.2
+opencv-python==4.4.0.46
+openpyxl==3.0.9
+opensmile==2.2.0
+opt-einsum==3.3.0
+optuna==2.10.0
+ordered-set==4.1.0
+orjson==3.8.4
+oyaml==1.0
+packaging==22.0
+pandas==1.2.5
+pandocfilters==1.4.3
+pangu==4.0.6.1
+parameterized==0.8.1
+parso==0.7.1
+pathlib2==2.3.7.post1
+pathspec==0.5.5
+pathtools==0.1.2
+pbr==5.6.0
+pefile==2019.4.18
+pescador==2.1.0
+pesq==0.0.3
+pexpect==4.8.0
+phonemizer==2.2.1
+pickleshare==0.7.5
+Pillow==9.3.0
+pip-api==0.0.23
+pipreqs==0.4.11
+pluggy==0.13.1
+pooch==1.3.0
+portalocker==2.3.2
+pptree==3.1
+pre-commit==2.9.0
+preprocessing==0.1.13
+pretty-midi==0.2.9
+prettytable==2.2.1
+primePy==1.3
+progressbar2==3.53.1
+prometheus-client==0.10.1
+promise==2.3
+prompt-toolkit==3.0.8
+protobuf==3.20.3
+psutil==5.6.6
+ptyprocess==0.6.0
+py==1.9.0
+py-espeak-ng==0.1.8
+py4j==0.10.9.7
+pyannote.audio==2.1.1
+pyannote.core==4.5
+pyannote.database==4.1.3
+pyannote.metrics==3.2.1
+pyannote.pipeline==2.3
+pyannotebook==0.1.0.dev0
+PyArabic==0.6.15
+pyarrow==3.0.0
+pyasn1==0.4.8
+pyasn1-modules==0.2.8
+pybind11==2.8.1
+pybtex==0.24.0
+pybtex-docutils==1.0.1
+pycodestyle==2.5.0
+pycparser==2.20
+pycryptodome==3.16.0
+pyctcdecode==0.4.0
+pydantic==1.10.4
+pyDeprecate==0.3.1
+pydub==0.25.1
+pyflakes==2.1.1
+Pygments==2.15.1
+pygtrie==2.5.0
+PyJWT==2.7.0
+pymodbus==2.5.3
+pyparsing==2.4.7
+pyperclip==1.8.2
+pypinyin==0.43.0
+pyrsistent==0.17.3
+pyserial==3.5
+PySocks==1.7.1
+pystoi==0.3.3
+pytest==5.4.1
+pytest-runner==5.3.1
+python-bidi==0.4.2
+python-crfsuite==0.9.7
+python-dateutil==2.8.2
+python-editor==1.0.4
+python-Levenshtein==0.12.2
+python-multipart==0.0.5
+python-utils==2.4.0
+pytorch-lightning==1.6.5
+pytorch-metric-learning==1.7.3
+pytorch-revgrad==0.2.0
+pytube==11.0.1
+pytz==2022.6
+PyWavelets==1.1.1
+PyYAML==6.0
+pyzmq==20.0.0
+rapidfuzz==1.8.2
+readchar==4.0.5
+regex==2020.11.13
+requests==2.28.1
+requests-oauthlib==1.3.0
+resampy==0.2.2
+rfc3986==1.4.0
+rich==13.4.2
+richenum==1.3.1
+rsa==4.7
+ruamel.yaml==0.17.21
+ruamel.yaml.clib==0.2.7
+s3m==1.1.0
+s3transfer==0.5.0
+sacrebleu==2.0.0
+sacremoses==0.0.44
+safetensors==0.3.1
+scikit-image==0.18.1
+scikit-learn==0.23.2
+scipy==1.5.4
+-e git+https://github.com/sanghack81/SDCIT@00d060dde733fde9345154a494f81e97fb395ca7#egg=SDCIT
+seaborn==0.11.1
+SecretStorage==3.3.3
+segments==2.1.3
+segtok==1.5.11
+semantic-version==2.10.0
+semver==2.13.0
+Send2Trash==1.5.0
+sentencepiece==0.1.99
+sentry-sdk==1.4.3
+shellingham==1.4.0
+shortuuid==1.0.7
+SIDEKIT==1.3.8.5.2
+simplejson==3.17.5
+singledispatchmethod==1.0
+six==1.15.0
+smart-open==5.0.0
+smmap==5.0.0
+sniffio==1.3.0
+snowballstemmer==2.0.0
+sortedcollections==2.1.0
+sortedcontainers==2.4.0
+sounddevice==0.4.5
+SoundFile==0.10.3.post1
+soupsieve==2.3
+sox==1.4.1
+sparsemax==0.1.9
+speechbrain==0.5.14
+sphfile==1.0.3
+Sphinx==3.3.1
+sphinx-rtd-theme==0.2.4
+sphinxcontrib-applehelp==1.0.2
+sphinxcontrib-bibtex==2.4.1
+sphinxcontrib-devhelp==1.0.2
+sphinxcontrib-htmlhelp==1.0.3
+sphinxcontrib-jsmath==1.0.1
+sphinxcontrib-qthelp==1.0.3
+sphinxcontrib-serializinghtml==1.1.4
+SQLAlchemy==1.4.25
+sqlitedict==2.1.0
+sqlparse==0.4.2
+stanza==1.4.2
+starlette==0.27.0
+starsessions==1.3.0
+stevedore==3.4.0
+subprocess32==3.5.4
+sympy==1.9
+tabulate==0.8.9
+tensorboard==2.4.0
+tensorboard-plugin-wit==1.7.0
+tensorboardX==2.6.1
+tensorflow==2.4.0
+tensorflow-estimator==2.4.0
+termcolor==1.1.0
+terminado==0.9.4
+testpath==0.4.4
+threadpoolctl==2.1.0
+tifffile==2020.12.8
+tikzplotlib==0.9.8
+tinycss2==1.2.1
+tkseem==0.0.3
+tokenizers==0.13.3
+toml==0.10.2
+toolz==0.12.0
+torch==1.13.1
+torch-audiomentations==0.11.0
+torch-pitch-shift==1.2.4
+torch-stft==0.1.4
+torchaudio==0.13.1
+torchmetrics==0.11.4
+torchvision==0.14.1
+tornado==6.1
+tqdm==4.61.1
+trackrip==1.2.1
+traitlets==5.9.0
+transformer-smaller-training-vocab==0.3.1
+transformers==4.30.2
+triton==2.0.0
+typed-ast==1.4.1
+typer==0.4.0
+typing-extensions==4.4.0
+uc-micro-py==1.0.1
+Unidecode==1.3.2
+uritemplate==3.0.1
+urllib3==1.26.2
+uvicorn==0.20.0
+versioneer==0.28
+virtualenv==20.2.1
+wandb==0.12.6
+wcwidth==0.2.5
+webdataset==0.1.62
+webencodings==0.5.1
+websocket-client==1.6.1
+websockets==10.4
+Werkzeug==1.0.1
+wget==3.2
+widgetsnbextension==3.5.1
+Wikipedia-API==0.6.0
+wordninja==2.0.0
+wrapt==1.12.1
+xmltodict==0.13.0
+xxhash==2.0.0
+yamllint==1.23.0
+yarg==0.1.9
+yarl==1.7.2
+yaspin==2.1.0
+youtokentome==1.0.6
+youtube-dl==2021.6.6
+zipp==3.6.0
+
+
+2023-09-25 11:54:18,932 - speechbrain.utils.superpowers - DEBUG - 082323a
+
+
+2023-09-25 11:54:18,967 - speechbrain.dataio.encoder - DEBUG - Loaded categorical encoding from semi_wavlm_large_tunisian_ctc/1234/save/label_encoder.txt
+2023-09-25 11:54:18,967 - speechbrain.dataio.encoder - INFO - Load called, but CTCTextEncoder is not empty. Loaded data will overwrite everything. This is normal if there is e.g. an unk label defined at init.
+2023-09-25 11:54:18,967 - speechbrain.dataio.encoder - DEBUG - Loaded categorical encoding from semi_wavlm_large_tunisian_ctc/1234/save/label_encoder.txt
+2023-09-25 11:54:21,194 - pyctcdecode.decoder - INFO - Using arpa instead of binary LM file, decoder instantiation might be slow.
+2023-09-25 11:54:21,296 - pyctcdecode.alphabet - INFO - Alphabet determined to be of regular style.
+2023-09-25 11:54:21,358 - pyctcdecode.alphabet - WARNING - Unigrams and labels don't seem to agree.
+2023-09-25 11:54:22,645 - speechbrain.core - INFO - Info: auto_mix_prec arg from hparam file is used
+2023-09-25 11:54:22,645 - speechbrain.core - INFO - Info: ckpt_interval_minutes arg from hparam file is used
+2023-09-25 11:54:23,707 - speechbrain.core - INFO - 314.4M trainable parameters in ASR
+2023-09-25 11:54:23,708 - speechbrain.utils.checkpoints - INFO - Loading a checkpoint from semi_wavlm_large_tunisian_ctc/1234/save/CKPT+2023-09-05+01-09-23+00
+2023-09-25 11:54:24,309 - urllib3.connectionpool - DEBUG - Starting new HTTPS connection (1): checkip.amazonaws.com:443
+2023-09-25 11:54:24,311 - asyncio - DEBUG - Using selector: EpollSelector
+2023-09-25 11:54:24,312 - urllib3.connectionpool - DEBUG - Starting new HTTPS connection (1): api.gradio.app:443
+2023-09-25 11:54:24,340 - asyncio - DEBUG - Using selector: EpollSelector
+2023-09-25 11:54:24,347 - urllib3.connectionpool - DEBUG - Starting new HTTP connection (1): 127.0.0.1:7860
+2023-09-25 11:54:24,349 - urllib3.connectionpool - DEBUG - http://127.0.0.1:7860 "GET /startup-events HTTP/1.1" 200 5
+2023-09-25 11:54:24,376 - botocore.hooks - DEBUG - Changing event name from creating-client-class.iot-data to creating-client-class.iot-data-plane
+2023-09-25 11:54:24,377 - botocore.hooks - DEBUG - Changing event name from before-call.apigateway to before-call.api-gateway
+2023-09-25 11:54:24,377 - botocore.hooks - DEBUG - Changing event name from request-created.machinelearning.Predict to request-created.machine-learning.Predict
+2023-09-25 11:54:24,378 - botocore.hooks - DEBUG - Changing event name from before-parameter-build.autoscaling.CreateLaunchConfiguration to before-parameter-build.auto-scaling.CreateLaunchConfiguration
+2023-09-25 11:54:24,378 - botocore.hooks - DEBUG - Changing event name from before-parameter-build.route53 to before-parameter-build.route-53
+2023-09-25 11:54:24,379 - botocore.hooks - DEBUG - Changing event name from request-created.cloudsearchdomain.Search to request-created.cloudsearch-domain.Search
+2023-09-25 11:54:24,379 - botocore.hooks - DEBUG - Changing event name from docs.*.autoscaling.CreateLaunchConfiguration.complete-section to docs.*.auto-scaling.CreateLaunchConfiguration.complete-section
+2023-09-25 11:54:24,381 - botocore.hooks - DEBUG - Changing event name from before-parameter-build.logs.CreateExportTask to before-parameter-build.cloudwatch-logs.CreateExportTask
+2023-09-25 11:54:24,381 - botocore.hooks - DEBUG - Changing event name from docs.*.logs.CreateExportTask.complete-section to docs.*.cloudwatch-logs.CreateExportTask.complete-section
+2023-09-25 11:54:24,381 - botocore.hooks - DEBUG - Changing event name from before-parameter-build.cloudsearchdomain.Search to before-parameter-build.cloudsearch-domain.Search
+2023-09-25 11:54:24,381 - botocore.hooks - DEBUG - Changing event name from docs.*.cloudsearchdomain.Search.complete-section to docs.*.cloudsearch-domain.Search.complete-section
+2023-09-25 11:54:24,382 - botocore.utils - DEBUG - IMDS ENDPOINT: http://169.254.169.254/
+2023-09-25 11:54:24,383 - botocore.credentials - DEBUG - Looking for credentials via: env
+2023-09-25 11:54:24,383 - botocore.credentials - DEBUG - Looking for credentials via: assume-role
+2023-09-25 11:54:24,383 - botocore.credentials - DEBUG - Looking for credentials via: assume-role-with-web-identity
+2023-09-25 11:54:24,383 - botocore.credentials - DEBUG - Looking for credentials via: sso
+2023-09-25 11:54:24,383 - botocore.credentials - DEBUG - Looking for credentials via: shared-credentials-file
+2023-09-25 11:54:24,383 - botocore.credentials - DEBUG - Looking for credentials via: custom-process
+2023-09-25 11:54:24,383 - botocore.credentials - DEBUG - Looking for credentials via: config-file
+2023-09-25 11:54:24,383 - botocore.credentials - DEBUG - Looking for credentials via: ec2-credentials-file
+2023-09-25 11:54:24,383 - botocore.credentials - DEBUG - Looking for credentials via: boto-config
+2023-09-25 11:54:24,383 - botocore.credentials - DEBUG - Looking for credentials via: container-role
+2023-09-25 11:54:24,383 - botocore.credentials - DEBUG - Looking for credentials via: iam-role
+2023-09-25 11:54:24,383 - urllib3.connectionpool - DEBUG - Starting new HTTP connection (1): 169.254.169.254:80
+2023-09-25 11:54:24,733 - urllib3.connectionpool - DEBUG - https://checkip.amazonaws.com:443 "GET / HTTP/1.1" 200 14
+2023-09-25 11:54:24,740 - urllib3.connectionpool - DEBUG - Starting new HTTPS connection (1): api.gradio.app:443
+2023-09-25 11:54:25,385 - botocore.utils - DEBUG - Caught retryable HTTP exception while making metadata service request to http://169.254.169.254/latest/api/token: Connect timeout on endpoint URL: "http://169.254.169.254/latest/api/token"
+Traceback (most recent call last):
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connection.py", line 169, in _new_conn
+    conn = connection.create_connection(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/util/connection.py", line 96, in create_connection
+    raise err
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/util/connection.py", line 86, in create_connection
+    sock.connect(sa)
+socket.timeout: timed out
+
+During handling of the above exception, another exception occurred:
+
+Traceback (most recent call last):
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/httpsession.py", line 385, in send
+    urllib_response = conn.urlopen(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connectionpool.py", line 755, in urlopen
+    retries = retries.increment(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/util/retry.py", line 506, in increment
+    raise six.reraise(type(error), error, _stacktrace)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/packages/six.py", line 735, in reraise
+    raise value
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connectionpool.py", line 699, in urlopen
+    httplib_response = self._make_request(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connectionpool.py", line 394, in _make_request
+    conn.request(method, url, **httplib_request_kw)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connection.py", line 234, in request
+    super(HTTPConnection, self).request(method, url, body=body, headers=headers)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/http/client.py", line 1255, in request
+    self._send_request(method, url, body, headers, encode_chunked)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/awsrequest.py", line 92, in _send_request
+    rval = super(AWSConnection, self)._send_request(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/http/client.py", line 1301, in _send_request
+    self.endheaders(body, encode_chunked=encode_chunked)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/http/client.py", line 1250, in endheaders
+    self._send_output(message_body, encode_chunked=encode_chunked)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/awsrequest.py", line 120, in _send_output
+    self.send(msg)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/awsrequest.py", line 204, in send
+    return super(AWSConnection, self).send(str)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/http/client.py", line 950, in send
+    self.connect()
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connection.py", line 200, in connect
+    conn = self._new_conn()
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connection.py", line 174, in _new_conn
+    raise ConnectTimeoutError(
+urllib3.exceptions.ConnectTimeoutError: (<botocore.awsrequest.AWSHTTPConnection object at 0x7fccca8c8a00>, 'Connection to 169.254.169.254 timed out. (connect timeout=1)')
+
+During handling of the above exception, another exception occurred:
+
+Traceback (most recent call last):
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/utils.py", line 430, in _fetch_metadata_token
+    response = self._session.send(request.prepare())
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/httpsession.py", line 418, in send
+    raise ConnectTimeoutError(endpoint_url=request.url, error=e)
+botocore.exceptions.ConnectTimeoutError: Connect timeout on endpoint URL: "http://169.254.169.254/latest/api/token"
+2023-09-25 11:54:25,391 - urllib3.connectionpool - DEBUG - Starting new HTTP connection (2): 169.254.169.254:80
+2023-09-25 11:54:25,437 - urllib3.connectionpool - DEBUG - https://api.gradio.app:443 "GET /pkg-version HTTP/1.1" 200 21
+2023-09-25 11:54:25,998 - urllib3.connectionpool - DEBUG - https://api.gradio.app:443 "POST /gradio-initiated-analytics/ HTTP/1.1" 200 None
+2023-09-25 11:54:26,392 - botocore.utils - DEBUG - Caught retryable HTTP exception while making metadata service request to http://169.254.169.254/latest/meta-data/iam/security-credentials/: Connect timeout on endpoint URL: "http://169.254.169.254/latest/meta-data/iam/security-credentials/"
+Traceback (most recent call last):
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connection.py", line 169, in _new_conn
+    conn = connection.create_connection(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/util/connection.py", line 96, in create_connection
+    raise err
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/util/connection.py", line 86, in create_connection
+    sock.connect(sa)
+socket.timeout: timed out
+
+During handling of the above exception, another exception occurred:
+
+Traceback (most recent call last):
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/httpsession.py", line 385, in send
+    urllib_response = conn.urlopen(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connectionpool.py", line 755, in urlopen
+    retries = retries.increment(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/util/retry.py", line 506, in increment
+    raise six.reraise(type(error), error, _stacktrace)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/packages/six.py", line 735, in reraise
+    raise value
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connectionpool.py", line 699, in urlopen
+    httplib_response = self._make_request(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connectionpool.py", line 394, in _make_request
+    conn.request(method, url, **httplib_request_kw)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connection.py", line 234, in request
+    super(HTTPConnection, self).request(method, url, body=body, headers=headers)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/http/client.py", line 1255, in request
+    self._send_request(method, url, body, headers, encode_chunked)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/awsrequest.py", line 92, in _send_request
+    rval = super(AWSConnection, self)._send_request(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/http/client.py", line 1301, in _send_request
+    self.endheaders(body, encode_chunked=encode_chunked)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/http/client.py", line 1250, in endheaders
+    self._send_output(message_body, encode_chunked=encode_chunked)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/awsrequest.py", line 120, in _send_output
+    self.send(msg)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/awsrequest.py", line 204, in send
+    return super(AWSConnection, self).send(str)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/http/client.py", line 950, in send
+    self.connect()
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connection.py", line 200, in connect
+    conn = self._new_conn()
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connection.py", line 174, in _new_conn
+    raise ConnectTimeoutError(
+urllib3.exceptions.ConnectTimeoutError: (<botocore.awsrequest.AWSHTTPConnection object at 0x7fccca8d92e0>, 'Connection to 169.254.169.254 timed out. (connect timeout=1)')
+
+During handling of the above exception, another exception occurred:
+
+Traceback (most recent call last):
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/utils.py", line 478, in _get_request
+    response = self._session.send(request.prepare())
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/httpsession.py", line 418, in send
+    raise ConnectTimeoutError(endpoint_url=request.url, error=e)
+botocore.exceptions.ConnectTimeoutError: Connect timeout on endpoint URL: "http://169.254.169.254/latest/meta-data/iam/security-credentials/"
+2023-09-25 11:54:26,393 - botocore.utils - DEBUG - Max number of attempts exceeded (1) when attempting to retrieve data from metadata service.
+2023-09-25 11:54:26,394 - botocore.loaders - DEBUG - Loading JSON file: /home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/data/endpoints.json
+2023-09-25 11:54:26,421 - botocore.hooks - DEBUG - Event choose-service-name: calling handler <function handle_service_name_alias at 0x7fcce336fa60>
+2023-09-25 11:54:26,432 - botocore.loaders - DEBUG - Loading JSON file: /home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/data/sts/2011-06-15/service-2.json
+2023-09-25 11:54:26,433 - botocore.hooks - DEBUG - Event creating-client-class.sts: calling handler <function add_generate_presigned_url at 0x7fcce3280430>
+2023-09-25 11:54:26,435 - botocore.endpoint - DEBUG - Setting sts timeout as (60, 60)
+2023-09-25 11:54:26,436 - botocore.loaders - DEBUG - Loading JSON file: /home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/data/_retry.json
+2023-09-25 11:54:26,436 - botocore.client - DEBUG - Registering retry handlers for service: sts
+2023-09-25 11:54:26,437 - botocore.hooks - DEBUG - Event before-parameter-build.sts.GetCallerIdentity: calling handler <function generate_idempotent_uuid at 0x7fcce3268dc0>
+2023-09-25 11:54:26,437 - botocore.hooks - DEBUG - Event before-call.sts.GetCallerIdentity: calling handler <function inject_api_version_header_if_needed at 0x7fcce3271670>
+2023-09-25 11:54:26,437 - botocore.endpoint - DEBUG - Making request for OperationModel(name=GetCallerIdentity) with params: {'url_path': '/', 'query_string': '', 'method': 'POST', 'headers': {'Content-Type': 'application/x-www-form-urlencoded; charset=utf-8', 'User-Agent': 'Boto3/1.20.2 Python/3.8.5 Linux/5.4.0-150-generic Botocore/1.23.2'}, 'body': {'Action': 'GetCallerIdentity', 'Version': '2011-06-15'}, 'url': 'https://sts.amazonaws.com/', 'context': {'client_region': 'aws-global', 'client_config': <botocore.config.Config object at 0x7fcce353d490>, 'has_streaming_input': False, 'auth_type': None}}
+2023-09-25 11:54:26,437 - botocore.hooks - DEBUG - Event request-created.sts.GetCallerIdentity: calling handler <bound method RequestSigner.handler of <botocore.signers.RequestSigner object at 0x7fcce353d550>>
+2023-09-25 11:54:26,437 - botocore.hooks - DEBUG - Event choose-signer.sts.GetCallerIdentity: calling handler <function set_operation_specific_signer at 0x7fcce3268ca0>
+2023-09-25 11:54:26,439 - urllib3.connectionpool - DEBUG - Starting new HTTP connection (1): 127.0.0.1:7860
+2023-09-25 11:54:26,448 - urllib3.connectionpool - DEBUG - http://127.0.0.1:7860 "HEAD / HTTP/1.1" 200 0
+2023-09-25 11:54:26,449 - urllib3.connectionpool - DEBUG - Starting new HTTPS connection (1): api.gradio.app:443
+2023-09-25 11:54:27,539 - urllib3.connectionpool - DEBUG - https://api.gradio.app:443 "POST /gradio-launched-telemetry/ HTTP/1.1" 200 None
+2023-09-25 11:54:36,120 - matplotlib.pyplot - DEBUG - Loaded backend agg version unknown.
+2023-09-25 11:54:36,189 - pydub.converter - DEBUG - subprocess.call(['ffmpeg', '-y', '-i', '/tmp/gradio/a676790f6947ae75dde556fb6a864b2ddb801922/audio.wav', '-acodec', 'pcm_s32le', '-vn', '-f', 'wav', '-'])
+2023-09-25 11:54:36,267 - matplotlib.pyplot - DEBUG - Loaded backend TkAgg version unknown.
+2023-09-25 11:55:11,177 - speechbrain.core - INFO - Beginning experiment!
+2023-09-25 11:55:11,177 - speechbrain.core - INFO - Experiment folder: semi_wavlm_large_tunisian_ctc/1234
+2023-09-25 11:55:11,673 - speechbrain.utils.superpowers - DEBUG - abkhazia==1.0
+absl-py==0.11.0
+aiofiles==23.2.1
+aiohttp==3.8.0
+aiosignal==1.2.0
+alabaster==0.7.12
+alembic==1.7.4
+altair==4.2.0
+altgraph==0.17
+antlr4-python3-runtime==4.9.3
+anyio==3.6.2
+appdirs==1.4.4
+argcomplete==1.12.2
+argon2-cffi==20.1.0
+arrow==1.2.3
+asgiref==3.6.0
+asteroid-filterbanks==0.4.0
+astunparse==1.6.3
+async-generator==1.10
+async-timeout==4.0.0
+attrdict==2.0.1
+attrs==20.3.0
+audeer==1.16.0
+audformat==0.11.5
+audinterface==0.7.0
+audiofile==1.0.0
+audiomentations==0.25.0
+audioread==2.1.9
+audobject==0.4.14
+audresample==0.1.6
+-e git+https://github.com/facebookresearch/WavAugment.git@54afcdb00ccc852c2f030f239f8532c9562b550e#egg=augment
+autopage==0.4.0
+Babel==2.9.0
+backcall==0.2.0
+backports.cached-property==1.0.2
+beautifulsoup4==4.10.0
+black==19.10b0
+bleach==3.3.0
+blessed==1.20.0
+boto3==1.20.2
+botocore==1.23.2
+bpemb==0.3.4
+braceexpand==0.1.7
+cachetools==4.2.0
+certifi @ file:///croot/certifi_1671487769961/work/certifi
+cffi==1.14.3
+cfgv==3.2.0
+chardet==3.0.4
+charset-normalizer==2.0.7
+click==7.1.2
+cliff==3.9.0
+clldutils==3.5.4
+cloudpickle==2.2.1
+cmaes==0.8.2
+cmake==3.18.4.post1
+cmd2==2.2.0
+colorama==0.4.4
+colorlog==4.6.2
+configparser==5.1.0
+conllu==4.5.3
+croniter==1.3.15
+cryptography==38.0.4
+csrgraph==0.1.28
+csvw==1.8.1
+cycler==0.10.0
+Cython==0.29.21
+dataclasses==0.6
+dateutils==0.6.12
+decorator==4.4.2
+deepdiff==6.3.0
+deepspeech==0.9.1
+defusedxml==0.7.1
+Deprecated==1.2.14
+dill==0.3.3
+Distance==0.1.3
+distlib==0.3.1
+Django==3.2.16
+django-auditlog==2.2.1
+django-filter==22.1
+django-js-asset==1.2.2
+django-mptt==0.14.0
+djangorestframework==3.14.0
+docker-pycreds==0.4.0
+docopt==0.6.2
+docutils==0.16
+drf-excel==2.2.0
+drf-flex-fields==1.0.0
+drf-renderer-xlsx==0.4.1
+easyocr==1.2.1
+editdistance==0.6.0
+einops==0.3.2
+emoji==2.2.0
+entrypoints==0.3
+et-xmlfile==1.1.0
+exceptiongroup==1.1.0
+farasapy==0.0.14
+fastapi==0.98.0
+fastjsonschema==2.17.1
+fasttext==0.9.2
+ffmpeg-python==0.2.0
+ffmpy==0.3.0
+filelock==3.0.12
+flair==0.12.2
+flake8==3.7.9
+flatbuffers==1.12
+frozendict==2.0.7
+frozenlist==1.2.0
+fsspec==2021.11.0
+ftfy==6.1.1
+future==0.18.2
+g2p-en==2.1.0
+gast==0.3.3
+gdown==4.4.0
+gdrive==0.1.5
+gensim==4.0.1
+gitdb==4.0.9
+GitPython==3.1.24
+google-api-core==2.11.1
+google-api-python-client==2.43.0
+google-auth==1.24.0
+google-auth-httplib2==0.1.0
+google-auth-oauthlib==0.5.3
+google-pasta==0.2.0
+googleapis-common-protos==1.59.1
+gradio==3.44.4
+gradio-client==0.5.1
+greenlet==1.1.2
+grpcio==1.32.0
+h11==0.14.0
+h5features==1.3.2
+h5py==2.10.0
+hierarchy==0.4.0
+hmmlearn==0.2.8
+htk-io==0.5
+httpcore==0.16.3
+httplib2==0.22.0
+httpx==0.23.3
+huggingface-hub==0.15.1
+hydra-colorlog==0.1.4
+hydra-core==1.3.2
+hyperopt==0.2.7
+HyperPyYAML==1.1.0
+hypothesis==6.61.2
+identify==1.5.10
+idna==2.10
+imageio==2.9.0
+imagesize==1.2.0
+importlib-metadata==4.8.1
+importlib-resources==5.2.2
+inflect==5.3.0
+inquirer==3.1.3
+ipadic==1.0.0
+ipyevents==2.0.1
+ipykernel==5.3.4
+ipython==7.19.0
+ipython-genutils==0.2.0
+ipywebrtc==0.6.0
+ipywidgets==7.6.3
+iso-639==0.4.5
+isodate==0.6.0
+isort==4.3.21
+itsdangerous==2.1.2
+Janome==0.5.0
+jedi==0.17.2
+jeepney==0.8.0
+jieba==0.42.1
+Jinja2==3.0.3
+jiwer==2.2.0
+jmespath==0.10.0
+joblib==0.17.0
+jsonschema==3.2.0
+julius==0.2.7
+jupyter-client==6.1.7
+jupyter-core==4.7.0
+jupyterlab-pygments==0.1.2
+jupyterlab-widgets==1.0.0
+kaitaistruct==0.9
+kaldi-io==0.9.4
+kaldi-python-io==1.2.2
+kaldiio==2.17.2
+kenlm @ https://github.com/kpu/kenlm/archive/master.zip
+Keras-Preprocessing==1.1.2
+kiwisolver==1.3.1
+lang-trans==0.6.0
+langdetect==1.0.9
+latexcodec==2.0.1
+ldap3==2.9.1
+librosa==0.9.0
+lightning-cloud==0.5.37
+lightning-utilities==0.8.0
+linkify-it-py==1.0.3
+lit==16.0.6
+llvmlite==0.35.0
+lxml==4.9.0
+Mako==1.1.5
+Markdown==3.3.3
+markdown-it-py==3.0.0
+MarkupSafe==2.1.3
+marshmallow==3.14.0
+matplotlib==3.3.3
+mccabe==0.6.1
+mcd==0.4
+mdit-py-plugins==0.3.3
+mdurl==0.1.2
+mecab-python3==1.0.3
+megatron-lm==2.2.0
+metrics==0.3.3
+mido==1.2.10
+mistune==0.8.4
+more-itertools==8.6.0
+mpld3==0.3
+mpmath==1.2.1
+multidict==5.2.0
+multiprocess==0.70.11.1
+nbclient==0.5.3
+nbconvert==5.6.1
+nbformat==5.9.0
+NEMO==4.3.2
+nemo-toolkit==1.4.0
+nest-asyncio==1.5.1
+networkx==2.8.8
+nltk==3.2.4
+nodeenv==1.5.0
+normalize==2.0.2
+notebook==6.3.0
+numba==0.52.0
+numpy==1.19.4
+nvidia-cublas-cu11==11.10.3.66
+nvidia-cuda-cupti-cu11==11.7.101
+nvidia-cuda-nvrtc-cu11==11.7.99
+nvidia-cuda-runtime-cu11==11.7.99
+nvidia-cudnn-cu11==8.5.0.96
+nvidia-cufft-cu11==10.9.0.58
+nvidia-curand-cu11==10.2.10.91
+nvidia-cusolver-cu11==11.4.0.1
+nvidia-cusparse-cu11==11.7.4.91
+nvidia-nccl-cu11==2.14.3
+nvidia-nvtx-cu11==11.7.91
+oauthlib==3.1.0
+omegaconf==2.3.0
+onnx==1.10.2
+OpenCC==1.1.2
+opencv-python==4.4.0.46
+openpyxl==3.0.9
+opensmile==2.2.0
+opt-einsum==3.3.0
+optuna==2.10.0
+ordered-set==4.1.0
+orjson==3.8.4
+oyaml==1.0
+packaging==22.0
+pandas==1.2.5
+pandocfilters==1.4.3
+pangu==4.0.6.1
+parameterized==0.8.1
+parso==0.7.1
+pathlib2==2.3.7.post1
+pathspec==0.5.5
+pathtools==0.1.2
+pbr==5.6.0
+pefile==2019.4.18
+pescador==2.1.0
+pesq==0.0.3
+pexpect==4.8.0
+phonemizer==2.2.1
+pickleshare==0.7.5
+Pillow==9.3.0
+pip-api==0.0.23
+pipreqs==0.4.11
+pluggy==0.13.1
+pooch==1.3.0
+portalocker==2.3.2
+pptree==3.1
+pre-commit==2.9.0
+preprocessing==0.1.13
+pretty-midi==0.2.9
+prettytable==2.2.1
+primePy==1.3
+progressbar2==3.53.1
+prometheus-client==0.10.1
+promise==2.3
+prompt-toolkit==3.0.8
+protobuf==3.20.3
+psutil==5.6.6
+ptyprocess==0.6.0
+py==1.9.0
+py-espeak-ng==0.1.8
+py4j==0.10.9.7
+pyannote.audio==2.1.1
+pyannote.core==4.5
+pyannote.database==4.1.3
+pyannote.metrics==3.2.1
+pyannote.pipeline==2.3
+pyannotebook==0.1.0.dev0
+PyArabic==0.6.15
+pyarrow==3.0.0
+pyasn1==0.4.8
+pyasn1-modules==0.2.8
+pybind11==2.8.1
+pybtex==0.24.0
+pybtex-docutils==1.0.1
+pycodestyle==2.5.0
+pycparser==2.20
+pycryptodome==3.16.0
+pyctcdecode==0.4.0
+pydantic==1.10.4
+pyDeprecate==0.3.1
+pydub==0.25.1
+pyflakes==2.1.1
+Pygments==2.15.1
+pygtrie==2.5.0
+PyJWT==2.7.0
+pymodbus==2.5.3
+pyparsing==2.4.7
+pyperclip==1.8.2
+pypinyin==0.43.0
+pyrsistent==0.17.3
+pyserial==3.5
+PySocks==1.7.1
+pystoi==0.3.3
+pytest==5.4.1
+pytest-runner==5.3.1
+python-bidi==0.4.2
+python-crfsuite==0.9.7
+python-dateutil==2.8.2
+python-editor==1.0.4
+python-Levenshtein==0.12.2
+python-multipart==0.0.5
+python-utils==2.4.0
+pytorch-lightning==1.6.5
+pytorch-metric-learning==1.7.3
+pytorch-revgrad==0.2.0
+pytube==11.0.1
+pytz==2022.6
+PyWavelets==1.1.1
+PyYAML==6.0
+pyzmq==20.0.0
+rapidfuzz==1.8.2
+readchar==4.0.5
+regex==2020.11.13
+requests==2.28.1
+requests-oauthlib==1.3.0
+resampy==0.2.2
+rfc3986==1.4.0
+rich==13.4.2
+richenum==1.3.1
+rsa==4.7
+ruamel.yaml==0.17.21
+ruamel.yaml.clib==0.2.7
+s3m==1.1.0
+s3transfer==0.5.0
+sacrebleu==2.0.0
+sacremoses==0.0.44
+safetensors==0.3.1
+scikit-image==0.18.1
+scikit-learn==0.23.2
+scipy==1.5.4
+-e git+https://github.com/sanghack81/SDCIT@00d060dde733fde9345154a494f81e97fb395ca7#egg=SDCIT
+seaborn==0.11.1
+SecretStorage==3.3.3
+segments==2.1.3
+segtok==1.5.11
+semantic-version==2.10.0
+semver==2.13.0
+Send2Trash==1.5.0
+sentencepiece==0.1.99
+sentry-sdk==1.4.3
+shellingham==1.4.0
+shortuuid==1.0.7
+SIDEKIT==1.3.8.5.2
+simplejson==3.17.5
+singledispatchmethod==1.0
+six==1.15.0
+smart-open==5.0.0
+smmap==5.0.0
+sniffio==1.3.0
+snowballstemmer==2.0.0
+sortedcollections==2.1.0
+sortedcontainers==2.4.0
+sounddevice==0.4.5
+SoundFile==0.10.3.post1
+soupsieve==2.3
+sox==1.4.1
+sparsemax==0.1.9
+speechbrain==0.5.14
+sphfile==1.0.3
+Sphinx==3.3.1
+sphinx-rtd-theme==0.2.4
+sphinxcontrib-applehelp==1.0.2
+sphinxcontrib-bibtex==2.4.1
+sphinxcontrib-devhelp==1.0.2
+sphinxcontrib-htmlhelp==1.0.3
+sphinxcontrib-jsmath==1.0.1
+sphinxcontrib-qthelp==1.0.3
+sphinxcontrib-serializinghtml==1.1.4
+SQLAlchemy==1.4.25
+sqlitedict==2.1.0
+sqlparse==0.4.2
+stanza==1.4.2
+starlette==0.27.0
+starsessions==1.3.0
+stevedore==3.4.0
+subprocess32==3.5.4
+sympy==1.9
+tabulate==0.8.9
+tensorboard==2.4.0
+tensorboard-plugin-wit==1.7.0
+tensorboardX==2.6.1
+tensorflow==2.4.0
+tensorflow-estimator==2.4.0
+termcolor==1.1.0
+terminado==0.9.4
+testpath==0.4.4
+threadpoolctl==2.1.0
+tifffile==2020.12.8
+tikzplotlib==0.9.8
+tinycss2==1.2.1
+tkseem==0.0.3
+tokenizers==0.13.3
+toml==0.10.2
+toolz==0.12.0
+torch==1.13.1
+torch-audiomentations==0.11.0
+torch-pitch-shift==1.2.4
+torch-stft==0.1.4
+torchaudio==0.13.1
+torchmetrics==0.11.4
+torchvision==0.14.1
+tornado==6.1
+tqdm==4.61.1
+trackrip==1.2.1
+traitlets==5.9.0
+transformer-smaller-training-vocab==0.3.1
+transformers==4.30.2
+triton==2.0.0
+typed-ast==1.4.1
+typer==0.4.0
+typing-extensions==4.4.0
+uc-micro-py==1.0.1
+Unidecode==1.3.2
+uritemplate==3.0.1
+urllib3==1.26.2
+uvicorn==0.20.0
+versioneer==0.28
+virtualenv==20.2.1
+wandb==0.12.6
+wcwidth==0.2.5
+webdataset==0.1.62
+webencodings==0.5.1
+websocket-client==1.6.1
+websockets==10.4
+Werkzeug==1.0.1
+wget==3.2
+widgetsnbextension==3.5.1
+Wikipedia-API==0.6.0
+wordninja==2.0.0
+wrapt==1.12.1
+xmltodict==0.13.0
+xxhash==2.0.0
+yamllint==1.23.0
+yarg==0.1.9
+yarl==1.7.2
+yaspin==2.1.0
+youtokentome==1.0.6
+youtube-dl==2021.6.6
+zipp==3.6.0
+
+
+2023-09-25 11:55:11,703 - speechbrain.utils.superpowers - DEBUG - 082323a
+
+
+2023-09-25 11:55:11,733 - speechbrain.dataio.encoder - DEBUG - Loaded categorical encoding from semi_wavlm_large_tunisian_ctc/1234/save/label_encoder.txt
+2023-09-25 11:55:11,733 - speechbrain.dataio.encoder - INFO - Load called, but CTCTextEncoder is not empty. Loaded data will overwrite everything. This is normal if there is e.g. an unk label defined at init.
+2023-09-25 11:55:11,734 - speechbrain.dataio.encoder - DEBUG - Loaded categorical encoding from semi_wavlm_large_tunisian_ctc/1234/save/label_encoder.txt
+2023-09-25 11:55:14,066 - pyctcdecode.decoder - INFO - Using arpa instead of binary LM file, decoder instantiation might be slow.
+2023-09-25 11:55:14,171 - pyctcdecode.alphabet - INFO - Alphabet determined to be of regular style.
+2023-09-25 11:55:14,236 - pyctcdecode.alphabet - WARNING - Unigrams and labels don't seem to agree.
+2023-09-25 11:55:15,539 - speechbrain.core - INFO - Info: auto_mix_prec arg from hparam file is used
+2023-09-25 11:55:15,539 - speechbrain.core - INFO - Info: ckpt_interval_minutes arg from hparam file is used
+2023-09-25 11:55:15,543 - speechbrain.core - INFO - 314.4M trainable parameters in ASR
+2023-09-25 11:55:15,543 - speechbrain.utils.checkpoints - INFO - Loading a checkpoint from semi_wavlm_large_tunisian_ctc/1234/save/CKPT+2023-09-05+01-09-23+00
+2023-09-25 11:55:16,121 - urllib3.connectionpool - DEBUG - Starting new HTTPS connection (1): api.gradio.app:443
+2023-09-25 11:55:16,121 - urllib3.connectionpool - DEBUG - Starting new HTTPS connection (1): checkip.amazonaws.com:443
+2023-09-25 11:55:16,122 - asyncio - DEBUG - Using selector: EpollSelector
+2023-09-25 11:55:16,149 - asyncio - DEBUG - Using selector: EpollSelector
+2023-09-25 11:55:16,157 - urllib3.connectionpool - DEBUG - Starting new HTTP connection (1): 127.0.0.1:7860
+2023-09-25 11:55:16,158 - urllib3.connectionpool - DEBUG - http://127.0.0.1:7860 "GET /startup-events HTTP/1.1" 200 5
+2023-09-25 11:55:16,185 - botocore.hooks - DEBUG - Changing event name from creating-client-class.iot-data to creating-client-class.iot-data-plane
+2023-09-25 11:55:16,186 - botocore.hooks - DEBUG - Changing event name from before-call.apigateway to before-call.api-gateway
+2023-09-25 11:55:16,186 - botocore.hooks - DEBUG - Changing event name from request-created.machinelearning.Predict to request-created.machine-learning.Predict
+2023-09-25 11:55:16,187 - botocore.hooks - DEBUG - Changing event name from before-parameter-build.autoscaling.CreateLaunchConfiguration to before-parameter-build.auto-scaling.CreateLaunchConfiguration
+2023-09-25 11:55:16,187 - botocore.hooks - DEBUG - Changing event name from before-parameter-build.route53 to before-parameter-build.route-53
+2023-09-25 11:55:16,187 - botocore.hooks - DEBUG - Changing event name from request-created.cloudsearchdomain.Search to request-created.cloudsearch-domain.Search
+2023-09-25 11:55:16,188 - botocore.hooks - DEBUG - Changing event name from docs.*.autoscaling.CreateLaunchConfiguration.complete-section to docs.*.auto-scaling.CreateLaunchConfiguration.complete-section
+2023-09-25 11:55:16,189 - botocore.hooks - DEBUG - Changing event name from before-parameter-build.logs.CreateExportTask to before-parameter-build.cloudwatch-logs.CreateExportTask
+2023-09-25 11:55:16,189 - botocore.hooks - DEBUG - Changing event name from docs.*.logs.CreateExportTask.complete-section to docs.*.cloudwatch-logs.CreateExportTask.complete-section
+2023-09-25 11:55:16,189 - botocore.hooks - DEBUG - Changing event name from before-parameter-build.cloudsearchdomain.Search to before-parameter-build.cloudsearch-domain.Search
+2023-09-25 11:55:16,189 - botocore.hooks - DEBUG - Changing event name from docs.*.cloudsearchdomain.Search.complete-section to docs.*.cloudsearch-domain.Search.complete-section
+2023-09-25 11:55:16,190 - botocore.utils - DEBUG - IMDS ENDPOINT: http://169.254.169.254/
+2023-09-25 11:55:16,191 - botocore.credentials - DEBUG - Looking for credentials via: env
+2023-09-25 11:55:16,191 - botocore.credentials - DEBUG - Looking for credentials via: assume-role
+2023-09-25 11:55:16,192 - botocore.credentials - DEBUG - Looking for credentials via: assume-role-with-web-identity
+2023-09-25 11:55:16,192 - botocore.credentials - DEBUG - Looking for credentials via: sso
+2023-09-25 11:55:16,192 - botocore.credentials - DEBUG - Looking for credentials via: shared-credentials-file
+2023-09-25 11:55:16,192 - botocore.credentials - DEBUG - Looking for credentials via: custom-process
+2023-09-25 11:55:16,192 - botocore.credentials - DEBUG - Looking for credentials via: config-file
+2023-09-25 11:55:16,192 - botocore.credentials - DEBUG - Looking for credentials via: ec2-credentials-file
+2023-09-25 11:55:16,192 - botocore.credentials - DEBUG - Looking for credentials via: boto-config
+2023-09-25 11:55:16,192 - botocore.credentials - DEBUG - Looking for credentials via: container-role
+2023-09-25 11:55:16,192 - botocore.credentials - DEBUG - Looking for credentials via: iam-role
+2023-09-25 11:55:16,192 - urllib3.connectionpool - DEBUG - Starting new HTTP connection (1): 169.254.169.254:80
+2023-09-25 11:55:16,411 - urllib3.connectionpool - DEBUG - https://checkip.amazonaws.com:443 "GET / HTTP/1.1" 200 14
+2023-09-25 11:55:16,418 - urllib3.connectionpool - DEBUG - Starting new HTTPS connection (1): api.gradio.app:443
+2023-09-25 11:55:17,100 - urllib3.connectionpool - DEBUG - https://api.gradio.app:443 "GET /pkg-version HTTP/1.1" 200 21
+2023-09-25 11:55:17,194 - botocore.utils - DEBUG - Caught retryable HTTP exception while making metadata service request to http://169.254.169.254/latest/api/token: Connect timeout on endpoint URL: "http://169.254.169.254/latest/api/token"
+Traceback (most recent call last):
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connection.py", line 169, in _new_conn
+    conn = connection.create_connection(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/util/connection.py", line 96, in create_connection
+    raise err
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/util/connection.py", line 86, in create_connection
+    sock.connect(sa)
+socket.timeout: timed out
+
+During handling of the above exception, another exception occurred:
+
+Traceback (most recent call last):
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/httpsession.py", line 385, in send
+    urllib_response = conn.urlopen(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connectionpool.py", line 755, in urlopen
+    retries = retries.increment(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/util/retry.py", line 506, in increment
+    raise six.reraise(type(error), error, _stacktrace)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/packages/six.py", line 735, in reraise
+    raise value
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connectionpool.py", line 699, in urlopen
+    httplib_response = self._make_request(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connectionpool.py", line 394, in _make_request
+    conn.request(method, url, **httplib_request_kw)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connection.py", line 234, in request
+    super(HTTPConnection, self).request(method, url, body=body, headers=headers)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/http/client.py", line 1255, in request
+    self._send_request(method, url, body, headers, encode_chunked)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/awsrequest.py", line 92, in _send_request
+    rval = super(AWSConnection, self)._send_request(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/http/client.py", line 1301, in _send_request
+    self.endheaders(body, encode_chunked=encode_chunked)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/http/client.py", line 1250, in endheaders
+    self._send_output(message_body, encode_chunked=encode_chunked)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/awsrequest.py", line 120, in _send_output
+    self.send(msg)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/awsrequest.py", line 204, in send
+    return super(AWSConnection, self).send(str)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/http/client.py", line 950, in send
+    self.connect()
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connection.py", line 200, in connect
+    conn = self._new_conn()
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connection.py", line 174, in _new_conn
+    raise ConnectTimeoutError(
+urllib3.exceptions.ConnectTimeoutError: (<botocore.awsrequest.AWSHTTPConnection object at 0x7fecf86e99a0>, 'Connection to 169.254.169.254 timed out. (connect timeout=1)')
+
+During handling of the above exception, another exception occurred:
+
+Traceback (most recent call last):
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/utils.py", line 430, in _fetch_metadata_token
+    response = self._session.send(request.prepare())
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/httpsession.py", line 418, in send
+    raise ConnectTimeoutError(endpoint_url=request.url, error=e)
+botocore.exceptions.ConnectTimeoutError: Connect timeout on endpoint URL: "http://169.254.169.254/latest/api/token"
+2023-09-25 11:55:17,200 - urllib3.connectionpool - DEBUG - Starting new HTTP connection (2): 169.254.169.254:80
+2023-09-25 11:55:17,412 - urllib3.connectionpool - DEBUG - https://api.gradio.app:443 "POST /gradio-initiated-analytics/ HTTP/1.1" 200 None
+2023-09-25 11:55:18,202 - botocore.utils - DEBUG - Caught retryable HTTP exception while making metadata service request to http://169.254.169.254/latest/meta-data/iam/security-credentials/: Connect timeout on endpoint URL: "http://169.254.169.254/latest/meta-data/iam/security-credentials/"
+Traceback (most recent call last):
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connection.py", line 169, in _new_conn
+    conn = connection.create_connection(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/util/connection.py", line 96, in create_connection
+    raise err
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/util/connection.py", line 86, in create_connection
+    sock.connect(sa)
+socket.timeout: timed out
+
+During handling of the above exception, another exception occurred:
+
+Traceback (most recent call last):
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/httpsession.py", line 385, in send
+    urllib_response = conn.urlopen(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connectionpool.py", line 755, in urlopen
+    retries = retries.increment(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/util/retry.py", line 506, in increment
+    raise six.reraise(type(error), error, _stacktrace)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/packages/six.py", line 735, in reraise
+    raise value
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connectionpool.py", line 699, in urlopen
+    httplib_response = self._make_request(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connectionpool.py", line 394, in _make_request
+    conn.request(method, url, **httplib_request_kw)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connection.py", line 234, in request
+    super(HTTPConnection, self).request(method, url, body=body, headers=headers)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/http/client.py", line 1255, in request
+    self._send_request(method, url, body, headers, encode_chunked)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/awsrequest.py", line 92, in _send_request
+    rval = super(AWSConnection, self)._send_request(
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/http/client.py", line 1301, in _send_request
+    self.endheaders(body, encode_chunked=encode_chunked)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/http/client.py", line 1250, in endheaders
+    self._send_output(message_body, encode_chunked=encode_chunked)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/awsrequest.py", line 120, in _send_output
+    self.send(msg)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/awsrequest.py", line 204, in send
+    return super(AWSConnection, self).send(str)
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/http/client.py", line 950, in send
+    self.connect()
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connection.py", line 200, in connect
+    conn = self._new_conn()
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/urllib3/connection.py", line 174, in _new_conn
+    raise ConnectTimeoutError(
+urllib3.exceptions.ConnectTimeoutError: (<botocore.awsrequest.AWSHTTPConnection object at 0x7fecf8a9e790>, 'Connection to 169.254.169.254 timed out. (connect timeout=1)')
+
+During handling of the above exception, another exception occurred:
+
+Traceback (most recent call last):
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/utils.py", line 478, in _get_request
+    response = self._session.send(request.prepare())
+  File "/home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/httpsession.py", line 418, in send
+    raise ConnectTimeoutError(endpoint_url=request.url, error=e)
+botocore.exceptions.ConnectTimeoutError: Connect timeout on endpoint URL: "http://169.254.169.254/latest/meta-data/iam/security-credentials/"
+2023-09-25 11:55:18,203 - botocore.utils - DEBUG - Max number of attempts exceeded (1) when attempting to retrieve data from metadata service.
+2023-09-25 11:55:18,203 - botocore.loaders - DEBUG - Loading JSON file: /home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/data/endpoints.json
+2023-09-25 11:55:18,212 - botocore.hooks - DEBUG - Event choose-service-name: calling handler <function handle_service_name_alias at 0x7fecf895f820>
+2023-09-25 11:55:18,216 - botocore.loaders - DEBUG - Loading JSON file: /home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/data/sts/2011-06-15/service-2.json
+2023-09-25 11:55:18,217 - botocore.hooks - DEBUG - Event creating-client-class.sts: calling handler <function add_generate_presigned_url at 0x7fecf87df430>
+2023-09-25 11:55:18,218 - botocore.endpoint - DEBUG - Setting sts timeout as (60, 60)
+2023-09-25 11:55:18,218 - botocore.loaders - DEBUG - Loading JSON file: /home/salah/anaconda3/envs/salah/lib/python3.8/site-packages/botocore/data/_retry.json
+2023-09-25 11:55:18,218 - botocore.client - DEBUG - Registering retry handlers for service: sts
+2023-09-25 11:55:18,218 - botocore.hooks - DEBUG - Event before-parameter-build.sts.GetCallerIdentity: calling handler <function generate_idempotent_uuid at 0x7fecf87c7e50>
+2023-09-25 11:55:18,219 - botocore.hooks - DEBUG - Event before-call.sts.GetCallerIdentity: calling handler <function inject_api_version_header_if_needed at 0x7fecf87d0700>
+2023-09-25 11:55:18,219 - botocore.endpoint - DEBUG - Making request for OperationModel(name=GetCallerIdentity) with params: {'url_path': '/', 'query_string': '', 'method': 'POST', 'headers': {'Content-Type': 'application/x-www-form-urlencoded; charset=utf-8', 'User-Agent': 'Boto3/1.20.2 Python/3.8.5 Linux/5.4.0-150-generic Botocore/1.23.2'}, 'body': {'Action': 'GetCallerIdentity', 'Version': '2011-06-15'}, 'url': 'https://sts.amazonaws.com/', 'context': {'client_region': 'aws-global', 'client_config': <botocore.config.Config object at 0x7fecf8b0fc40>, 'has_streaming_input': False, 'auth_type': None}}
+2023-09-25 11:55:18,219 - botocore.hooks - DEBUG - Event request-created.sts.GetCallerIdentity: calling handler <bound method RequestSigner.handler of <botocore.signers.RequestSigner object at 0x7fecf8b0fe50>>
+2023-09-25 11:55:18,219 - botocore.hooks - DEBUG - Event choose-signer.sts.GetCallerIdentity: calling handler <function set_operation_specific_signer at 0x7fecf87c7d30>
+2023-09-25 11:55:18,220 - urllib3.connectionpool - DEBUG - Starting new HTTP connection (1): 127.0.0.1:7860
+2023-09-25 11:55:18,226 - urllib3.connectionpool - DEBUG - http://127.0.0.1:7860 "HEAD / HTTP/1.1" 200 0
+2023-09-25 11:55:18,228 - urllib3.connectionpool - DEBUG - Starting new HTTPS connection (1): api.gradio.app:443
+2023-09-25 11:55:19,252 - urllib3.connectionpool - DEBUG - https://api.gradio.app:443 "POST /gradio-launched-telemetry/ HTTP/1.1" 200 None
+2023-09-25 11:55:26,648 - matplotlib.pyplot - DEBUG - Loaded backend agg version unknown.
+2023-09-25 11:55:26,688 - pydub.converter - DEBUG - subprocess.call(['ffmpeg', '-y', '-i', '/tmp/gradio/8c62f8e9d42a7333da0ae55f015d665ae74f7913/audio.wav', '-acodec', 'pcm_s32le', '-vn', '-f', 'wav', '-'])
+2023-09-25 11:55:26,941 - matplotlib.pyplot - DEBUG - Loaded backend TkAgg version unknown.
+2023-09-25 11:55:34,439 - matplotlib.pyplot - DEBUG - Loaded backend agg version unknown.
+2023-09-25 11:55:34,479 - pydub.converter - DEBUG - subprocess.call(['ffmpeg', '-y', '-i', '/tmp/gradio/a0134345072496f4b50c5c2ed175ba6f070ef6f0/audio.wav', '-acodec', 'pcm_s32le', '-vn', '-f', 'wav', '-'])
+2023-09-25 11:55:34,955 - matplotlib.pyplot - DEBUG - Loaded backend TkAgg version unknown.
+2023-09-25 11:55:41,408 - matplotlib.pyplot - DEBUG - Loaded backend agg version unknown.
+2023-09-25 11:55:41,456 - pydub.converter - DEBUG - subprocess.call(['ffmpeg', '-y', '-i', '/tmp/gradio/31eca7d10f9506b4380f815f4bc19a04caed0f0f/audio.wav', '-acodec', 'pcm_s32le', '-vn', '-f', 'wav', '-'])
+2023-09-25 11:55:41,780 - matplotlib.pyplot - DEBUG - Loaded backend TkAgg version unknown.

semi_wavlm_large_tunisian_ctc/1234/save/CKPT+2023-09-05+01-09-23+00/CKPT.yaml

@@ -0,0 +1,4 @@
+# yamllint disable
+WER: 27.83210816487267
+end-of-epoch: true
+unixtime: 1693868963.5220973

semi_wavlm_large_tunisian_ctc/1234/save/CKPT+2023-09-05+01-09-23+00/brain.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3947a24e8dff5a14299b9cf2fe66ffb4d738cb88717de7f0cf7e8547a76e9776
+size 51

semi_wavlm_large_tunisian_ctc/1234/save/CKPT+2023-09-05+01-09-23+00/counter.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b51d431df5d7f141cbececcf79edf3dd861c3b4069f0b11661a3eefacbba918
+size 2

semi_wavlm_large_tunisian_ctc/1234/save/CKPT+2023-09-05+01-09-23+00/dataloader-TRAIN.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b363886c229e536bd3c84e0c3e89312d70e00422578e076a62df1b45c9390793
+size 5

semi_wavlm_large_tunisian_ctc/1234/save/CKPT+2023-09-05+01-09-23+00/model.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc1dbeca1e1f1340b08d8ebea6e492f474708dddbbe8cabbcdde5ee9660704f2
+size 12814446

semi_wavlm_large_tunisian_ctc/1234/save/CKPT+2023-09-05+01-09-23+00/modelopt.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3af1791eb9a5bfbfc087d2c10b94634df24cad3ac503ce9ba280a3ecc4737781
+size 25575663

semi_wavlm_large_tunisian_ctc/1234/save/CKPT+2023-09-05+01-09-23+00/scheduler_model.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c275ab9245b440d1586f72058d9edaac1a2fb3e7a52712aa9a9ad022b99a1c0d
+size 639

semi_wavlm_large_tunisian_ctc/1234/save/CKPT+2023-09-05+01-09-23+00/scheduler_wav2vec.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a88187f7882dc3e10c108f1b7abfbd819285b34bded4e88e91c4ff699c1bb5d2
+size 643

semi_wavlm_large_tunisian_ctc/1234/save/CKPT+2023-09-05+01-09-23+00/wav2vec2.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:788267bd25ef37623715fa21a975090e5e316fff05971375cd3f62e5160f0743
+size 1262005979

semi_wavlm_large_tunisian_ctc/1234/save/CKPT+2023-09-05+01-09-23+00/wav2vec_opt.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:efa967fdd8067be7d88c18cd197980c9c91f344a3dff2b2518b8381c49f28b1e
+size 2490361859

semi_wavlm_large_tunisian_ctc/1234/save/label_encoder.txt

@@ -0,0 +1,44 @@
+'ب' => 38
+'ا' => 1
+'ه' => 2
+'ي' => 3
+'و' => 4
+'ن' => 5
+'أ' => 6
+' ' => 7
+'م' => 8
+'ش' => 9
+'ل' => 10
+'س' => 11
+'ت' => 12
+'د' => 13
+'ر' => 14
+'ى' => 15
+'ح' => 16
+'ط' => 17
+'ع' => 18
+'ك' => 19
+'ف' => 20
+'ق' => 21
+'آ' => 22
+'ة' => 23
+'ج' => 24
+'ض' => 25
+'ز' => 26
+'ص' => 27
+'إ' => 28
+'ث' => 29
+'خ' => 30
+'ڨ' => 31
+'ذ' => 32
+'ظ' => 33
+'ء' => 34
+'غ' => 35
+'ئ' => 36
+'ؤ' => 37
+'<blank>' => 0
+1 => 39
+================
+'starting_index' => 0
+'unk_label' => 1
+'blank_label' => '<blank>'

taric_test.csv

@@ -0,0 +1,381 @@
+Unnamed: 0,ID,wav,wrd,duration
+270681,audio4_1h_transcrit_360.719_361.907,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_360.719_361.907.wav,ڨابس وقتاش,1.188
+275905,audio7_test_transcrit_3968.362_3971.002,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_3968.362_3971.002.wav,الساعتين مع الستة الستة و درجين,2.64
+270058,audio7_test_transcrit_3694.359_3697.671,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_3694.359_3697.671.wav,أيوه يوصل العشرة العشرة غير ربع العشرة غير درجين,3.312
+271296,audio5_Transcrit_2485.137_2487.215,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_2485.137_2487.215.wav,الثلاثة غير درج موش الأربعة و نصف,2.0780000000000003
+274681,audio1_2hTranscrit_2666.861_2667.791,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_2666.861_2667.791.wav,بالله,0.93
+273082,audio1_2hTranscrit_3182.755_3184.453,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_3182.755_3184.453.wav,باهي أعطيني تونس,1.6980000000000002
+267991,audio7_test_transcrit_4503.851_4505.820,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_4503.851_4505.820.wav,السلام عليكم ڨابس كنفور,1.969
+262671,audio10_1752.729_1755.454,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio10_1752.729_1755.454.wav,وقتاه هو التران يخرج,2.725
+272639,audio8-1h30.tst_1371.833_1373.792,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_1371.833_1373.792.wav,أي آهي البطاقة متاعي,1.959
+270677,audio12_2881.038_2882.838,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_2881.038_2882.838.wav,ساعتين و نصف إنتي تحب,1.8
+276090,audio1_2hTranscrit_95.063_96.160,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_95.063_96.160.wav,سلام عليكم,1.097
+278161,audio12_3594.075_3595.275,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_3594.075_3595.275.wav,تونس,1.2
+269161,audio17_1h_1591.741_1593.705,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_1591.741_1593.705.wav,وين ماشية تكلم هنا هنا,1.964
+277093,audio3_transcrit_1067.879_1069.395,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_1067.879_1069.395.wav,و إلا نصف الليل و درجين,1.516
+268969,audio8-1h30.tst_5033.163_5036.389,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_5033.163_5036.389.wav,ثلاثة ألاي سامبل و إلا ثمة لألاي رتور,3.2260000000000004
+271109,audio11_Transcrit_2899.443_2901.130,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_2899.443_2901.130.wav,آكسبراس الأربعة و نصف,1.6869999999999998
+10415,audio5_Transcrit_423.305_430.317,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_423.305_430.317.wav,و الله ما نعرفش توا كي نقرب إنق توا إذا آني غادي توا نقلك,7.0120000000000005
+276418,audio14-1h_1087.372_1089.260,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_1087.372_1089.260.wav,الساعتين و ربع أي تفضل مدام,1.8880000000000001
+275484,audio8-1h30.tst_4388.209_4391.520,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_4388.209_4391.520.wav,حاشتي بثلاثة بلايص لبير بورڨبة,3.3110000000000004
+269977,audio12_3352.756_3357.172,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_3352.756_3357.172.wav,أعطيني ألفين و خمسة مية أعطيني نحب ألفين و خمسة مية,4.416
+273952,audio1_2hTranscrit_5535.218_5536.281,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_5535.218_5536.281.wav,ثلاثة غير درج,1.063
+266742,audio5_Transcrit_1174.151_1174.891,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_1174.151_1174.891.wav,توا,0.74
+15939,audio11_Transcrit_1892.293_1897.957,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_1892.293_1897.957.wav,بالله القلعة يعيشك بروميار قول هو كنفور بكلو,5.664
+271436,audio17_1h_376.916_377.956,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_376.916_377.956.wav,سرس,1.04
+267117,audio10_1234.611_1237.386,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio10_1234.611_1237.386.wav,باهي أختي بقداه التكاي,2.775
+266559,audio14-1h_2584.370_2589.563,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_2584.370_2589.563.wav,خمسة آلاف و مية و خمسين أعطيني مية و خمسين فرنك عندكشي خمسة مية,5.193
+273290,audio1_2hTranscrit_882.335_883.710,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_882.335_883.710.wav,نابل بير بورڨبة,1.375
+266645,audio1_2hTranscrit_5063.235_5065.076,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_5063.235_5065.076.wav,سلام باللاهي تونس,1.841
+272054,audio3_transcrit_714.997_715.765,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_714.997_715.765.wav,أي ب,0.768
+270920,audio12_2740.304_2741.264,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_2740.304_2741.264.wav,تفضل,0.96
+270263,audio5_Transcrit_1830.538_1832.065,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_1830.538_1832.065.wav,أعطيني ألفين و مية,1.527
+269434,audio12_842.457_847.441,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_842.457_847.441.wav,إليوم باريود روج تتسمى خاطر ويكاند دانك بسبعطاش دينار,4.984
+262602,audio4_1h_transcrit_3315.769_3316.914,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_3315.769_3316.914.wav,ثلاثة غير درج,1.145
+276898,audio12_1499.442_1503.066,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_1499.442_1503.066.wav,عندكشي مية و خمسين عيش ولدي,3.6239999999999997
+261117,audio1_2hTranscrit_5260.824_5262.199,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_5260.824_5262.199.wav,ثمة تران لتونس,1.375
+266521,audio10_3322.288_3324.667,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio10_3322.288_3324.667.wav,عسلامة باللاهي تكاي للمهدية,2.379
+274371,audio3_transcrit_2182.048_2183.364,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_2182.048_2183.364.wav,كارت جون ما عندكش,1.3159999999999998
+270951,audio1_2hTranscrit_4864.609_4865.997,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_4864.609_4865.997.wav,تران وقتاش يخرج,1.3880000000000001
+265283,audio1_2hTranscrit_489.928_491.030,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_489.928_491.030.wav,غدوة الصباح,1.102
+268083,audio7_test_transcrit_3774.600_3775.177,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_3774.600_3775.177.wav,ياقف,0.5770000000000001
+261488,audio6.tst_495.711_496.789,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio6.tst_495.711_496.789.wav,ألاي و رتور,1.078
+275219,audio6.tst_451.267_452.974,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio6.tst_451.267_452.974.wav,إسمع التران أماهو إلي نركب فيه,1.7069999999999999
+261387,audio1_2hTranscrit_1558.619_1561.075,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_1558.619_1561.075.wav,ماضي ثلاثة أربعة سوايع معناها,2.456
+263997,audio1_2hTranscrit_3990.908_3991.843,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_3990.908_3991.843.wav,يا ولدي,0.935
+268313,audio17_1h_719.875_720.580,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_719.875_720.580.wav,سلام,0.705
+264510,audio1_2hTranscrit_5682.582_5683.551,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_5682.582_5683.551.wav,و الساعتين,0.9690000000000001
+274380,audio1_2hTranscrit_2849.132_2850.866,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_2849.132_2850.866.wav,ثلاثة غير درج آكسبراس,1.734
+263438,audio1_2hTranscrit_752.713_754.220,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_752.713_754.220.wav,خرج الحداش و درجين,1.507
+276901,audio11_Transcrit_995.823_996.761,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_995.823_996.761.wav,لا لا,0.938
+274005,audio5_Transcrit_3757.520_3759.929,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_3757.520_3759.929.wav,بالله وريقة متاع الأوقات متاع التران,2.4090000000000003
+271846,audio6.tst_3293.175_3295.577,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio6.tst_3293.175_3295.577.wav,سوسة الأربعة و أربعة آه الأربعة غير أربعة,2.4019999999999997
+271819,audio6.tst_3404.361_3407.239,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio6.tst_3404.361_3407.239.wav,عشرة و ثلاثة مية دوزيام كلاس أما هاذا بروميار كلاس,2.878
+267595,audio8-1h30.tst_2409.161_2411.080,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_2409.161_2411.080.wav,كل نهار هاذوم موجودين,1.919
+15792,audio1_2hTranscrit_1844.951_1845.619,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_1844.951_1845.619.wav,نعم,0.6679999999999999
+267084,audio14-1h_2532.547_2533.797,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_2532.547_2533.797.wav,أعطيني ألفين و خمسة مية,1.25
+267000,audio7_test_transcrit_779.182_780.854,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_779.182_780.854.wav,حداش ألف,1.672
+261576,audio7_test_transcrit_150.267_151.533,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_150.267_151.533.wav,سوسة بروميار,1.266
+272817,audio1_2hTranscrit_3170.793_3172.465,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_3170.793_3172.465.wav,مازلنا نستحقوها,1.672
+265865,audio14-1h_84.796_86.662,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_84.796_86.662.wav,لازمك تزيد أربعة آلاف و خمسة مية,1.8659999999999999
+263750,audio11_Transcrit_2013.200_2014.638,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_2013.200_2014.638.wav,بقداش يا خويا,1.4380000000000002
+262824,audio4_1h_transcrit_3512.147_3515.152,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_3512.147_3515.152.wav,قلي آني كنت باش نسافر إليوم,3.005
+273921,audio4_1h_transcrit_3642.557_3644.227,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_3642.557_3644.227.wav,متاع الثلاثة و درج هاذا,1.67
+270512,audio10_1654.348_1655.621,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio10_1654.348_1655.621.wav,عسلامة,1.273
+273531,audio5_Transcrit_408.042_409.358,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_408.042_409.358.wav,صباح الخير,1.3159999999999998
+276096,audio5_Transcrit_3027.938_3028.988,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_3027.938_3028.988.wav,وحدة تونس,1.05
+271977,audio6.tst_147.396_148.771,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio6.tst_147.396_148.771.wav,هذي التكاي و هاذا الدينار,1.375
+266323,audio11_Transcrit_2596.029_2596.873,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_2596.029_2596.873.wav,ثلاثة,0.8440000000000001
+276428,audio14-1h_577.224_583.781,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_577.224_583.781.wav,آه أيا باهي و اللاهي فجعتني نستخايل الأربعة قلت قداش باش نقعد نستنى توا في اللاڨار أيا باهي بارك الله فيك يرحم والديك,6.557
+263205,audio4_1h_transcrit_2136.598_2138.442,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_2136.598_2138.442.wav,نحساب فيه أكثر من رحلة,1.844
+261969,audio6.tst_197.733_198.380,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio6.tst_197.733_198.380.wav,صباح الخير,0.647
+266581,audio8-1h30.tst_4040.022_4042.205,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_4040.022_4042.205.wav,آكسبراس و إلا دراكت إنتي شتحب,2.1830000000000003
+270532,audio3_transcrit_3285.837_3286.815,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_3285.837_3286.815.wav,آكسبراس,0.978
+260617,audio5_Transcrit_1663.042_1664.545,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_1663.042_1664.545.wav,توا ماضي ساعة و نصف,1.5030000000000001
+270653,audio18_30_524.952_528.157,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_524.952_528.157.wav,أعطيني ألفين و خمسة مية يعيشك,3.205
+261778,audio7_test_transcrit_3921.025_3922.275,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_3921.025_3922.275.wav,آ ڨابس,1.25
+264599,audio7_test_transcrit_4489.914_4490.777,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_4489.914_4490.777.wav,ميتين,0.863
+272815,audio7_test_transcrit_3183.614_3185.455,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_3183.614_3185.455.wav,معناتها حق الكارت خلص,1.841
+261970,audio5_Transcrit_1741.696_1743.696,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_1741.696_1743.696.wav,وقتاش يقصوا,2.0
+268946,audio5_Transcrit_1513.905_1517.889,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_1513.905_1517.889.wav,بربي كيفاش نعمل آني باش روح للجم,3.984
+263151,audio7_test_transcrit_5114.512_5117.355,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_5114.512_5117.355.wav,آ بالله ستة تران تونس,2.843
+269245,audio10_1555.684_1558.644,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio10_1555.684_1558.644.wav,باهي أوكاي يعيشك أعطيني التكاي,2.96
+270610,audio5_Transcrit_3350.606_3351.699,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_3350.606_3351.699.wav,الستة و أربعة,1.093
+267644,audio5_Transcrit_713.453_715.600,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_713.453_715.600.wav,نجم نقص الجم تونس,2.147
+277094,audio17_1h_1294.057_1295.965,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_1294.057_1295.965.wav,باللاهي شنوة أول تران لصفاقس,1.9080000000000001
+261658,audio4_1h_transcrit_1693.084_1694.210,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_1693.084_1694.210.wav,زوز تونس,1.126
+10506,audio1_2hTranscrit_6401.597_6402.718,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_6401.597_6402.718.wav,الخمسة و خمسة,1.121
+262447,audio14-1h_2563.909_2564.914,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_2563.909_2564.914.wav,أعطيني صرف أمان,1.005
+271333,audio3_transcrit_3615.353_3616.571,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_3615.353_3616.571.wav,ما مشاش بيها,1.218
+268231,audio7_test_transcrit_239.503_240.266,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_239.503_240.266.wav,نعم,0.763
+261534,audio11_Transcrit_1546.871_1548.278,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_1546.871_1548.278.wav,بلاصة و نصف تونس,1.4069999999999998
+273816,audio3_transcrit_3287.798_3288.985,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_3287.798_3288.985.wav,و فمة في لآكسبراس,1.187
+269668,audio18_30_1283.188_1285.143,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_1283.188_1285.143.wav,أي تكاي وحدة أي,1.955
+15820,audio1_2hTranscrit_2486.442_2487.895,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_2486.442_2487.895.wav,الستة متاع الصباح,1.453
+16015,audio6.tst_45.661_46.510,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio6.tst_45.661_46.510.wav,تفضل يا مدام,0.8490000000000001
+267908,audio3_transcrit_3022.185_3022.977,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_3022.185_3022.977.wav,ثلاثين,0.792
+277986,audio1_2hTranscrit_2394.213_2395.072,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_2394.213_2395.072.wav,شنوة,0.8590000000000001
+266729,audio4_1h_transcrit_2117.490_2119.286,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_2117.490_2119.286.wav,يا ذهاب و عودة,1.796
+264898,audio1_2hTranscrit_3633.724_3636.849,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_3633.724_3636.849.wav,و متاع الستة غير درج متاع ماطر موجود,3.125
+267899,audio11_Transcrit_2115.211_2117.993,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_2115.211_2117.993.wav,الأربعة و نصف و بعد السبعة و نصف,2.782
+266459,audio4_1h_transcrit_2958.173_2961.079,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_2958.173_2961.079.wav,بربي تران الصباح أول تران,2.906
+261835,audio4_1h_transcrit_699.018_700.143,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_699.018_700.143.wav,بارك الله فيك,1.125
+266103,audio1_2hTranscrit_1977.211_1979.076,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_1977.211_1979.076.wav,و أعطيني وحدة ألاي رتور,1.865
+277806,audio1_2hTranscrit_2067.017_2070.451,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_2067.017_2070.451.wav,تران بالله تونس الماضي ساعة و ربع متاع الليل,3.4339999999999997
+267121,audio12_2189.848_2193.592,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_2189.848_2193.592.wav,ماه أقل من عشرة سنين يخلص نصف صحيح و إلا لا,3.7439999999999998
+277523,audio10_1121.954_1125.472,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio10_1121.954_1125.472.wav,دو تكاي بلاصة كاملة ألاي و رتور,3.5180000000000002
+265383,audio8-1h30.tst_662.141_663.125,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_662.141_663.125.wav,أي باهي,0.9840000000000001
+269709,audio7_test_transcrit_2919.721_2921.467,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_2919.721_2921.467.wav,هات أي هات,1.746
+270049,audio12_324.787_328.267,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_324.787_328.267.wav,حداش و تسعة مية شوفلي ألفين,3.48
+278087,audio1_2hTranscrit_429.786_431.341,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_429.786_431.341.wav,أربعة و ثلاثين دينار و ميتين,1.555
+274394,audio7_test_transcrit_1671.513_1673.201,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_1671.513_1673.201.wav,ستة و أربعة,1.6880000000000002
+273948,audio10_3460.409_3462.097,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio10_3460.409_3462.097.wav,ألاي رتور آه,1.6880000000000002
+260666,audio18_30_1535.886_1537.195,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_1535.886_1537.195.wav,و الكنفور,1.3090000000000002
+263924,audio5_Transcrit_788.685_789.651,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_788.685_789.651.wav,صباح الخير,0.966
+270402,audio5_Transcrit_3418.982_3422.045,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_3418.982_3422.045.wav,سانكنت سانك بروميار سوسة,3.063
+272296,audio8-1h30.tst_203.630_205.678,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_203.630_205.678.wav,معناتها بعد الثالث جور,2.048
+262699,audio1_2hTranscrit_5274.256_5278.148,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_5274.256_5278.148.wav,غدوة الحي الصباح فمة تران لتونس مع الخمسة هكا,3.892
+262641,audio10_1797.784_1800.744,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio10_1797.784_1800.744.wav,أنا عندي كارت دو فدالتاي نجم نقص بيها,2.96
+273705,audio3_transcrit_1823.346_1824.627,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_1823.346_1824.627.wav,نعطيوك بروميار,1.281
+276782,audio18_30_979.813_980.937,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_979.813_980.937.wav,أي,1.124
+276670,audio12_2193.592_2195.392,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_2193.592_2195.392.wav,أي باهي أي تفضل أي,1.8
+276849,audio1_2hTranscrit_778.288_779.015,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_778.288_779.015.wav,أيه,0.727
+269385,audio7_test_transcrit_127.799_129.139,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_127.799_129.139.wav,تفضل أي,1.34
+277418,audio1_2hTranscrit_5281.062_5282.283,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_5281.062_5282.283.wav,يخرج من هنا,1.2209999999999999
+275328,audio8-1h30.tst_3560.657_3562.766,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_3560.657_3562.766.wav,أي تفضل هذي التكاي متاعك أي,2.109
+265993,audio5_Transcrit_61.587_62.853,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_61.587_62.853.wav,كيف كيف آكسبراس,1.266
+274176,audio7_test_transcrit_520.620_523.979,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_520.620_523.979.wav,فمة توا الماضي ساعة و نصف,3.359
+263194,audio7_test_transcrit_1884.408_1885.534,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_1884.408_1885.534.wav,بلاصة,1.126
+268452,audio7_test_transcrit_3594.026_3595.885,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_3594.026_3595.885.wav,إنتي ماشية لتونس,1.859
+266674,audio18_30_1608.682_1610.099,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_1608.682_1610.099.wav,تفضل يا بنتي,1.4169999999999998
+267528,audio7_test_transcrit_4701.219_4702.583,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_4701.219_4702.583.wav,لا ما فماش,1.364
+265248,audio10_1256.284_1257.972,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio10_1256.284_1257.972.wav,عسلامة أختي,1.6880000000000002
+275581,audio3_transcrit_2789.373_2790.202,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_2789.373_2790.202.wav,وقتاش,0.8290000000000001
+264044,audio1_2hTranscrit_919.723_921.254,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_919.723_921.254.wav,باللاهي صفاقس تونس,1.531
+271795,audio11_Transcrit_3774.621_3776.074,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_3774.621_3776.074.wav,ماضي ساعة و نصف,1.453
+265193,audio11_Transcrit_647.450_649.106,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_647.450_649.106.wav,فمة ماضي ساعة,1.656
+265175,audio8-1h30.tst_2983.300_2984.321,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_2983.300_2984.321.wav,آ,1.021
+261423,audio17_1h_2633.243_2633.904,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_2633.243_2633.904.wav,لحظة,0.6609999999999999
+277832,audio11_Transcrit_3142.920_3144.600,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_3142.920_3144.600.wav,قداش ألفين و مية,1.68
+272719,audio1_2hTranscrit_3397.359_3399.935,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_3397.359_3399.935.wav,و أربعة أعطيني بروميار كلاس,2.576
+272799,audio17_1h_1226.287_1228.681,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_1226.287_1228.681.wav,مشي و جي أيه مشي و جي,2.394
+269438,audio5_Transcrit_1711.912_1714.084,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_1711.912_1714.084.wav,ماضي ساعة هو,2.1719999999999997
+270789,audio4_1h_transcrit_3338.252_3339.155,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_3338.252_3339.155.wav,عادي,0.903
+276257,audio8-1h30.tst_5176.098_5179.674,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_5176.098_5179.674.wav,فمة تران نصف نهار و خمسة,3.576
+264308,audio5_Transcrit_1128.412_1130.086,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_1128.412_1130.086.wav,فمة تران تونس توا,1.6740000000000002
+262418,audio4_1h_transcrit_3474.612_3475.690,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_3474.612_3475.690.wav,سلام عليكم,1.078
+261258,audio3_transcrit_3261.976_3262.787,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_3261.976_3262.787.wav,ثناش,0.8109999999999999
+276712,audio11_Transcrit_1465.894_1467.358,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_1465.894_1467.358.wav,تخلص البنية,1.464
+266969,audio18_30_2032.507_2035.578,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_2032.507_2035.578.wav,و و ثلاثة بالكارت جون و زوز بلاش,3.071
+269967,audio11_Transcrit_1011.482_1013.794,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_1011.482_1013.794.wav,أيه ألفين و ستة مية أي,2.312
+270368,audio7_test_transcrit_4679.559_4680.518,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_4679.559_4680.518.wav,تونس,0.9590000000000001
+262552,audio14-1h_1962.602_1968.272,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_1962.602_1968.272.wav,أيا باهي هاتهم أي بارك الله فيك وقتاش يخرج بالظبط التران,5.67
+267089,audio1_2hTranscrit_6440.628_6441.331,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_6440.628_6441.331.wav,تونس,0.703
+268039,audio3_transcrit_2340.341_2342.511,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_2340.341_2342.511.wav,السبعة غير درجين الصباح آكسبراس,2.17
+271467,audio8-1h30.tst_3912.400_3914.830,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_3912.400_3914.830.wav,أعطيني تسعطاش ألف و ثلاثة مية,2.43
+270777,audio4_1h_transcrit_554.402_555.823,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_554.402_555.823.wav,وقتاش تخرج,1.421
+266326,audio5_Transcrit_3455.400_3457.077,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_3455.400_3457.077.wav,لا ثنين درجة أولى,1.6769999999999998
+274156,audio12_2696.115_2698.635,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_2696.115_2698.635.wav,توزر قداش توزر,2.52
+275220,audio11_Transcrit_2110.315_2111.563,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_2110.315_2111.563.wav,الأربعة و نصف,1.248
+268048,audio11_Transcrit_1721.042_1721.618,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_1721.042_1721.618.wav,آه,0.5760000000000001
+268080,audio8-1h30.tst_1432.735_1434.248,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_1432.735_1434.248.wav,ما فهمتكش,1.5130000000000001
+270118,audio3_transcrit_794.348_795.497,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_794.348_795.497.wav,ڨابس وقتاش,1.149
+266241,audio18_30_697.857_700.185,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_697.857_700.185.wav,بالله ألاي رتور تونس,2.3280000000000003
+264005,audio5_Transcrit_1647.592_1650.268,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_1647.592_1650.268.wav,بالله تونس وقتاش خويا,2.676
+262583,audio7_test_transcrit_5845.331_5847.315,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_5845.331_5847.315.wav,ألاي رتور تونس,1.984
+10333,audio3_transcrit_658.518_659.234,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_658.518_659.234.wav,وين,0.716
+272669,audio8-1h30.tst_416.916_418.561,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_416.916_418.561.wav,واحد عمرو عشرة سنين,1.645
+269999,audio10_1352.457_1355.044,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio10_1352.457_1355.044.wav,باهي تفضل آهوم الفلوس,2.5869999999999997
+264676,audio1_2hTranscrit_6289.956_6291.201,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_6289.956_6291.201.wav,معناتها,1.245
+262052,audio4_1h_transcrit_583.769_584.556,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_583.769_584.556.wav,لا,0.787
+10195,audio8-1h30.tst_654.714_656.449,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_654.714_656.449.wav,باهي أهم أربعطاش ألف,1.735
+274425,audio1_2hTranscrit_1109.892_1111.638,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_1109.892_1111.638.wav,ثلاثة و ثلاثين و ثمنية مية,1.746
+263783,audio7_test_transcrit_3062.611_3065.001,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_3062.611_3065.001.wav,ألاي رتور و إلا ألاي برك,2.39
+277560,audio11_Transcrit_3935.162_3936.002,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_3935.162_3936.002.wav,كيفاش,0.84
+270758,audio18_30_1206.902_1209.960,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_1206.902_1209.960.wav,سلام عليكم باللاهي وحدة لمتلوي,3.0580000000000003
+271291,audio14-1h_108.104_114.925,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_108.104_114.925.wav,باهي مالا أعطيني تكاي في التران إلي بعد مع الأربعة و أربعة بقداش هي التكاي أي تفضل آهم الفلوس,6.821000000000001
+277950,audio1_2hTranscrit_5813.629_5815.316,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_5813.629_5815.316.wav,أول تران ماضي ساعة و أربعة,1.6869999999999998
+260700,audio8-1h30.tst_65.031_65.991,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_65.031_65.991.wav,أي نقصلك و إلا لا,0.96
+260523,audio18_30_2055.504_2057.200,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_2055.504_2057.200.wav,بروميار كلاس تحب,1.696
+275514,audio18_30_1447.038_1449.239,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_1447.038_1449.239.wav,باهي تحب بروميار كلاس,2.201
+274103,audio4_1h_transcrit_3493.503_3496.083,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_3493.503_3496.083.wav,تونس فمة توا السريع ثلاثة غير درج,2.58
+261578,audio1_2hTranscrit_2563.159_2564.050,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_2563.159_2564.050.wav,تفضل,0.8909999999999999
+261877,audio3_transcrit_3488.734_3492.483,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_3488.734_3492.483.wav,عسلامة بربي سوسة بالله الأربعة و نصف لآكسبراس فمة,3.7489999999999997
+277820,audio1_2hTranscrit_6456.997_6461.214,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_6456.997_6461.214.wav,إيه أنا قلت لك سومها الحقاني في وسط الجمعة يرخصوا فيها,4.217
+276100,audio11_Transcrit_672.272_673.016,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_672.272_673.016.wav,أيه,0.7440000000000001
+273868,audio1_2hTranscrit_5966.482_5969.587,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_5966.482_5969.587.wav,كهو يا العشرة متاع الصباح يا الساعتين و ربع متاع الليل,3.105
+268135,audio8-1h30.tst_4329.116_4330.780,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_4329.116_4330.780.wav,أي باهي بقداش البليصة,1.6640000000000001
+264432,audio17_1h_936.617_937.867,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_936.617_937.867.wav,أمم,1.25
+266764,audio11_Transcrit_1290.678_1293.072,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_1290.678_1293.072.wav,ألفين ماضي ساعة و نصف,2.394
+269390,audio1_2hTranscrit_2533.552_2538.159,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_2533.552_2538.159.wav,و ثمنية مية تولي ثمنية و عشرين دينار فمة تقريب سانك دينار,4.607
+270107,audio14-1h_1824.744_1825.985,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_1824.744_1825.985.wav,قداش قلت لي سامحني,1.2409999999999999
+265311,audio7_test_transcrit_1734.718_1735.911,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_1734.718_1735.911.wav,ثمنطاش,1.193
+272622,audio4_1h_transcrit_2840.216_2841.675,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_2840.216_2841.675.wav,ماضي ساعة و نصف,1.459
+265931,audio3_transcrit_3682.575_3684.392,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_3682.575_3684.392.wav,ستة و أربعة يخرج من هنا,1.817
+269885,audio4_1h_transcrit_658.838_659.900,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_658.838_659.900.wav,أوكي البنوتة,1.062
+278136,audio11_Transcrit_386.294_388.325,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_386.294_388.325.wav,و و ثمة تران بعد,2.031
+273526,audio1_2hTranscrit_4799.695_4800.633,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_4799.695_4800.633.wav,عشرة أيام,0.938
+266007,audio7_test_transcrit_196.373_198.967,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_196.373_198.967.wav,بروميار دينار و خمسين يعيش إختي,2.594
+260330,audio1_2hTranscrit_3724.770_3725.880,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_3724.770_3725.880.wav,بروميار بروميار,1.11
+274129,audio17_1h_53.699_53.967,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_53.699_53.967.wav,آ,0.268
+277101,audio11_Transcrit_3102.009_3103.305,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_3102.009_3103.305.wav,هاكي خليها عندك,1.296
+265488,audio3_transcrit_3928.512_3929.606,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_3928.512_3929.606.wav,لاي زورار,1.094
+268974,audio18_30_1745.633_1749.758,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_1745.633_1749.758.wav,ستطاش و تسعة مية عندكشي ألفين غادي,4.125
+277997,audio5_Transcrit_1480.513_1484.404,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_1480.513_1484.404.wav,تران إلي مروح لتوز لتوزر وقتاش يخلط,3.891
+10386,audio18_30_1364.881_1375.204,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_1364.881_1375.204.wav,بالله عندي بياي هذي آ جا بيها ولدي و الرجوع موش باش يرجع بيها حتى شد في خدمة بعثتلو شركة,10.323
+271656,audio14-1h_2247.508_2249.061,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_2247.508_2249.061.wav,ثلاثة آلاف و خمسة مية عندك,1.5530000000000002
+265906,audio1_2hTranscrit_5353.019_5354.129,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_5353.019_5354.129.wav,نحب نعمل,1.11
+275584,audio5_Transcrit_1633.263_1634.091,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_1633.263_1634.091.wav,بقداه,0.828
+265821,audio14-1h_1777.493_1779.421,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_1777.493_1779.421.wav,عندكشي دينار و مية و إلا دينار و خمسة مية,1.9280000000000002
+263826,audio1_2hTranscrit_1601.475_1604.006,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_1601.475_1604.006.wav,تي طالع في التران هاذيا آني,2.531
+267542,audio1_2hTranscrit_4075.012_4077.144,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_4075.012_4077.144.wav,وقتاش التران غدوة الحي الصباح,2.1319999999999997
+271853,audio18_30_426.864_428.767,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_426.864_428.767.wav,ثناش ألف و متين و خمسين,1.903
+266601,audio1_2hTranscrit_139.222_140.729,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_139.222_140.729.wav,عندك لاي زورار خويا,1.507
+10221,audio4_1h_transcrit_3289.633_3290.977,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_3289.633_3290.977.wav,ثلاثة غير درج,1.344
+260143,audio4_1h_transcrit_3310.179_3311.018,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_3310.179_3311.018.wav,سلام,0.8390000000000001
+276691,audio14-1h_455.150_456.391,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_455.150_456.391.wav,آكسبراس راهو آ,1.2409999999999999
+266039,audio17_1h_2920.520_2921.791,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_2920.520_2921.791.wav,لا موش عليك إنتي,1.271
+260208,audio10_350.185_352.140,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio10_350.185_352.140.wav,باهي ملا أعطيني تكاي,1.955
+276788,audio12_2186.248_2187.184,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_2186.248_2187.184.wav,سلام,0.9359999999999999
+276798,audio18_30_1213.953_1214.670,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_1213.953_1214.670.wav,أيه,0.7170000000000001
+271406,audio4_1h_transcrit_469.362_470.638,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_469.362_470.638.wav,الحداش و ربع,1.276
+267052,audio1_2hTranscrit_6024.415_6029.118,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_6024.415_6029.118.wav,ڨابس عنا يا الصباح العشرة متاع الصباح يا الساعتين و أربعة متاع الليل,4.703
+271628,audio7_test_transcrit_4602.135_4603.428,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_4602.135_4603.428.wav,لا لا,1.2930000000000001
+274505,audio5_Transcrit_270.526_272.009,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_270.526_272.009.wav,ثناش دينار,1.483
+264057,audio7_test_transcrit_4951.108_4952.405,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_4951.108_4952.405.wav,السلام,1.297
+261199,audio18_30_1008.406_1010.421,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_1008.406_1010.421.wav,بياي تسكرة لتونس,2.015
+264957,audio1_2hTranscrit_2680.594_2681.469,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_2680.594_2681.469.wav,تفضل,0.875
+15823,audio3_transcrit_452.619_453.578,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_452.619_453.578.wav,غدوة أيه,0.9590000000000001
+271878,audio6.tst_2983.904_2985.792,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio6.tst_2983.904_2985.792.wav,دهماني ألاي و رتور ستطاش و ستة مية,1.8880000000000001
+272147,audio1_2hTranscrit_5317.898_5319.191,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_5317.898_5319.191.wav,باللاهي تران الفجر,1.2930000000000001
+272744,audio11_Transcrit_136.254_138.723,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_136.254_138.723.wav,سوسة ثلاثة آلاف و ميتين و خمسين,2.469
+263580,audio7_test_transcrit_1902.440_1905.128,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_1902.440_1905.128.wav,فمة خويا تران ماشي لڨابس توا,2.688
+263274,audio8-1h30.tst_2826.379_2829.136,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_2826.379_2829.136.wav,تران إلي ماشي تو لصفاقس ياقف في الحنشة,2.7569999999999997
+273392,audio14-1h_3386.498_3393.431,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_3386.498_3393.431.wav,باهي مدام فمة قصلي تكاي تعبى و إلا مازال تجم تشوفلي,6.933
+275359,audio5_Transcrit_2675.636_2677.191,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_2675.636_2677.191.wav,ما فمشي ساعتين و درجين,1.555
+269379,audio3_transcrit_1788.077_1789.451,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_1788.077_1789.451.wav,ديمة يمشي التران,1.374
+275423,audio4_1h_transcrit_253.374_254.452,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_253.374_254.452.wav,آه فهمتك,1.078
+274564,audio17_1h_2647.038_2649.239,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_2647.038_2649.239.wav,لا لا ماضي ساعتين بعشرة آلاف و ثلاثة مية,2.201
+276797,audio1_2hTranscrit_3264.944_3265.803,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_3264.944_3265.803.wav,الخمسة,0.8590000000000001
+262526,audio4_1h_transcrit_3519.707_3520.446,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_3519.707_3520.446.wav,أيه,0.7390000000000001
+267207,audio5_Transcrit_2307.038_2311.068,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_2307.038_2311.068.wav,باهي نعمل نجم نعمل كارت جون توا,4.03
+263844,audio11_Transcrit_3821.778_3823.997,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_3821.778_3823.997.wav,ثلاثة و عشرين و ثمنية مية الزوز,2.219
+270236,audio5_Transcrit_1654.877_1656.050,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_1654.877_1656.050.wav,تفضل أي,1.173
+269475,audio17_1h_266.321_267.857,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_266.321_267.857.wav,معاش تخلط عليه تو,1.536
+270182,audio8-1h30.tst_3906.619_3908.069,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_3906.619_3908.069.wav,دوزيام كلاس,1.45
+273181,audio5_Transcrit_501.615_502.812,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_501.615_502.812.wav,أوفات,1.197
+266933,audio11_Transcrit_4704.435_4705.825,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_4704.435_4705.825.wav,الله يرحم والديك,1.39
+271906,audio1_2hTranscrit_2789.980_2790.743,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_2789.980_2790.743.wav,تو,0.763
+268075,audio5_Transcrit_542.212_543.171,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_542.212_543.171.wav,أيوه,0.9590000000000001
+263713,audio12_1818.663_1823.919,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_1818.663_1823.919.wav,باللاهي تران تونس ماضي ساعة متاع الليل نحب بروميار آه,5.256
+266219,audio12_1359.678_1361.838,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_1359.678_1361.838.wav,عيش ولدي راني ما عنديش صروف,2.16
+264321,audio5_Transcrit_364.591_366.997,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_364.591_366.997.wav,ستة آلاف و ستة مية و خمسين,2.406
+10274,audio11_Transcrit_3336.707_3338.435,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_3336.707_3338.435.wav,آكسبراس بالله وقتاش لتونس,1.7280000000000002
+276993,audio3_transcrit_96.053_97.538,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_96.053_97.538.wav,باهي بحول الله,1.485
+269844,audio1_2hTranscrit_6267.480_6268.081,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_6267.480_6268.081.wav,سلام,0.601
+265757,audio5_Transcrit_3117.800_3118.806,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_3117.800_3118.806.wav,و العادي,1.006
+261212,audio7_test_transcrit_5929.296_5931.020,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_5929.296_5931.020.wav,أيا خويا أي,1.724
+15738,audio11_Transcrit_5027.236_5028.004,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_5027.236_5028.004.wav,الحداش,0.768
+10007,audio3_transcrit_2291.171_2292.124,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_2291.171_2292.124.wav,مارسي,0.953
+262514,audio17_1h_1507.474_1508.903,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_1507.474_1508.903.wav,أي تفضل أي,1.429
+269635,audio7_test_transcrit_3733.809_3738.276,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_3733.809_3738.276.wav,غدوة الصباح الخمسة و خمسة و عشرين دقيقة,4.467
+262507,audio18_30_1496.319_1497.909,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_1496.319_1497.909.wav,متلوي الحداش,1.59
+276334,audio1_2hTranscrit_5106.666_5107.557,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_5106.666_5107.557.wav,قداش,0.8909999999999999
+261156,audio7_test_transcrit_5412.974_5414.143,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_5412.974_5414.143.wav,أيه,1.169
+266598,audio12_942.207_945.116,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_942.207_945.116.wav,تونس وقتاه,2.909
+262634,audio5_Transcrit_1029.867_1033.195,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_1029.867_1033.195.wav,لآكسبراس العشية آني وقت يخرج,3.3280000000000003
+276318,audio8-1h30.tst_1968.306_1969.204,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_1968.306_1969.204.wav,تفضل,0.898
+260501,audio5_Transcrit_3188.954_3192.011,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_3188.954_3192.011.wav,كالعادة أوك الأوقات متاع العادة,3.057
+271033,audio1_2hTranscrit_6495.551_6497.129,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_6495.551_6497.129.wav,السبعة غير درجين آكسبراس,1.578
+273504,audio3_transcrit_1743.128_1743.867,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_1743.128_1743.867.wav,تونس,0.7390000000000001
+277874,audio7_test_transcrit_1519.128_1520.333,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_1519.128_1520.333.wav,الستة و أربعة,1.205
+276374,audio4_1h_transcrit_2534.849_2536.459,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_2534.849_2536.459.wav,أوقات المشيان لتونس,1.61
+269593,audio7_test_transcrit_1325.396_1326.595,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_1325.396_1326.595.wav,وقتاه خويا التران,1.199
+274663,audio12_3403.484_3417.404,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_3403.484_3417.404.wav,لا راهو ثلاثة غير درج في رمضان أما تويكة في الإفطار يولي يولي الساعتين غير درج أعطيني ستطاش ألف ميتين و خمسين و أعطيني الكارت جون أعطيني نومرو متاعها,13.92
+270417,audio8-1h30.tst_3472.231_3474.008,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_3472.231_3474.008.wav,ثمة ترأنات,1.777
+271566,audio1_2hTranscrit_4845.372_4846.688,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_4845.372_4846.688.wav,قلعة صغرى,1.3159999999999998
+278077,audio7_test_transcrit_900.132_901.687,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_900.132_901.687.wav,أربعة سوايع يقعد,1.555
+268090,audio5_Transcrit_777.604_778.463,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_777.604_778.463.wav,زوز نابل,0.8590000000000001
+277004,audio4_1h_transcrit_2182.144_2183.079,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_2182.144_2183.079.wav,زوز من ناس,0.935
+265691,audio3_transcrit_2701.965_2703.658,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_2701.965_2703.658.wav,دوزيام قلت لي,1.693
+274626,audio17_1h_1002.457_1003.163,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_1002.457_1003.163.wav,ما يقلقنيش,0.706
+269983,audio1_2hTranscrit_1207.347_1209.175,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_1207.347_1209.175.wav,توا الماضي ساعة و ربع,1.828
+266105,audio3_transcrit_3205.183_3206.183,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_3205.183_3206.183.wav,تونس,1.0
+273774,audio1_2hTranscrit_2445.416_2446.041,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_2445.416_2446.041.wav,آه,0.625
+264815,audio7_test_transcrit_5699.819_5700.928,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_5699.819_5700.928.wav,أيه,1.109
+274959,audio7_test_transcrit_2764.021_2768.147,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_2764.021_2768.147.wav,تران لتونس فمة توا الأربعة و نصف و فمة الستة و الأربعة أدراج,4.126
+267827,audio1_2hTranscrit_6019.722_6020.785,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_6019.722_6020.785.wav,تفضل أي,1.063
+274447,audio3_transcrit_525.416_526.056,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_525.416_526.056.wav,أيه,0.64
+268407,audio4_1h_transcrit_2432.428_2433.124,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_2432.428_2433.124.wav,سلام,0.696
+266124,audio11_Transcrit_3362.002_3363.564,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_3362.002_3363.564.wav,تسكرة بروميار,1.5619999999999998
+276676,audio3_transcrit_1354.877_1357.861,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_1354.877_1357.861.wav,تران متاع ماضي ساعة و نصف ياقف في حمام الأنف,2.984
+271181,audio6.tst_2726.323_2730.577,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio6.tst_2726.323_2730.577.wav,أهلا و سهلا باللاهي عيشك فمة تران لسوسة توا آكسبراس ولا نورمال,4.254
+269695,audio3_transcrit_3095.700_3097.443,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_3095.700_3097.443.wav,أيه ييه توا,1.743
+265729,audio1_2hTranscrit_6478.999_6479.882,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_6478.999_6479.882.wav,عسلامة,0.883
+260520,audio11_Transcrit_2232.785_2234.661,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_2232.785_2234.661.wav,ألاي و رتور تونس,1.876
+272736,audio3_transcrit_1465.640_1466.694,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_1465.640_1466.694.wav,حمام الأنف,1.054
+266451,audio14-1h_2450.561_2452.321,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_2450.561_2452.321.wav,نابل مع أما وقت يا بابا,1.76
+268854,audio7_test_transcrit_2730.348_2734.833,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_2730.348_2734.833.wav,توصل السبعة و نصف فهمت الفازة,4.485
+277372,audio1_2hTranscrit_1474.951_1475.623,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_1474.951_1475.623.wav,وين,0.672
+268383,audio7_test_transcrit_3819.817_3820.299,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_3819.817_3820.299.wav,آ,0.48200000000000004
+9990,audio12_1928.252_1929.487,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_1928.252_1929.487.wav,إنشاء الله أي,1.235
+276300,audio1_2hTranscrit_3343.543_3344.184,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_3343.543_3344.184.wav,آه,0.6409999999999999
+264611,audio3_transcrit_2954.143_2955.080,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_2954.143_2955.080.wav,ثناش دينار,0.937
+268241,audio1_2hTranscrit_1618.121_1619.199,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_1618.121_1619.199.wav,مازالو درجين,1.078
+263329,audio14-1h_3290.693_3298.943,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_3290.693_3298.943.wav,خمسين فرنك توا نشوفلك في بإلي عندي مية ما عنديش خمسينات أي توا نشوفلك أي باهي تفضل آهو لقيتلك خمسين,8.25
+275013,audio8-1h30.tst_495.522_497.706,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_495.522_497.706.wav,أعطينا ثلاثة و ثلاثين دينار و,2.184
+271636,audio11_Transcrit_3386.284_3388.847,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_3386.284_3388.847.wav,عندكشي مية و خمسين شوية صرف صغير,2.563
+271077,audio18_30_366.225_367.488,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_366.225_367.488.wav,لا ما عنديش,1.263
+268101,audio14-1h_2410.073_2410.667,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_2410.073_2410.667.wav,أهلا,0.594
+266854,audio11_Transcrit_687.765_689.952,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_687.765_689.952.wav,لا لا ثلاثة غير درج,2.187
+277414,audio17_1h_1203.933_1205.500,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_1203.933_1205.500.wav,ألاي و رتور في الدوزيام,1.567
+276158,audio1_2hTranscrit_230.681_231.945,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_230.681_231.945.wav,الخمسة و خمسة,1.264
+275436,audio10_1390.549_1393.420,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio10_1390.549_1393.420.wav,فمة كان في الليل مع التسعة و نصف,2.8710000000000004
+277486,audio5_Transcrit_3481.043_3482.049,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_3481.043_3482.049.wav,تونس,1.006
+261999,audio6.tst_2633.243_2633.904,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio6.tst_2633.243_2633.904.wav,لحظة,0.6609999999999999
+260207,audio4_1h_transcrit_2276.239_2277.388,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_2276.239_2277.388.wav,تونس,1.149
+271284,audio17_1h_1580.155_1582.098,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_1580.155_1582.098.wav,وقتاش يخرج هو,1.943
+276316,audio11_Transcrit_2956.430_2958.258,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_2956.430_2958.258.wav,و عليكم السلام و رحمة الله,1.828
+9983,audio5_Transcrit_3105.026_3105.551,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_3105.026_3105.551.wav,أيه,0.525
+268072,audio12_3202.436_3210.428,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_3202.436_3210.428.wav,و إلي بعدو آ آ فمة الأربعة غير ربع و فمة في الليل الكلهم القلعة الصغرى راهو موش سنتر آه,7.992000000000001
+274969,audio8-1h30.tst_1529.745_1531.751,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_1529.745_1531.751.wav,أي باهي أي آكسبراس راهو,2.006
+273084,audio17_1h_2787.469_2791.447,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_2787.469_2791.447.wav,أيه أيه سوسة راهو موش صفاقس أيا تفضل آهم الفلوس أي,3.978
+268372,audio1_2hTranscrit_5001.155_5002.202,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_5001.155_5002.202.wav,ستطاش و تسعة مية,1.047
+269976,audio4_1h_transcrit_3307.283_3308.314,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_3307.283_3308.314.wav,آكسبراس أيه,1.031
+269098,audio17_1h_3209.685_3216.039,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_3209.685_3216.039.wav,عسلامة باللاهي عيش أختي مازالوا فمة ترأنات المهدية وقتاش كان فمة وقتاش يخرج هو التران بالظبط,6.354
+273193,audio7_test_transcrit_4132.130_4134.658,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_4132.130_4134.658.wav,ميتين و خمسة و عشرين ثلاثة تسعات,2.528
+272699,audio5_Transcrit_2760.797_2761.859,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_2760.797_2761.859.wav,أوكاي مارسي,1.062
+267294,audio12_3389.300_3393.140,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_3389.300_3393.140.wav,خمسطاش ألف و ميتين و خمسين أعطيني ميتين و خمسين,3.84
+264347,audio12_1282.511_1284.214,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_1282.511_1284.214.wav,قفلهم و إلا مازال,1.703
+272462,audio5_Transcrit_1867.794_1871.106,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_1867.794_1871.106.wav,القلعة الصغيرة محطة بعد سوسة بثلاثة كيلو,3.312
+269180,audio4_1h_transcrit_2210.432_2211.558,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_2210.432_2211.558.wav,ستين ألف,1.126
+272203,audio1_2hTranscrit_5167.562_5168.664,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_5167.562_5168.664.wav,زوز تونس إيه,1.102
+260980,audio1_2hTranscrit_5838.906_5841.249,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_5838.906_5841.249.wav,البرومسيون مازالت تبطى واشي,2.343
+270861,audio11_Transcrit_2140.479_2141.448,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_2140.479_2141.448.wav,السريع,0.9690000000000001
+272005,audio7_test_transcrit_888.890_889.796,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_888.890_889.796.wav,أيه,0.9059999999999999
+269510,audio18_30_8.133_12.113,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_8.133_12.113.wav,آه دو تكاي الجم باهي أعطيني سبعطاش ألف و خمسة مية,3.98
+265896,audio12_2962.210_2964.394,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_2962.210_2964.394.wav,بلاصة و نصف لتونس,2.184
+270403,audio5_Transcrit_465.707_466.761,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_465.707_466.761.wav,دوزيام,1.054
+272435,audio7_test_transcrit_2895.068_2896.766,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_2895.068_2896.766.wav,كلهم بروميار كلاس,1.6980000000000002
+278025,audio6.tst_10.268_11.607,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio6.tst_10.268_11.607.wav,دراكت و إلا آكسبراس,1.339
+276201,audio11_Transcrit_1513.094_1513.934,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_1513.094_1513.934.wav,لا متاع,0.84
+262825,audio11_Transcrit_3507.485_3508.204,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_3507.485_3508.204.wav,دوزيام,0.7190000000000001
+267472,audio7_test_transcrit_5681.454_5682.365,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_5681.454_5682.365.wav,أوه,0.9109999999999999
+261872,audio18_30_1982.346_1984.579,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_1982.346_1984.579.wav,أيا تفضل خويا,2.233
+261978,audio14-1h_3617.414_3619.053,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_3617.414_3619.053.wav,وريني الكارت,1.639
+277008,audio3_transcrit_2529.953_2530.999,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_2529.953_2530.999.wav,ماضي ساعة و نصف,1.046
+261954,audio1_2hTranscrit_2160.048_2160.845,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_2160.048_2160.845.wav,أيه ييه,0.797
+271357,audio7_test_transcrit_1299.893_1302.044,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_1299.893_1302.044.wav,الأربعة و نصف إنشاء الله,2.151
+270227,audio14-1h_2936.169_2937.656,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_2936.169_2937.656.wav,بقداش الزوز يطلعولي,1.4869999999999999
+260883,audio8-1h30.tst_1981.030_1983.458,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_1981.030_1983.458.wav,آ القلعة الكبرى و إلا القلعة الصغرى,2.428
+266418,audio8-1h30.tst_1055.615_1056.490,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_1055.615_1056.490.wav,أهلا,0.875
+261452,audio7_test_transcrit_974.326_975.233,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio7_test_transcrit_974.326_975.233.wav,سوسة,0.907
+264155,audio4_1h_transcrit_1151.247_1152.622,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_1151.247_1152.622.wav,أعطيني بروميار مالا,1.375
+277815,audio5_Transcrit_3352.252_3355.042,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_3352.252_3355.042.wav,فمة تران تخرج لڨابس,2.79
+261656,audio14-1h_891.336_895.109,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_891.336_895.109.wav,لا لا أعطيني نورمال و برة لين آكسبراس ما نحبهاش آكسبراس أنا,3.773
+274992,audio1_2hTranscrit_3933.068_3933.665,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_3933.068_3933.665.wav,نعم,0.597
+10294,audio3_transcrit_2874.441_2875.270,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_2874.441_2875.270.wav,أي ولدي أي,0.8290000000000001
+269322,audio8-1h30.tst_901.156_902.516,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_901.156_902.516.wav,التسعة و سبعة يا مدام,1.36
+274220,audio18_30_898.599_899.907,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio18_30_898.599_899.907.wav,الله يبارك,1.308
+275271,audio10_2632.204_2633.892,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio10_2632.204_2633.892.wav,بقداه التكاي,1.6880000000000002
+10051,audio11_Transcrit_3921.994_3924.150,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_3921.994_3924.150.wav,وحيدة آكسبراس تونس,2.156
+263861,audio3_transcrit_3168.504_3170.597,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_3168.504_3170.597.wav,لا وحدة برك زوز بير بورڨبة,2.093
+265224,audio4_1h_transcrit_776.833_777.614,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio4_1h_transcrit_776.833_777.614.wav,مرحبا,0.7809999999999999
+260463,audio12_2664.386_2665.730,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_2664.386_2665.730.wav,السلام,1.344
+264094,audio8-1h30.tst_3925.655_3926.618,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_3925.655_3926.618.wav,آ,0.963
+262188,audio8-1h30.tst_4523.169_4524.719,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_4523.169_4524.719.wav,أي أكهو كان إنتي,1.55
+265237,audio5_Transcrit_2711.174_2712.799,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_2711.174_2712.799.wav,الخمسة و خمسة أول تران,1.625
+266458,audio17_1h_1021.866_1023.713,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio17_1h_1021.866_1023.713.wav,بحول الله,1.847
+260336,audio3_transcrit_1679.016_1680.610,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_1679.016_1680.610.wav,تونس بروميار قلت لي,1.594
+10299,audio8-1h30.tst_4792.160_4793.181,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_4792.160_4793.181.wav,داكوردو,1.021
+269875,audio3_transcrit_1134.090_1134.965,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_1134.090_1134.965.wav,وقتاش,0.875
+275840,audio12_629.114_632.042,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_629.114_632.042.wav,نورمال عندكشي ألفين,2.928
+267377,audio14-1h_857.516_858.581,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio14-1h_857.516_858.581.wav,بلاصة و نصف,1.065
+266325,audio12_889.550_894.590,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio12_889.550_894.590.wav,بالله تسكرة لتونس ماضي ساعة ماضي ساعة و قداش هو,5.04
+266617,audio11_Transcrit_4437.900_4438.932,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio11_Transcrit_4437.900_4438.932.wav,العشرة,1.032
+275039,audio1_2hTranscrit_4258.629_4259.421,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio1_2hTranscrit_4258.629_4259.421.wav,الجم,0.792
+263210,audio6.tst_375.099_376.515,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio6.tst_375.099_376.515.wav,قداش من بلاصة حاشتك,1.416
+263778,audio8-1h30.tst_1155.278_1155.903,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio8-1h30.tst_1155.278_1155.903.wav,أكهو,0.625
+270184,audio6.tst_161.784_162.454,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio6.tst_161.784_162.454.wav,بالسلامة,0.67
+276421,audio5_Transcrit_1408.025_1409.289,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio5_Transcrit_1408.025_1409.289.wav,كان مشي برك,1.264
+275401,audio3_transcrit_3025.983_3026.623,/gpfsscratch/rech/nou/uzn19yk/TARIC_2/audio3_transcrit_3025.983_3026.623.wav,باهي,0.64

train_semi.yaml

@@ -0,0 +1,175 @@
+# ################################
+# Model: wav2vec2 + DNN + CTC
+# Augmentation: SpecAugment
+# Authors: Titouan Parcollet 2021
+# ################################
+
+# Seed needs to be set at top of yaml, before objects with parameters are made
+seed: 1234
+__set_seed: !!python/object/apply:torch.manual_seed [!ref <seed>]
+output_folder: !ref semi_wavlm_large_tunisian_ctc/<seed>
+wer_file: !ref <output_folder>/wer.txt
+save_folder: !ref <output_folder>/save
+train_log: !ref <output_folder>/train_log.txt
+
+# URL for the biggest LeBenchmark wav2vec french.
+wav2vec2_folder: !ref <save_folder>/wav2vec2_checkpoint
+
+# Data files
+data_folder: /path/to/data  # e.g, /localscratch/cv-corpus-5.1-2020-06-22/fr
+train_tsv_file: !ref <data_folder>/train.tsv  # Standard CommonVoice .tsv files
+dev_tsv_file: !ref <data_folder>/dev.tsv  # Standard CommonVoice .tsv files
+test_tsv_file: !ref <data_folder>/test.tsv  # Standard CommonVoice .tsv files
+accented_letters: True
+language: fr # use 'it' for Italian, 'rw' for Kinyarwanda, 'en' for english
+test_csv: 
+    - /path/to/test_data
+
+skip_prep: True # Skip data preparation
+
+use_language_modelling: True
+ngram_lm_path: outdomain.arpa
+
+# We remove utterance slonger than 10s in the train/dev/test sets as
+# longer sentences certainly correspond to "open microphones".
+avoid_if_longer_than: 10.0
+avoid_if_shorter_than: 1.2
+
+
+# Training parameters
+number_of_epochs: 12
+lr: 1.0
+lr_wav2vec: 0.0001
+sorting: ascending
+auto_mix_prec: False
+sample_rate: 16000
+ckpt_interval_minutes: 30 # save checkpoint every N min
+
+# With data_parallel batch_size is split into N jobs
+# With DDP batch_size is multiplied by N jobs
+# Must be 6 per GPU to fit 16GB of VRAM
+batch_size: 10
+test_batch_size: 4
+
+dataloader_options:
+    batch_size: !ref <batch_size>
+    num_workers: 6
+test_dataloader_options:
+    batch_size: !ref <test_batch_size>
+    num_workers: 6
+
+# BPE parameters
+token_type: char  # ["unigram", "bpe", "char"]
+character_coverage: 1.0
+
+# Model parameters
+# activation: !name:torch.nn.LeakyReLU
+wav2vec_output_dim: 1024
+dnn_neurons: 1024
+freeze_wav2vec: False
+freeze_feature_extractor: True
+dropout: 0.15
+warmup_steps: 500 # The wav2vec 2 model isn't updated for this amount of steps
+
+# Outputs
+output_neurons: 40  # BPE size, index(blank/eos/bos) = 0
+
+# Decoding parameters
+# Be sure that the bos and eos index match with the BPEs ones
+blank_index: 0
+unk_index: 1
+
+#
+# Functions and classes
+#
+epoch_counter: !new:speechbrain.utils.epoch_loop.EpochCounter
+    limit: !ref <number_of_epochs>
+
+augmentation: !new:speechbrain.lobes.augment.TimeDomainSpecAugment
+    sample_rate: !ref <sample_rate>
+    speeds: [95, 100, 105]
+
+enc: !new:speechbrain.nnet.containers.Sequential
+    input_shape: [null, null, !ref <wav2vec_output_dim>]
+    linear1: !name:speechbrain.nnet.linear.Linear
+        n_neurons: !ref <dnn_neurons>
+        bias: True
+    bn1: !name:speechbrain.nnet.normalization.BatchNorm1d
+    activation: !new:torch.nn.LeakyReLU
+    drop: !new:torch.nn.Dropout
+        p: !ref <dropout>
+    linear2: !name:speechbrain.nnet.linear.Linear
+        n_neurons: !ref <dnn_neurons>
+        bias: True
+    bn2: !name:speechbrain.nnet.normalization.BatchNorm1d
+    activation2: !new:torch.nn.LeakyReLU
+    drop2: !new:torch.nn.Dropout
+        p: !ref <dropout>
+    linear3: !name:speechbrain.nnet.linear.Linear
+        n_neurons: !ref <dnn_neurons>
+        bias: True
+    bn3: !name:speechbrain.nnet.normalization.BatchNorm1d
+    activation3: !new:torch.nn.LeakyReLU
+
+wav2vec2: !new:speechbrain.lobes.models.huggingface_wav2vec.HuggingFaceWav2Vec2
+    source: wavlm-large/ 
+    output_norm: False
+    freeze: !ref <freeze_wav2vec>
+    freeze_feature_extractor: !ref <freeze_feature_extractor>
+    save_path: !ref <wav2vec2_folder>
+
+
+ctc_lin: !new:speechbrain.nnet.linear.Linear
+    input_size: !ref <dnn_neurons>
+    n_neurons: !ref <output_neurons>
+
+log_softmax: !new:speechbrain.nnet.activations.Softmax
+    apply_log: True
+
+ctc_cost: !name:speechbrain.nnet.losses.ctc_loss
+    blank_index: !ref <blank_index>
+
+modules:
+    wav2vec2: !ref <wav2vec2>
+    enc: !ref <enc>
+    ctc_lin: !ref <ctc_lin>
+
+model: !new:torch.nn.ModuleList
+    - [!ref <enc>, !ref <ctc_lin>]
+
+model_opt_class: !name:torch.optim.Adadelta
+    lr: !ref <lr>
+    rho: 0.95
+    eps: 1.e-8
+
+wav2vec_opt_class: !name:torch.optim.Adam
+    lr: !ref <lr_wav2vec>
+
+lr_annealing_model: !new:speechbrain.nnet.schedulers.NewBobScheduler
+    initial_value: !ref <lr>
+    improvement_threshold: 0.0025
+    annealing_factor: 0.8
+    patient: 0
+
+lr_annealing_wav2vec: !new:speechbrain.nnet.schedulers.NewBobScheduler
+    initial_value: !ref <lr_wav2vec>
+    improvement_threshold: 0.0025
+    annealing_factor: 0.9
+    patient: 0
+
+checkpointer: !new:speechbrain.utils.checkpoints.Checkpointer
+    checkpoints_dir: !ref <save_folder>
+    recoverables:
+        wav2vec2: !ref <wav2vec2>
+        model: !ref <model>
+        scheduler_model: !ref <lr_annealing_model>
+        scheduler_wav2vec: !ref <lr_annealing_wav2vec>
+        counter: !ref <epoch_counter>
+
+train_logger: !new:speechbrain.utils.train_logger.FileTrainLogger
+    save_file: !ref <train_log>
+
+error_rate_computer: !name:speechbrain.utils.metric_stats.ErrorRateStats
+
+cer_computer: !name:speechbrain.utils.metric_stats.ErrorRateStats
+    split_tokens: True

train_with_wavlm.py

@@ -0,0 +1,399 @@
+#!/usr/bin/env python3
+import sys
+import torch
+import logging
+import speechbrain as sb
+from pathlib import Path
+import os
+import torchaudio
+from hyperpyyaml import load_hyperpyyaml
+from speechbrain.tokenizers.SentencePiece import SentencePiece
+from speechbrain.utils.data_utils import undo_padding
+from speechbrain.utils.distributed import run_on_main
+
+"""Recipe for training a sequence-to-sequence ASR system with CommonVoice.
+The system employs a wav2vec2 encoder and a CTC decoder.
+Decoding is performed with greedy decoding (will be extended to beam search).
+
+To run this recipe, do the following:
+> python train_with_wav2vec2.py hparams/train_with_wav2vec2.yaml
+
+With the default hyperparameters, the system employs a pretrained wav2vec2 encoder.
+The wav2vec2 model is pretrained following the model given in the hprams file.
+It may be dependent on the language.
+
+The neural network is trained with CTC on sub-word units estimated with
+Byte Pairwise Encoding (BPE).
+
+The experiment file is flexible enough to support a large variety of
+different systems. By properly changing the parameter files, you can try
+different encoders, decoders, tokens (e.g, characters instead of BPE),
+training languages (all CommonVoice languages), and many
+other possible variations.
+
+Authors
+ * Titouan Parcollet 2021
+"""
+
+logger = logging.getLogger(__name__)
+
+
+# Define training procedure
+class ASR(sb.core.Brain):
+    def compute_forward(self, batch, stage):
+        """Forward computations from the waveform batches to the output probabilities."""
+
+        batch = batch.to(self.device)
+        wavs, wav_lens = batch.sig
+        wavs, wav_lens = wavs.to(self.device), wav_lens.to(self.device)
+        if stage == sb.Stage.TRAIN:
+            if hasattr(self.hparams, "augmentation"):
+                wavs = self.hparams.augmentation(wavs, wav_lens)
+
+        # Forward pass
+        feats = self.modules.wav2vec2(wavs, wav_lens)
+        x = self.modules.enc(feats)
+        logits = self.modules.ctc_lin(x)
+        p_ctc = self.hparams.log_softmax(logits)
+
+        return p_ctc, wav_lens
+
+    def compute_objectives(self, predictions, batch, stage):
+        """Computes the loss (CTC) given predictions and targets."""
+
+        p_ctc, wav_lens = predictions
+
+        ids = batch.id
+        tokens, tokens_lens = batch.tokens
+
+        loss = self.hparams.ctc_cost(p_ctc, tokens, wav_lens, tokens_lens)
+
+        if stage != sb.Stage.TRAIN:
+            predicted_tokens = sb.decoders.ctc_greedy_decode(
+                p_ctc, wav_lens, blank_id=self.hparams.blank_index
+            )
+            # Decode token terms to words
+            if self.hparams.use_language_modelling:
+                predicted_words = []
+                for logs in p_ctc:
+                    text = decoder.decode(logs.detach().cpu().numpy())
+                    predicted_words.append(text.split(" "))
+            else:
+                predicted_words = [
+                    "".join(self.tokenizer.decode_ndim(utt_seq)).split(" ")
+                    for utt_seq in predicted_tokens
+                ]
+            # Convert indices to words
+            target_words = [wrd.split(" ") for wrd in batch.wrd]
+
+            self.wer_metric.append(ids, predicted_words, target_words)
+            self.cer_metric.append(ids, predicted_words, target_words)
+
+        return loss
+
+    def fit_batch(self, batch):
+        """Train the parameters given a single batch in input"""
+        should_step = self.step % self.grad_accumulation_factor == 0
+        # Managing automatic mixed precision
+        # TOFIX: CTC fine-tuning currently is unstable
+        # This is certainly due to CTC being done in fp16 instead of fp32
+        if self.auto_mix_prec:
+            with torch.cuda.amp.autocast():
+                with self.no_sync():
+                    outputs = self.compute_forward(batch, sb.Stage.TRAIN)
+                loss = self.compute_objectives(outputs, batch, sb.Stage.TRAIN)
+            with self.no_sync(not should_step):
+                self.scaler.scale(
+                    loss / self.grad_accumulation_factor
+                ).backward()
+            if should_step:
+
+                if not self.hparams.wav2vec2.freeze:
+                    self.scaler.unscale_(self.wav2vec_optimizer)
+                self.scaler.unscale_(self.model_optimizer)
+                if self.check_gradients(loss):
+                    if not self.hparams.wav2vec2.freeze:
+                        if self.optimizer_step >= self.hparams.warmup_steps:
+                            self.scaler.step(self.wav2vec_optimizer)
+                    self.scaler.step(self.model_optimizer)
+                self.scaler.update()
+                self.zero_grad()
+                self.optimizer_step += 1
+        else:
+            # This is mandatory because HF models have a weird behavior with DDP
+            # on the forward pass
+            with self.no_sync():
+                outputs = self.compute_forward(batch, sb.Stage.TRAIN)
+
+            loss = self.compute_objectives(outputs, batch, sb.Stage.TRAIN)
+
+            with self.no_sync(not should_step):
+                (loss / self.grad_accumulation_factor).backward()
+            if should_step:
+                if self.check_gradients(loss):
+                    if not self.hparams.wav2vec2.freeze:
+                        if self.optimizer_step >= self.hparams.warmup_steps:
+                            self.wav2vec_optimizer.step()
+                    self.model_optimizer.step()
+                self.zero_grad()
+                self.optimizer_step += 1
+
+        self.on_fit_batch_end(batch, outputs, loss, should_step)
+        return loss.detach().cpu()
+
+    def evaluate_batch(self, batch, stage):
+        """Computations needed for validation/test batches"""
+        predictions = self.compute_forward(batch, stage=stage)
+        with torch.no_grad():
+            loss = self.compute_objectives(predictions, batch, stage=stage)
+        return loss.detach()
+
+    def on_stage_start(self, stage, epoch):
+        """Gets called at the beginning of each epoch"""
+        if stage != sb.Stage.TRAIN:
+            self.cer_metric = self.hparams.cer_computer()
+            self.wer_metric = self.hparams.error_rate_computer()
+
+    def on_stage_end(self, stage, stage_loss, epoch):
+        """Gets called at the end of an epoch."""
+        # Compute/store important stats
+        stage_stats = {"loss": stage_loss}
+        if stage == sb.Stage.TRAIN:
+            self.train_stats = stage_stats
+        else:
+            stage_stats["CER"] = self.cer_metric.summarize("error_rate")
+            stage_stats["WER"] = self.wer_metric.summarize("error_rate")
+
+        # Perform end-of-iteration things, like annealing, logging, etc.
+        if stage == sb.Stage.VALID:
+            old_lr_model, new_lr_model = self.hparams.lr_annealing_model(
+                stage_stats["loss"]
+            )
+            old_lr_wav2vec, new_lr_wav2vec = self.hparams.lr_annealing_wav2vec(
+                stage_stats["loss"]
+            )
+            sb.nnet.schedulers.update_learning_rate(
+                self.model_optimizer, new_lr_model
+            )
+            if not self.hparams.wav2vec2.freeze:
+                sb.nnet.schedulers.update_learning_rate(
+                    self.wav2vec_optimizer, new_lr_wav2vec
+                )
+            self.hparams.train_logger.log_stats(
+                stats_meta={
+                    "epoch": epoch,
+                    "lr_model": old_lr_model,
+                    "lr_wav2vec": old_lr_wav2vec,
+                },
+                train_stats=self.train_stats,
+                valid_stats=stage_stats,
+            )
+            self.checkpointer.save_and_keep_only(
+                meta={"WER": stage_stats["WER"]}, min_keys=["WER"],
+            )
+        elif stage == sb.Stage.TEST:
+            self.hparams.train_logger.log_stats(
+                stats_meta={"Epoch loaded": self.hparams.epoch_counter.current},
+                test_stats=stage_stats,
+            )
+            with open(self.hparams.wer_file, "w") as w:
+                self.wer_metric.write_stats(w)
+
+    def init_optimizers(self):
+        "Initializes the wav2vec2 optimizer and model optimizer"
+
+        # If the wav2vec encoder is unfrozen, we create the optimizer
+        if not self.hparams.wav2vec2.freeze:
+            self.wav2vec_optimizer = self.hparams.wav2vec_opt_class(
+                self.modules.wav2vec2.parameters()
+            )
+            if self.checkpointer is not None:
+                self.checkpointer.add_recoverable(
+                    "wav2vec_opt", self.wav2vec_optimizer
+                )
+
+        self.model_optimizer = self.hparams.model_opt_class(
+            self.hparams.model.parameters()
+        )
+
+        if self.checkpointer is not None:
+            self.checkpointer.add_recoverable("modelopt", self.model_optimizer)
+
+    def zero_grad(self, set_to_none=False):
+        if not self.hparams.wav2vec2.freeze:
+            self.wav2vec_optimizer.zero_grad(set_to_none)
+        self.model_optimizer.zero_grad(set_to_none)
+
+
+# Define custom data procedure
+def dataio_prepare(hparams):
+    """This function prepares the datasets to be used in the brain class.
+    It also defines the data processing pipeline through user-defined functions."""
+
+    # 1. Define datasets
+    data_folder = hparams["data_folder"]
+
+    train_data = sb.dataio.dataset.DynamicItemDataset.from_csv(
+        csv_path=hparams["train_csv"], replacements={"data_root": data_folder},
+    )
+
+    if hparams["sorting"] == "ascending":
+        # we sort training data to speed up training and get better results.
+        train_data = train_data.filtered_sorted(
+            sort_key="duration",
+            key_max_value={"duration": hparams["avoid_if_longer_than"]},
+        )
+        # when sorting do not shuffle in dataloader ! otherwise is pointless
+        hparams["dataloader_options"]["shuffle"] = False
+
+    elif hparams["sorting"] == "descending":
+        train_data = train_data.filtered_sorted(
+            sort_key="duration",
+            reverse=True,
+            key_max_value={"duration": hparams["avoid_if_longer_than"]},
+        )
+        # when sorting do not shuffle in dataloader ! otherwise is pointless
+        hparams["dataloader_options"]["shuffle"] = False
+
+    elif hparams["sorting"] == "random":
+        pass
+
+    else:
+        raise NotImplementedError(
+            "sorting must be random, ascending or descending"
+        )
+
+    valid_data = sb.dataio.dataset.DynamicItemDataset.from_csv(
+        csv_path=hparams["valid_csv"], replacements={"data_root": data_folder},
+    )
+    # We also sort the validation data so it is faster to validate
+    valid_data = valid_data.filtered_sorted(sort_key="duration")
+    test_datasets = {}
+    for csv_file in hparams["test_csv"]:
+        name = Path(csv_file).stem
+        test_datasets[name] = sb.dataio.dataset.DynamicItemDataset.from_csv(
+            csv_path=csv_file, replacements={"data_root": data_folder}
+        )
+        test_datasets[name] = test_datasets[name].filtered_sorted(
+            sort_key="duration"
+        )
+
+    datasets = [train_data, valid_data] + [i for k, i in test_datasets.items()]
+
+
+    # 2. Define audio pipeline:
+    @sb.utils.data_pipeline.takes("wav")
+    @sb.utils.data_pipeline.provides("sig")
+    def audio_pipeline(wav):
+        info = torchaudio.info(wav)
+        sig = sb.dataio.dataio.read_audio(wav)
+        resampled = torchaudio.transforms.Resample(
+            info.sample_rate, hparams["sample_rate"],
+        )(sig)
+        return resampled
+
+    sb.dataio.dataset.add_dynamic_item(datasets, audio_pipeline)
+    label_encoder = sb.dataio.encoder.CTCTextEncoder()
+
+    # 3. Define text pipeline:
+    @sb.utils.data_pipeline.takes("wrd")
+    @sb.utils.data_pipeline.provides(
+        "wrd", "char_list", "tokens_list", "tokens"
+    )
+    def text_pipeline(wrd):
+        yield wrd
+        char_list = list(wrd)
+        yield char_list
+        tokens_list = label_encoder.encode_sequence(char_list)
+        yield tokens_list
+        tokens = torch.LongTensor(tokens_list)
+        yield tokens
+
+    sb.dataio.dataset.add_dynamic_item(datasets, text_pipeline)
+    lab_enc_file = os.path.join(hparams["save_folder"], "label_encoder.txt")
+    special_labels = {
+        "blank_label": hparams["blank_index"],
+        "unk_label": hparams["unk_index"]
+    }
+    label_encoder.load_or_create(
+        path=lab_enc_file,
+        from_didatasets=[train_data],
+        output_key="char_list",
+        special_labels=special_labels,
+        sequence_input=True,
+    )
+
+    # 4. Set output:
+    sb.dataio.dataset.set_output_keys(
+        datasets, ["id", "sig", "wrd", "char_list", "tokens"],
+    )
+    return train_data, valid_data,test_datasets, label_encoder
+
+
+if __name__ == "__main__":
+
+    # Load hyperparameters file with command-line overrides
+    hparams_file, run_opts, overrides = sb.parse_arguments(sys.argv[1:])
+    with open(hparams_file) as fin:
+        hparams = load_hyperpyyaml(fin, overrides)
+
+    # If --distributed_launch then
+    # create ddp_group with the right communication protocol
+    sb.utils.distributed.ddp_init_group(run_opts)
+
+
+    # Create experiment directory
+    sb.create_experiment_directory(
+        experiment_directory=hparams["output_folder"],
+        hyperparams_to_save=hparams_file,
+        overrides=overrides,
+    )
+
+    # Due to DDP, we do the preparation ONLY on the main python process
+    # Defining tokenizer and loading it
+    # Create the datasets objects as well as tokenization and encoding :-D
+    train_data, valid_data, test_datasets, label_encoder = dataio_prepare(hparams)
+    if hparams["use_language_modelling"]:
+        print("using langauge_modeeling")
+        from pyctcdecode import build_ctcdecoder
+        ind2lab = label_encoder.ind2lab
+        print(ind2lab)
+        labels = [ind2lab[x] for x in range(len(ind2lab))]
+        labels = [""] + labels[1:-1] + ["1"] 
+        # Replace the <blank> token with a blank character, needed for PyCTCdecode
+        print(labels)
+        decoder = build_ctcdecoder(
+            labels,
+            kenlm_model_path=hparams["ngram_lm_path"],  # .arpa or .bin
+            alpha=0.5,  # Default by KenLM
+            beta=1.0,  # Default by KenLM
+        )
+    # Trainer initialization
+    asr_brain = ASR(
+        modules=hparams["modules"],
+        hparams=hparams,
+        run_opts=run_opts,
+        checkpointer=hparams["checkpointer"],
+    )
+
+    # Adding objects to trainer.
+    asr_brain.tokenizer = label_encoder
+
+    # Training
+    asr_brain.fit(
+        asr_brain.hparams.epoch_counter,
+        train_data,
+        valid_data,
+        train_loader_kwargs=hparams["dataloader_options"],
+        valid_loader_kwargs=hparams["test_dataloader_options"],
+    )
+
+    # Test
+    for k in test_datasets.keys():  # keys are test_clean, test_other etc
+        asr_brain.hparams.wer_file = os.path.join(
+            hparams["output_folder"], "wer_{}.txt".format(k)
+        )
+        asr_brain.evaluate(
+            test_datasets[k], test_loader_kwargs=hparams["test_dataloader_options"]
+        )
+

wavlm-large/README.md

@@ -0,0 +1,64 @@
+---
+language:
+- en
+tags:
+- speech
+inference: false
+---
+
+# WavLM-Large
+
+[Microsoft's WavLM](https://github.com/microsoft/unilm/tree/master/wavlm)
+
+The large model pretrained on 16kHz sampled speech audio. When using the model, make sure that your speech input is also sampled at 16kHz. 
+
+**Note**: This model does not have a tokenizer as it was pretrained on audio alone. In order to use this model **speech recognition**, a tokenizer should be created and the model should be fine-tuned on labeled text data. Check out [this blog](https://huggingface.co/blog/fine-tune-wav2vec2-english) for more in-detail explanation of how to fine-tune the model.
+
+The model was pre-trained on:
+
+- 60,000 hours of [Libri-Light](https://arxiv.org/abs/1912.07875)
+- 10,000 hours of [GigaSpeech](https://arxiv.org/abs/2106.06909)
+- 24,000 hours of [VoxPopuli](https://arxiv.org/abs/2101.00390)
+
+[Paper: WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://arxiv.org/abs/2110.13900)
+
+Authors: Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, Jian Wu, Long Zhou, Shuo Ren, Yanmin Qian, Yao Qian, Jian Wu, Michael Zeng, Furu Wei
+
+**Abstract**
+*Self-supervised learning (SSL) achieves great success in speech recognition, while limited exploration has been attempted for other speech processing tasks. As speech signal contains multi-faceted information including speaker identity, paralinguistics, spoken content, etc., learning universal representations for all speech tasks is challenging. In this paper, we propose a new pre-trained model, WavLM, to solve full-stack downstream speech tasks. WavLM is built based on the HuBERT framework, with an emphasis on both spoken content modeling and speaker identity preservation. We first equip the Transformer structure with gated relative position bias to improve its capability on recognition tasks. For better speaker discrimination, we propose an utterance mixing training strategy, where additional overlapped utterances are created unsupervisely and incorporated during model training. Lastly, we scale up the training dataset from 60k hours to 94k hours. WavLM Large achieves state-of-the-art performance on the SUPERB benchmark, and brings significant improvements for various speech processing tasks on their representative benchmarks.*
+
+The original model can be found under https://github.com/microsoft/unilm/tree/master/wavlm.
+
+# Usage
+
+This is an English pre-trained speech model that has to be fine-tuned on a downstream task like speech recognition or audio classification before it can be 
+used in inference. The model was pre-trained in English and should therefore perform well only in English. The model has been shown to work well on the [SUPERB benchmark](https://superbbenchmark.org/).
+
+**Note**: The model was pre-trained on phonemes rather than characters. This means that one should make sure that the input text is converted to a sequence 
+of phonemes before fine-tuning.
+
+## Speech Recognition
+
+To fine-tune the model for speech recognition, see [the official speech recognition example](https://github.com/huggingface/transformers/tree/master/examples/pytorch/speech-recognition).
+
+## Speech Classification
+
+To fine-tune the model for speech classification, see [the official audio classification example](https://github.com/huggingface/transformers/tree/master/examples/pytorch/audio-classification).
+
+## Speaker Verification
+
+TODO
+
+## Speaker Diarization
+
+TODO
+
+# Contribution
+
+The model was contributed by [cywang](https://huggingface.co/cywang) and [patrickvonplaten](https://huggingface.co/patrickvonplaten).
+
+# License
+
+The official license can be found [here](https://github.com/microsoft/UniSpeech/blob/main/LICENSE)
+
+![design](https://raw.githubusercontent.com/patrickvonplaten/scientific_images/master/wavlm.png)

wavlm-large/config.json

@@ -0,0 +1,99 @@
+{
+  "_name_or_path": "./wavlm-large/",
+  "activation_dropout": 0.0,
+  "adapter_kernel_size": 3,
+  "adapter_stride": 2,
+  "add_adapter": false,
+  "apply_spec_augment": true,
+  "architectures": [
+    "WavLMModel"
+  ],
+  "attention_dropout": 0.1,
+  "bos_token_id": 1,
+  "classifier_proj_size": 256,
+  "codevector_dim": 768,
+  "contrastive_logits_temperature": 0.1,
+  "conv_bias": false,
+  "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": "sum",
+  "ctc_zero_infinity": false,
+  "diversity_loss_weight": 0.1,
+  "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": 0.1,
+  "feat_quantizer_dropout": 0.0,
+  "final_dropout": 0.0,
+  "gradient_checkpointing": false,
+  "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_min_masks": 0,
+  "mask_feature_prob": 0.0,
+  "mask_time_length": 10,
+  "mask_time_min_masks": 2,
+  "mask_time_min_space": 1,
+  "mask_time_other": 0.0,
+  "mask_time_prob": 0.075,
+  "mask_time_selection": "static",
+  "max_bucket_distance": 800,
+  "model_type": "wavlm",
+  "num_adapter_layers": 3,
+  "num_attention_heads": 16,
+  "num_buckets": 320,
+  "num_codevector_groups": 2,
+  "num_codevectors_per_group": 320,
+  "num_conv_pos_embedding_groups": 16,
+  "num_conv_pos_embeddings": 128,
+  "num_ctc_classes": 80,
+  "num_feat_extract_layers": 7,
+  "num_hidden_layers": 24,
+  "num_negatives": 100,
+  "output_hidden_size": 1024,
+  "pad_token_id": 0,
+  "proj_codevector_dim": 768,
+  "replace_prob": 0.5,
+  "tokenizer_class": "Wav2Vec2CTCTokenizer",
+  "torch_dtype": "float32",
+  "transformers_version": "4.15.0.dev0",
+  "use_weighted_layer_sum": false,
+  "vocab_size": 32
+}

wavlm-large/preprocessor_config.json

@@ -0,0 +1,9 @@
+{
+  "do_normalize": true,
+  "feature_extractor_type": "Wav2Vec2FeatureExtractor",
+  "feature_size": 1,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "return_attention_mask": true,
+  "sampling_rate": 16000
+}

wavlm-large/pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fdee460e529396ddb2f8c8e8ce0ad74cfb747b726bc6f612e666c7c1e1963c9d
+size 1261990257