First model version

.gitignore

@@ -0,0 +1 @@
+checkpoint-*/

added_tokens.json

@@ -0,0 +1 @@
+{"<s>": 44, "</s>": 45}

alphabet.json

@@ -0,0 +1 @@
+{"labels": [" ", "#", "'", "-", "a", "b", "c", "d", "e", "f", "g", "h", "i", "j", "k", "l", "m", "n", "o", "p", "q", "r", "s", "t", "u", "v", "w", "x", "y", "z", "\u00b7", "\u00e0", "\u00e7", "\u00e8", "\u00e9", "\u00ed", "\u00ef", "\u00f2", "\u00f3", "\u00fa", "\u00fc", "\u0903", "\u2047", "", "<s>", "</s>"], "is_bpe": false}

config.json

@@ -0,0 +1,107 @@
+{
+  "_name_or_path": "facebook/wav2vec2-xls-r-300m",
+  "activation_dropout": 0.1,
+  "adapter_kernel_size": 3,
+  "adapter_stride": 2,
+  "add_adapter": false,
+  "apply_spec_augment": true,
+  "architectures": [
+    "Wav2Vec2ForCTC"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "classifier_proj_size": 256,
+  "codevector_dim": 768,
+  "contrastive_logits_temperature": 0.1,
+  "conv_bias": true,
+  "conv_dim": [
+    512,
+    512,
+    512,
+    512,
+    512,
+    512,
+    512
+  ],
+  "conv_kernel": [
+    10,
+    3,
+    3,
+    3,
+    3,
+    2,
+    2
+  ],
+  "conv_stride": [
+    5,
+    2,
+    2,
+    2,
+    2,
+    2,
+    2
+  ],
+  "ctc_loss_reduction": "mean",
+  "ctc_zero_infinity": false,
+  "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": 0.0,
+  "feat_quantizer_dropout": 0.0,
+  "final_dropout": 0.0,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0.0,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-05,
+  "layerdrop": 0.0,
+  "mask_feature_length": 64,
+  "mask_feature_min_masks": 0,
+  "mask_feature_prob": 0.25,
+  "mask_time_length": 10,
+  "mask_time_min_masks": 2,
+  "mask_time_prob": 0.75,
+  "model_type": "wav2vec2",
+  "num_adapter_layers": 3,
+  "num_attention_heads": 16,
+  "num_codevector_groups": 2,
+  "num_codevectors_per_group": 320,
+  "num_conv_pos_embedding_groups": 16,
+  "num_conv_pos_embeddings": 128,
+  "num_feat_extract_layers": 7,
+  "num_hidden_layers": 24,
+  "num_negatives": 100,
+  "output_hidden_size": 1024,
+  "pad_token_id": 43,
+  "proj_codevector_dim": 768,
+  "tdnn_dilation": [
+    1,
+    2,
+    3,
+    1,
+    1
+  ],
+  "tdnn_dim": [
+    512,
+    512,
+    512,
+    512,
+    1500
+  ],
+  "tdnn_kernel": [
+    5,
+    3,
+    3,
+    1,
+    1
+  ],
+  "torch_dtype": "float32",
+  "transformers_version": "4.16.0.dev0",
+  "use_weighted_layer_sum": false,
+  "vocab_size": 46,
+  "xvector_output_dim": 512
+}

language_model/5gram.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af7f2cbb75fbcb5091a6a24c9b6984b81a0c257652ca96f7b434f5b6e5224740
+size 630272442

language_model/attrs.json

@@ -0,0 +1 @@
+{"alpha": 0.5, "beta": 1.5, "unk_score_offset": -10.0, "score_boundary": true}

preprocessor_config.json

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

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8266cc58fff449061f92483f6f7afebb0e00e0f6bba42881c5d847151579245c
+size 1262112241

run.sh

@@ -0,0 +1,39 @@
+python run_speech_recognition_ctc.py \
+--dataset_name "mozilla-foundation/common_voice_8_0" "collectivat/tv3_parla" "projecte-aina/parlament_parla" \
+--dataset_config_name "ca" "ca" "clean" \
+--model_name_or_path="facebook/wav2vec2-xls-r-300m" \
+--train_split_name "train+validation" "train" "train+validation" \
+--eval_split_name "test" "test" "test" \
+--audio_column_name "audio" "audio" "audio" \
+--output_dir="wav2vec2-xls-r-300m-ca" \
+--overwrite_output_dir \
+--num_train_epochs="10" \
+--per_device_train_batch_size="32" \
+--per_device_eval_batch_size="32" \
+--gradient_accumulation_steps="4" \
+--learning_rate="7.5e-5" \
+--warmup_steps="2000" \
+--length_column_name="input_length" \
+--evaluation_strategy="steps" \
+--text_column_name "sentence" "text" "sentence" \
+--chars_to_ignore [ , ? . ! \; \: \" “ % ” � — … – ] \
+--save_steps="500" \
+--eval_steps="500" \
+--logging_steps="500" \
+--layerdrop="0.0" \
+--activation_dropout="0.1" \
+--save_total_limit="3" \
+--freeze_feature_encoder \
+--feat_proj_dropout="0.0" \
+--mask_time_prob="0.75" \
+--preprocessing_num_workers="12" \
+--mask_time_length="10" \
+--mask_feature_prob="0.25" \
+--mask_feature_length="64" \
+--gradient_checkpointing \
+--use_auth_token \
+--fp16 \
+--group_by_length \
+--do_train --do_eval \
+--push_to_hub 
+#&> train.log

run_speech_recognition_ctc.py

@@ -0,0 +1,814 @@
+#!/usr/bin/env python
+# coding=utf-8
+# Copyright 2021 The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+
+""" Fine-tuning a 🤗 Transformers CTC model for automatic speech recognition"""
+
+import functools
+import json
+import logging
+import os
+import re
+import sys
+import warnings
+from dataclasses import dataclass, field
+from typing import Dict, List, Optional, Union
+from text.numbers_ca import normalize_numbers_ca
+
+import datasets
+import numpy as np
+import torch
+from datasets import DatasetDict, load_dataset, load_metric, concatenate_datasets
+
+import transformers
+from transformers import (
+    AutoConfig,
+    AutoFeatureExtractor,
+    AutoModelForCTC,
+    AutoProcessor,
+    AutoTokenizer,
+    HfArgumentParser,
+    Trainer,
+    TrainingArguments,
+    Wav2Vec2Processor,
+    set_seed,
+)
+from transformers.trainer_utils import get_last_checkpoint, is_main_process
+from transformers.utils import check_min_version
+from transformers.utils.versions import require_version
+
+
+# Will error if the minimal version of Transformers is not installed. Remove at your own risks.
+check_min_version("4.16.0.dev0")
+
+require_version("datasets>=1.13.3", "To fix: pip install -r examples/pytorch/text-classification/requirements.txt")
+
+
+logger = logging.getLogger(__name__)
+
+
+def list_field(default=None, metadata=None):
+    return field(default_factory=lambda: default, metadata=metadata)
+
+
+@dataclass
+class ModelArguments:
+    """
+    Arguments pertaining to which model/config/tokenizer we are going to fine-tune from.
+    """
+
+    model_name_or_path: str = field(
+        metadata={"help": "Path to pretrained model or model identifier from huggingface.co/models"}
+    )
+    tokenizer_name_or_path: Optional[str] = field(
+        default=None,
+        metadata={"help": "Path to pretrained tokenizer or tokenizer identifier from huggingface.co/models"},
+    )
+    cache_dir: Optional[str] = field(
+        default=None,
+        metadata={"help": "Where do you want to store the pretrained models downloaded from huggingface.co"},
+    )
+    freeze_feature_encoder: bool = field(
+        default=True, metadata={"help": "Whether to freeze the feature encoder layers of the model."}
+    )
+    attention_dropout: float = field(
+        default=0.0, metadata={"help": "The dropout ratio for the attention probabilities."}
+    )
+    activation_dropout: float = field(
+        default=0.0, metadata={"help": "The dropout ratio for activations inside the fully connected layer."}
+    )
+    feat_proj_dropout: float = field(default=0.0, metadata={"help": "The dropout ratio for the projected features."})
+    hidden_dropout: float = field(
+        default=0.0,
+        metadata={
+            "help": "The dropout probability for all fully connected layers in the embeddings, encoder, and pooler."
+        },
+    )
+    final_dropout: float = field(
+        default=0.0,
+        metadata={"help": "The dropout probability for the final projection layer."},
+    )
+    mask_time_prob: float = field(
+        default=0.05,
+        metadata={
+            "help": "Probability of each feature vector along the time axis to be chosen as the start of the vector"
+            "span to be masked. Approximately ``mask_time_prob * sequence_length // mask_time_length`` feature"
+            "vectors will be masked along the time axis."
+        },
+    )
+    mask_time_length: int = field(
+        default=10,
+        metadata={"help": "Length of vector span to mask along the time axis."},
+    )
+    mask_feature_prob: float = field(
+        default=0.0,
+        metadata={
+            "help": "Probability of each feature vector along the feature axis to be chosen as the start of the vector"
+            "span to be masked. Approximately ``mask_feature_prob * sequence_length // mask_feature_length`` feature bins will be masked along the time axis."
+        },
+    )
+    mask_feature_length: int = field(
+        default=10,
+        metadata={"help": "Length of vector span to mask along the feature axis."},
+    )
+    layerdrop: float = field(default=0.0, metadata={"help": "The LayerDrop probability."})
+    ctc_loss_reduction: Optional[str] = field(
+        default="mean", metadata={"help": "The way the ctc loss should be reduced. Should be one of 'mean' or 'sum'."}
+    )
+
+
+@dataclass
+class DataTrainingArguments:
+    """
+    Arguments pertaining to what data we are going to input our model for training and eval.
+
+    Using `HfArgumentParser` we can turn this class
+    into argparse arguments to be able to specify them on
+    the command line.
+    """
+
+    dataset_name: List[str] = field(
+        metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
+    )
+    dataset_config_name: List[str] = list_field(
+        default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
+    )
+    train_split_name: List[str] = list_field(
+        default=["train+validation"],
+        metadata={
+            "help": "The name of the training data set split to use (via the datasets library). Defaults to 'train'"
+        },
+    )
+    eval_split_name: List[str] = list_field(
+        default=["test"],
+        metadata={
+            "help": "The name of the training data set split to use (via the datasets library). Defaults to 'train'"
+        },
+    )
+    audio_column_name: List[str] = list_field(
+        default=["audio"],
+        metadata={"help": "The name of the dataset column containing the audio data. Defaults to 'audio'"},
+    )
+    text_column_name: List[str] = list_field(
+        default=["text"],
+        metadata={"help": "The name of the dataset column containing the text data. Defaults to 'text'"},
+    )
+    overwrite_cache: bool = field(
+        default=False, metadata={"help": "Overwrite the cached preprocessed datasets or not."}
+    )
+    preprocessing_num_workers: Optional[int] = field(
+        default=None,
+        metadata={"help": "The number of processes to use for the preprocessing."},
+    )
+    max_train_samples: Optional[int] = field(
+        default=None,
+        metadata={
+            "help": "For debugging purposes or quicker training, truncate the number of training examples to this "
+            "value if set."
+        },
+    )
+    max_eval_samples: Optional[int] = field(
+        default=None,
+        metadata={
+            "help": "For debugging purposes or quicker training, truncate the number of validation examples to this "
+            "value if set."
+        },
+    )
+    chars_to_ignore: Optional[List[str]] = list_field(
+        default=None,
+        metadata={"help": "A list of characters to remove from the transcripts."},
+    )
+    eval_metrics: List[str] = list_field(
+        default=["wer"],
+        metadata={"help": "A list of metrics the model should be evaluated on. E.g. `'wer cer'`"},
+    )
+    max_duration_in_seconds: float = field(
+        default=20.0,
+        metadata={
+            "help": "Filter audio files that are longer than `max_duration_in_seconds` seconds to 'max_duration_in_seconds`"
+        },
+    )
+    min_duration_in_seconds: float = field(
+        default=0.0, metadata={"help": "Filter audio files that are shorter than `min_duration_in_seconds` seconds"}
+    )
+    preprocessing_only: bool = field(
+        default=False,
+        metadata={
+            "help": "Whether to only do data preprocessing and skip training. "
+            "This is especially useful when data preprocessing errors out in distributed training due to timeout. "
+            "In this case, one should run the preprocessing in a non-distributed setup with `preprocessing_only=True` "
+            "so that the cached datasets can consequently be loaded in distributed training"
+        },
+    )
+    use_auth_token: bool = field(
+        default=False,
+        metadata={
+            "help": "If :obj:`True`, will use the token generated when running"
+            ":obj:`transformers-cli login` as HTTP bearer authorization for remote files."
+        },
+    )
+    unk_token: str = field(
+        default="[UNK]",
+        metadata={"help": "The unk token for the tokenizer"},
+    )
+    pad_token: str = field(
+        default="[PAD]",
+        metadata={"help": "The padding token for the tokenizer"},
+    )
+    word_delimiter_token: str = field(
+        default="|",
+        metadata={"help": "The word delimiter token for the tokenizer"},
+    )
+    phoneme_language: Optional[str] = field(
+        default=None,
+        metadata={
+            "help": "The target language that should be used be"
+            " passed to the tokenizer for tokenization. Note that"
+            " this is only relevant if the model classifies the"
+            " input audio to a sequence of phoneme sequences."
+        },
+    )
+
+
+@dataclass
+class DataCollatorCTCWithPadding:
+    """
+    Data collator that will dynamically pad the inputs received.
+    Args:
+        processor (:class:`~transformers.AutoProcessor`)
+            The processor used for proccessing the data.
+        padding (:obj:`bool`, :obj:`str` or :class:`~transformers.tokenization_utils_base.PaddingStrategy`, `optional`, defaults to :obj:`True`):
+            Select a strategy to pad the returned sequences (according to the model's padding side and padding index)
+            among:
+            * :obj:`True` or :obj:`'longest'`: Pad to the longest sequence in the batch (or no padding if only a single
+              sequence if provided).
+            * :obj:`'max_length'`: Pad to a maximum length specified with the argument :obj:`max_length` or to the
+              maximum acceptable input length for the model if that argument is not provided.
+            * :obj:`False` or :obj:`'do_not_pad'` (default): No padding (i.e., can output a batch with sequences of
+              different lengths).
+        max_length (:obj:`int`, `optional`):
+            Maximum length of the ``input_values`` of the returned list and optionally padding length (see above).
+        max_length_labels (:obj:`int`, `optional`):
+            Maximum length of the ``labels`` returned list and optionally padding length (see above).
+        pad_to_multiple_of (:obj:`int`, `optional`):
+            If set will pad the sequence to a multiple of the provided value.
+            This is especially useful to enable the use of Tensor Cores on NVIDIA hardware with compute capability >=
+            7.5 (Volta).
+    """
+
+    processor: AutoProcessor
+    padding: Union[bool, str] = "longest"
+    pad_to_multiple_of: Optional[int] = None
+    pad_to_multiple_of_labels: Optional[int] = None
+
+    def __call__(self, features: List[Dict[str, Union[List[int], torch.Tensor]]]) -> Dict[str, torch.Tensor]:
+        # split inputs and labels since they have to be of different lenghts and need
+        # different padding methods
+        input_features = [{"input_values": feature["input_values"]} for feature in features]
+        label_features = [{"input_ids": feature["labels"]} for feature in features]
+
+        batch = self.processor.pad(
+            input_features,
+            padding=self.padding,
+            pad_to_multiple_of=self.pad_to_multiple_of,
+            return_tensors="pt",
+        )
+
+        with self.processor.as_target_processor():
+            labels_batch = self.processor.pad(
+                label_features,
+                padding=self.padding,
+                pad_to_multiple_of=self.pad_to_multiple_of_labels,
+                return_tensors="pt",
+            )
+
+        # replace padding with -100 to ignore loss correctly
+        labels = labels_batch["input_ids"].masked_fill(labels_batch.attention_mask.ne(1), -100)
+
+        batch["labels"] = labels
+
+        return batch
+
+
+def create_vocabulary_from_data(
+    datasets: DatasetDict,
+    word_delimiter_token: Optional[str] = None,
+    unk_token: Optional[str] = None,
+    pad_token: Optional[str] = None,
+):
+    # Given training and test labels create vocabulary
+    def extract_all_chars(batch):
+        all_text = " ".join(batch["target_text"])
+        vocab = list(set(all_text))
+        return {"vocab": [vocab], "all_text": [all_text]}
+
+    vocab_1 = set()
+    for string in datasets["train"]["target_text"]:
+        vocab_1.update(string.lower())
+
+    vocab_2 = set()
+    for string in datasets["eval"]["target_text"]:
+        vocab_2.update(string.lower())
+
+    # vocabs = datasets.map(
+    #     extract_all_chars,
+    #     batched=True,
+    #     batch_size=-1,
+    #     keep_in_memory=True,
+    #     remove_columns=datasets["train"].column_names,
+    #     desc="extract characters"
+    # )
+
+    # take union of all unique characters in each dataset
+    vocab_set = functools.reduce(
+        lambda vocab_1, vocab_2: vocab_1 | vocab_2, [vocab_1, vocab_2]
+    )
+
+    vocab_dict = {v: k for k, v in enumerate(sorted(list(vocab_set)))}
+
+    # replace white space with delimiter token
+    if word_delimiter_token is not None:
+        vocab_dict[word_delimiter_token] = vocab_dict[" "]
+        del vocab_dict[" "]
+
+    # add unk and pad token
+    if unk_token is not None:
+        vocab_dict[unk_token] = len(vocab_dict)
+
+    if pad_token is not None:
+        vocab_dict[pad_token] = len(vocab_dict)
+
+    return vocab_dict
+
+
+def main():
+    # See all possible arguments in src/transformers/training_args.py
+    # or by passing the --help flag to this script.
+    # We now keep distinct sets of args, for a cleaner separation of concerns.
+
+    parser = HfArgumentParser((ModelArguments, DataTrainingArguments, TrainingArguments))
+    if len(sys.argv) == 2 and sys.argv[1].endswith(".json"):
+        # If we pass only one argument to the script and it's the path to a json file,
+        # let's parse it to get our arguments.
+        model_args, data_args, training_args = parser.parse_json_file(json_file=os.path.abspath(sys.argv[1]))
+    else:
+        model_args, data_args, training_args = parser.parse_args_into_dataclasses()
+
+    # Detecting last checkpoint.
+    last_checkpoint = None
+    if os.path.isdir(training_args.output_dir) and training_args.do_train and not training_args.overwrite_output_dir:
+        last_checkpoint = get_last_checkpoint(training_args.output_dir)
+        if last_checkpoint is None and len(os.listdir(training_args.output_dir)) > 0:
+            raise ValueError(
+                f"Output directory ({training_args.output_dir}) already exists and is not empty. "
+                "Use --overwrite_output_dir to overcome."
+            )
+        elif last_checkpoint is not None:
+            logger.info(
+                f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
+                "the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
+            )
+
+    # Setup logging
+    logging.basicConfig(
+        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
+        datefmt="%m/%d/%Y %H:%M:%S",
+        handlers=[logging.StreamHandler(sys.stdout)],
+    )
+    logger.setLevel(logging.INFO if is_main_process(training_args.local_rank) else logging.WARN)
+
+    # Log on each process the small summary:
+    logger.warning(
+        f"Process rank: {training_args.local_rank}, device: {training_args.device}, n_gpu: {training_args.n_gpu}"
+        f"distributed training: {bool(training_args.local_rank != -1)}, 16-bits training: {training_args.fp16}"
+    )
+    # Set the verbosity to info of the Transformers logger (on main process only):
+    if is_main_process(training_args.local_rank):
+        transformers.utils.logging.set_verbosity_info()
+    logger.info("Training/evaluation parameters %s", training_args)
+
+    # Set seed before initializing model.
+    set_seed(training_args.seed)
+
+    # 1. First, let's load the dataset
+    raw_datasets = DatasetDict()
+    train_datasets = []
+    eval_datasests = []
+    if training_args.do_train:
+        print(data_args.dataset_name,data_args.dataset_config_name, data_args.train_split_name, data_args.audio_column_name, data_args.text_column_name)
+        for dataset_name, dataset_config_name, train_split_name, audio_column_name, text_column_name in zip(data_args.dataset_name, data_args.dataset_config_name, data_args.train_split_name, data_args.audio_column_name, data_args.text_column_name):
+            raw_datasets["train"] = load_dataset(
+                dataset_name,
+                dataset_config_name,
+                split=train_split_name,
+                use_auth_token=data_args.use_auth_token,
+                data_dir="datasets", 
+                cache_dir="datasets"
+            )
+
+            if audio_column_name not in raw_datasets["train"].column_names:
+                raise ValueError(
+                    f"--audio_column_name '{audio_column_name}' not found in dataset '{dataset_name}'. "
+                    "Make sure to set `--audio_column_name` to the correct audio column - one of "
+                    f"{', '.join(raw_datasets['train'].column_names)}."
+                )
+
+            if text_column_name not in raw_datasets["train"].column_names:
+                raise ValueError(
+                    f"--text_column_name {text_column_name} not found in dataset '{dataset_name}'. "
+                    "Make sure to set `--text_column_name` to the correct text column - one of "
+                    f"{', '.join(raw_datasets['train'].column_names)}."
+                )
+
+            if data_args.max_train_samples is not None:
+                raw_datasets["train"] = raw_datasets["train"].select(range(data_args.max_train_samples))
+            if text_column_name != "text":
+                raw_datasets["train"] = raw_datasets["train"].rename_column(text_column_name, "text")
+            if audio_column_name != "audio":
+                raw_datasets["train"] = raw_datasets["train"].rename_column(audio_column_name, "audio")
+            raw_datasets["train"] = raw_datasets["train"].remove_columns([column for column in raw_datasets["train"].column_names if column not in ["text", "audio"]])
+            train_datasets.append(raw_datasets["train"])
+        raw_datasets["train"] = concatenate_datasets(train_datasets)
+
+    if training_args.do_eval:
+        for dataset_name, dataset_config_name, eval_split_name, audio_column_name, text_column_name in zip(data_args.dataset_name, data_args.dataset_config_name, data_args.eval_split_name, data_args.audio_column_name, data_args.text_column_name):
+            raw_datasets["eval"] = load_dataset(
+                dataset_name,
+                dataset_config_name,
+                split=eval_split_name,
+                use_auth_token=data_args.use_auth_token,
+                data_dir="datasets", 
+                cache_dir="datasets"
+            )
+
+            if audio_column_name not in raw_datasets["eval"].column_names:
+                raise ValueError(
+                    f"--audio_column_name '{audio_column_name}' not found in dataset '{dataset_name}'. "
+                    "Make sure to set `--audio_column_name` to the correct audio column - one of "
+                    f"{', '.join(raw_datasets['eval'].column_names)}."
+                )
+
+            if text_column_name not in raw_datasets["eval"].column_names:
+                raise ValueError(
+                    f"--text_column_name {text_column_name} not found in dataset '{dataset_name}'. "
+                    "Make sure to set `--text_column_name` to the correct text column - one of "
+                    f"{', '.join(raw_datasets['eval'].column_names)}."
+                )
+
+            if data_args.max_eval_samples is not None:
+                raw_datasets["eval"] = raw_datasets["eval"].select(range(data_args.max_eval_samples))
+            if text_column_name != "text":
+                raw_datasets["eval"] = raw_datasets["eval"].rename_column(text_column_name, "text")
+            if audio_column_name != "audio":
+                raw_datasets["eval"] = raw_datasets["eval"].rename_column(audio_column_name, "audio")
+            raw_datasets["eval"] = raw_datasets["eval"].remove_columns([column for column in raw_datasets["eval"].column_names if column not in ["text", "audio"]])
+            eval_datasests.append(raw_datasets["eval"])
+        raw_datasets["eval"] = concatenate_datasets(eval_datasests)
+
+    # 2. We remove some special characters from the datasets
+    # that make training complicated and do not help in transcribing the speech
+    # E.g. characters, such as `,` and `.` do not really have an acoustic characteristic
+    # that could be easily picked up by the model
+    chars_to_ignore_regex = (
+        f'[{"".join(data_args.chars_to_ignore)}]' if data_args.chars_to_ignore is not None else None
+    )
+    text_column_name = "text"
+
+    def normalize_numbers(batch):
+        text = batch["text"]
+        text = normalize_numbers_ca(text)
+        batch["text"] = text.lower()
+        return batch
+
+    with training_args.main_process_first(desc="dataset verbalize numbers"):
+        raw_datasets = raw_datasets.map(
+            normalize_numbers,
+            desc="remove special characters from datasets",
+        )
+
+    def remove_special_characters(batch):
+        if chars_to_ignore_regex is not None:
+            batch["target_text"] = re.sub(chars_to_ignore_regex, "", batch[text_column_name]).lower() + " "
+            batch["target_text"] = re.sub("á", "a", batch["target_text"])
+            batch["target_text"] = re.sub("ñ", "ny", batch["target_text"])
+        else:
+            batch["target_text"] = batch[text_column_name].lower() + " "
+        return batch
+
+    with training_args.main_process_first(desc="dataset map special characters removal"):
+        raw_datasets = raw_datasets.map(
+            remove_special_characters,
+            remove_columns=[text_column_name],
+            desc="remove special characters from datasets",
+        )
+
+    set_characters = set()
+    for string in raw_datasets["train"]["target_text"]:
+        set_characters.update(string.lower())
+
+    vocab = [character for character in "aàbcçdeéèfghiíïjklmnoóòpqrstuúüvwxyz'·-"]
+
+    unwanted_chars = set_characters-set(vocab)-set([' '])
+
+    with training_args.main_process_first(desc="dataset filter non vocab chars"):
+        raw_datasets = raw_datasets.filter(
+            lambda example: not any((c in unwanted_chars) for c in example),
+            input_columns="target_text", 
+            desc="remove examples with weird characters"
+        )
+
+    # save special tokens for tokenizer
+    word_delimiter_token = data_args.word_delimiter_token
+    unk_token = data_args.unk_token
+    pad_token = data_args.pad_token
+
+    # 3. Next, let's load the config as we might need it to create
+    # the tokenizer
+    # load config
+    config = AutoConfig.from_pretrained(
+        model_args.model_name_or_path, cache_dir=model_args.cache_dir, use_auth_token=data_args.use_auth_token
+    )
+
+    # 4. Next, if no tokenizer file is defined,
+    # we create the vocabulary of the model by extracting all unique characters from
+    # the training and evaluation datasets
+    # We need to make sure that only first rank saves vocabulary
+    # make sure all processes wait until vocab is created
+    tokenizer_name_or_path = model_args.tokenizer_name_or_path
+    tokenizer_kwargs = {}
+    if tokenizer_name_or_path is None:
+        # save vocab in training output dir
+        tokenizer_name_or_path = training_args.output_dir
+
+        vocab_file = os.path.join(tokenizer_name_or_path, "vocab.json")
+
+        with training_args.main_process_first():
+            if training_args.overwrite_output_dir and os.path.isfile(vocab_file):
+                os.remove(vocab_file)
+                
+        with training_args.main_process_first(desc="dataset map vocabulary creation"):
+            if not os.path.isfile(vocab_file):
+                os.makedirs(tokenizer_name_or_path, exist_ok=True)
+                vocab_dict = create_vocabulary_from_data(
+                    raw_datasets,
+                    word_delimiter_token=word_delimiter_token,
+                    unk_token=unk_token,
+                    pad_token=pad_token,
+                )
+
+                # save vocab dict to be loaded into tokenizer
+                with open(vocab_file, "w") as file:
+                    json.dump(vocab_dict, file)
+
+        # if tokenizer has just been created
+        # it is defined by `tokenizer_class` if present in config else by `model_type`
+        tokenizer_kwargs = {
+            "config": config if config.tokenizer_class is not None else None,
+            "tokenizer_type": config.model_type if config.tokenizer_class is None else None,
+            "unk_token": unk_token,
+            "pad_token": pad_token,
+            "word_delimiter_token": word_delimiter_token,
+        }
+
+    # 5. Now we can instantiate the feature extractor, tokenizer and model
+    # Note for distributed training, the .from_pretrained methods guarantee that only
+    # one local process can concurrently download model & vocab.
+
+    # load feature_extractor and tokenizer
+    tokenizer = AutoTokenizer.from_pretrained(
+        tokenizer_name_or_path,
+        use_auth_token=data_args.use_auth_token,
+        **tokenizer_kwargs,
+    )
+    feature_extractor = AutoFeatureExtractor.from_pretrained(
+        model_args.model_name_or_path, cache_dir=model_args.cache_dir, use_auth_token=data_args.use_auth_token
+    )
+
+    # adapt config
+    config.update(
+        {
+            "feat_proj_dropout": model_args.feat_proj_dropout,
+            "attention_dropout": model_args.attention_dropout,
+            "hidden_dropout": model_args.hidden_dropout,
+            "final_dropout": model_args.final_dropout,
+            "mask_time_prob": model_args.mask_time_prob,
+            "mask_time_length": model_args.mask_time_length,
+            "mask_feature_prob": model_args.mask_feature_prob,
+            "mask_feature_length": model_args.mask_feature_length,
+            "gradient_checkpointing": training_args.gradient_checkpointing,
+            "layerdrop": model_args.layerdrop,
+            "ctc_loss_reduction": model_args.ctc_loss_reduction,
+            "pad_token_id": tokenizer.pad_token_id,
+            "vocab_size": len(tokenizer),
+            "activation_dropout": model_args.activation_dropout,
+        }
+    )
+
+    # create model
+    model = AutoModelForCTC.from_pretrained(
+        model_args.model_name_or_path,
+        cache_dir=model_args.cache_dir,
+        config=config,
+        use_auth_token=data_args.use_auth_token,
+    )
+
+    # freeze encoder
+    if model_args.freeze_feature_encoder:
+        model.freeze_feature_encoder()
+
+    # 6. Now we preprocess the datasets including loading the audio, resampling and normalization
+    # Thankfully, `datasets` takes care of automatically loading and resampling the audio,
+    # so that we just need to set the correct target sampling rate and normalize the input
+    # via the `feature_extractor`
+
+    # make sure that dataset decodes audio with correct sampling rate
+    dataset_sampling_rate = next(iter(raw_datasets.values())).features["audio"].sampling_rate
+    # if dataset_sampling_rate != feature_extractor.sampling_rate:
+    raw_datasets = raw_datasets.cast_column(
+        "audio", datasets.features.Audio(sampling_rate=feature_extractor.sampling_rate)
+    )
+
+    # derive max & min input length for sample rate & max duration
+    max_input_length = data_args.max_duration_in_seconds * feature_extractor.sampling_rate
+    min_input_length = data_args.min_duration_in_seconds * feature_extractor.sampling_rate
+    audio_column_name = "audio"
+    num_workers = data_args.preprocessing_num_workers
+
+    # `phoneme_language` is only relevant if the model is fine-tuned on phoneme classification
+    phoneme_language = data_args.phoneme_language
+
+    # Preprocessing the datasets.
+    # We need to read the audio files as arrays and tokenize the targets.
+    def prepare_dataset(batch):
+        # load audio
+        sample = batch[audio_column_name]
+
+        inputs = feature_extractor(sample["array"], sampling_rate=sample["sampling_rate"])
+        batch["input_values"] = inputs.input_values[0]
+        batch["input_length"] = len(batch["input_values"])
+
+        # encode targets
+        additional_kwargs = {}
+        if phoneme_language is not None:
+            additional_kwargs["phonemizer_lang"] = phoneme_language
+
+        batch["labels"] = tokenizer(batch["target_text"], **additional_kwargs).input_ids
+        return batch
+
+    raw_datasets = raw_datasets.shuffle(seed=42)
+
+    with training_args.main_process_first(desc="dataset map preprocessing"):
+        vectorized_datasets = raw_datasets.map(
+            prepare_dataset,
+            remove_columns=next(iter(raw_datasets.values())).column_names,
+            num_proc=num_workers,
+            desc="preprocess datasets",
+        )
+
+        def is_audio_in_length_range(length):
+            return length > min_input_length and length < max_input_length
+
+        # filter data that is shorter than min_input_length
+        vectorized_datasets = vectorized_datasets.filter(
+            is_audio_in_length_range,
+            num_proc=num_workers,
+            input_columns=["input_length"],
+        )
+
+    # 7. Next, we can prepare the training.
+    # Let's use word error rate (WER) as our evaluation metric,
+    # instantiate a data collator and the trainer
+
+    # Define evaluation metrics during training, *i.e.* word error rate, character error rate
+    eval_metrics = {metric: load_metric(metric) for metric in data_args.eval_metrics}
+
+    # for large datasets it is advised to run the preprocessing on a
+    # single machine first with ``args.preprocessing_only`` since there will mostly likely
+    # be a timeout when running the script in distributed mode.
+    # In a second step ``args.preprocessing_only`` can then be set to `False` to load the
+    # cached dataset
+    if data_args.preprocessing_only:
+        logger.info(f"Data preprocessing finished. Files cached at {vectorized_datasets.cache_files}")
+        return
+
+    def compute_metrics(pred):
+        pred_logits = pred.predictions
+        pred_ids = np.argmax(pred_logits, axis=-1)
+
+        pred.label_ids[pred.label_ids == -100] = tokenizer.pad_token_id
+
+        pred_str = tokenizer.batch_decode(pred_ids)
+        # we do not want to group tokens when computing the metrics
+        label_str = tokenizer.batch_decode(pred.label_ids, group_tokens=False)
+
+        metrics = {k: v.compute(predictions=pred_str, references=label_str) for k, v in eval_metrics.items()}
+
+        return metrics
+
+    # Now save everything to be able to create a single processor later
+    if is_main_process(training_args.local_rank):
+        # save feature extractor, tokenizer and config
+        feature_extractor.save_pretrained(training_args.output_dir)
+        tokenizer.save_pretrained(training_args.output_dir)
+        config.save_pretrained(training_args.output_dir)
+
+    try:
+        processor = AutoProcessor.from_pretrained(training_args.output_dir)
+    except (OSError, KeyError):
+        warnings.warn(
+            "Loading a processor from a feature extractor config that does not"
+            " include a `processor_class` attribute is deprecated and will be removed in v5. Please add the following "
+            " attribute to your `preprocessor_config.json` file to suppress this warning: "
+            " `'processor_class': 'Wav2Vec2Processor'`",
+            FutureWarning,
+        )
+        processor = Wav2Vec2Processor.from_pretrained(training_args.output_dir)
+
+    # Instantiate custom data collator
+    data_collator = DataCollatorCTCWithPadding(processor=processor)
+
+    # Initialize Trainer
+    trainer = Trainer(
+        model=model,
+        data_collator=data_collator,
+        args=training_args,
+        compute_metrics=compute_metrics,
+        train_dataset=vectorized_datasets["train"] if training_args.do_train else None,
+        eval_dataset=vectorized_datasets["eval"] if training_args.do_eval else None,
+        tokenizer=feature_extractor,
+    )
+
+    # 8. Finally, we can start training
+
+    # Training
+    if training_args.do_train:
+
+        # use last checkpoint if exist
+        if last_checkpoint is not None:
+            checkpoint = last_checkpoint
+        elif os.path.isdir(model_args.model_name_or_path):
+            checkpoint = model_args.model_name_or_path
+        else:
+            checkpoint = None
+
+        train_result = trainer.train(resume_from_checkpoint=checkpoint)
+        trainer.save_model()
+
+        metrics = train_result.metrics
+        max_train_samples = (
+            data_args.max_train_samples
+            if data_args.max_train_samples is not None
+            else len(vectorized_datasets["train"])
+        )
+        metrics["train_samples"] = min(max_train_samples, len(vectorized_datasets["train"]))
+
+        trainer.log_metrics("train", metrics)
+        trainer.save_metrics("train", metrics)
+        trainer.save_state()
+
+    # Evaluation
+    results = {}
+    if training_args.do_eval:
+        logger.info("*** Evaluate ***")
+        metrics = trainer.evaluate()
+        max_eval_samples = (
+            data_args.max_eval_samples if data_args.max_eval_samples is not None else len(vectorized_datasets["eval"])
+        )
+        metrics["eval_samples"] = min(max_eval_samples, len(vectorized_datasets["eval"]))
+
+        trainer.log_metrics("eval", metrics)
+        trainer.save_metrics("eval", metrics)
+
+    # Write model card and (optionally) push to hub
+    config_name = data_args.dataset_config_name[0] if data_args.dataset_config_name is not None else "na"
+    kwargs = {
+        "finetuned_from": model_args.model_name_or_path,
+        "tasks": "speech-recognition",
+        "tags": ["automatic-speech-recognition"]+data_args.dataset_name[0],
+        "dataset_args": f"Config: {config_name}, Training split: {data_args.train_split_name[0]}, Eval split: {data_args.eval_split_name[0]}",
+        "dataset": f"{data_args.dataset_name[0].upper()} - {config_name.upper()}",
+    }
+    if "common_voice" in data_args.dataset_name[0]:
+        kwargs["language"] = config_name
+
+    if training_args.push_to_hub:
+        trainer.push_to_hub(**kwargs)
+    else:
+        trainer.create_model_card(**kwargs)
+
+    return results
+
+
+if __name__ == "__main__":
+    main()

special_tokens_map.json

@@ -0,0 +1 @@
+{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "[UNK]", "pad_token": "[PAD]", "additional_special_tokens": [{"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}]}

text/LICENSE

@@ -0,0 +1,19 @@
+Copyright (c) 2017 Keith Ito
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.

text/__init__.py

@@ -0,0 +1,74 @@
+""" from https://github.com/keithito/tacotron """
+import re
+from text import cleaners
+from text.symbols import symbols
+
+
+# Mappings from symbol to numeric ID and vice versa:
+_symbol_to_id = {s: i for i, s in enumerate(symbols)}
+_id_to_symbol = {i: s for i, s in enumerate(symbols)}
+
+# Regular expression matching text enclosed in curly braces:
+_curly_re = re.compile(r'(.*?)\{(.+?)\}(.*)')
+
+
+def text_to_sequence(text, cleaner_names):
+  '''Converts a string of text to a sequence of IDs corresponding to the symbols in the text.
+
+    The text can optionally have ARPAbet sequences enclosed in curly braces embedded
+    in it. For example, "Turn left on {HH AW1 S S T AH0 N} Street."
+
+    Args:
+      text: string to convert to a sequence
+      cleaner_names: names of the cleaner functions to run the text through
+
+    Returns:
+      List of integers corresponding to the symbols in the text
+  '''
+  sequence = []
+
+  # Check for curly braces and treat their contents as ARPAbet:
+  while len(text):
+    m = _curly_re.match(text)
+    if not m:
+      sequence += _symbols_to_sequence(_clean_text(text, cleaner_names))
+      break
+    sequence += _symbols_to_sequence(_clean_text(m.group(1), cleaner_names))
+    sequence += _arpabet_to_sequence(m.group(2))
+    text = m.group(3)
+
+  return sequence
+
+
+def sequence_to_text(sequence):
+  '''Converts a sequence of IDs back to a string'''
+  result = ''
+  for symbol_id in sequence:
+    if symbol_id in _id_to_symbol:
+      s = _id_to_symbol[symbol_id]
+      # Enclose ARPAbet back in curly braces:
+      if len(s) > 1 and s[0] == '@':
+        s = '{%s}' % s[1:]
+      result += s
+  return result.replace('}{', ' ')
+
+
+def _clean_text(text, cleaner_names):
+  for name in cleaner_names:
+    cleaner = getattr(cleaners, name)
+    if not cleaner:
+      raise Exception('Unknown cleaner: %s' % name)
+    text = cleaner(text)
+  return text
+
+
+def _symbols_to_sequence(symbols):
+  return [_symbol_to_id[s] for s in symbols if _should_keep_symbol(s)]
+
+
+def _arpabet_to_sequence(text):
+  return _symbols_to_sequence(['@' + s for s in text.split()])
+
+
+def _should_keep_symbol(s):
+  return s in _symbol_to_id and s is not '_' and s is not '~'

text/ca.sor

@@ -0,0 +1,485 @@
+^0 zero
+1$ u
+1 un
+2 dos
+3 tres
+4 quatre
+5 cinc
+6 sis
+7 set
+8 huit # [:ca-valencia:] [:ca-ES-valencia:]
+8 vuit
+9 nou
+#10-19
+10 deu
+11 onze
+12 dotze
+13 tretze
+14 catorze
+15 quinze
+16 setze
+17 dèsset # [:ca-valencia:] [:ca-ES-valencia:]
+17 desset # [:ca-balear:] [:ca-ES-balear:]
+17 disset
+18 devuit # [:ca-balear:] [:ca-ES-balear:]
+18 díhuit # [:ca-valencia:] [:ca-ES-valencia:]
+19 denou # [:ca-balear:] [:ca-ES-balear:]
+19 dènou # [:ca-valencia:] [:ca-ES-valencia:]
+1(\d) di$1
+# 20-29
+20 vint
+2(\d) vint-i-$1
+# 30, 40, 50, 60, 70, 80, 90
+30 trenta
+40 quaranta
+50 cinquanta
+60 seixanta
+70 setanta
+80 huitanta # [:ca-valencia:] [:ca-ES-valencia:]
+80 vuitanta
+90 noranta
+(\d)(\d) $(\10)-$2
+
+#100-199
+100 cent
+1(\d\d) cent $1
+#200-999
+(\d)00 $1-cents
+(\d)(\d\d) $1-cents $2
+
+#1000-1999
+1000 mil
+1(\d{3}) mil $1
+
+#2000-999999
+(\d{1,3})000 $1 mil
+(\d{1,3})(\d{3}) $1 mil $2
+
+# our limit is number <10^606
+(\d{606,}) ""
+
+# x-lions
+# 10000000=10^6 -> un milió
+1((0{6})+) un $(pre:$(count:\1))lió
+1((\d{6})+) un $(pre:$(count:\1))lió $1
+# 2000000=2·10^6 -> dos milions
+(\d{1,6})((0{6})+) $1 $(pre:$(count:\2))lions
+(\d{1,6})((\d{6})+) $1 $(pre:$(count:\2))lions $2
+
+
+# count number of 10^6, usefull for x-lions, and x-liards prefixes.
+count:.{0,5}? 0
+count:.{6}.{0,5} 1
+count:(.{12}).{0,5} 2
+count:(.{18}).{0,5} 3
+count:(.{24}).{0,5} 4
+count:(.{30}).{0,5} 5
+count:(.{36}).{0,5} 6
+count:(.{42}).{0,5} 7
+count:(.{48}).{0,5} 8
+count:(.{54}).{0,5} 9
+count:(.{60})(.{0,59}) 1|$(count:\2)
+count:(.{120})(.{0,59}) 2|$(count:\2)
+count:(.{180})(.{0,59}) 3|$(count:\2)
+count:(.{240})(.{0,59}) 4|$(count:\2)
+count:(.{300})(.{0,59}) 5|$(count:\2)
+count:(.{360})(.{0,59}) 6|$(count:\2)
+count:(.{420})(.{0,59}) 7|$(count:\2)
+count:(.{480})(.{0,59}) 8|$(count:\2)
+count:(.{540})(.{0,59}) 9|$(count:\2)
+count:(.{600})(.{0,5}) 10|$(count:\2) # our limit is 10^606-1
+
+# prefixes needed for x-lions and x-liards, up to 10^606-1
+pre:1 mi
+pre:2 bi
+pre:3 tri
+pre:4 quadri
+pre:5 quinti
+pre:6 sexti
+pre:7 septi
+pre:8 octi
+pre:9 noni
+pre:10 deci
+pre:1(\d) $(pre2:\1)|deci
+pre:(\d)0 $(pre3:\1)
+pre:(\d)(\d) $(pre2:\2)|$(pre3:\1)
+pre:100 centi
+
+pre2:1 uno
+pre2:2 duo
+pre2:3 tre
+pre2:4 quattour
+pre2:5 quin
+pre2:6 sex
+pre2:7 septen
+pre2:8 octo
+pre2:9 novem
+
+pre3:1 deci
+pre3:2 viginti
+pre3:3 triginti
+pre3:4 quadraginti
+pre3:5 quinquaginti
+pre3:6 sexaginti
+pre3:7 septuaginti
+pre3:8 octoginti
+pre3:9 nonoginti
+pre3:10 centi
+
+# negative number
+[--](\d+) menys |$1
+
+# decimals
+"([^,]*\d)[.]((\d{3})+)([,][^,.]*)?" $(\1\2\4)
+"([--]?\d+)([,]0*)?" $1
+"([--]?\d+)[,](\d*)" $(\1·\2)
+"([--]?\d+·0*)([^0]00?)0*" $1| |$2
+"([--]?\d+·0*)([^0])" $1| |$2
+"([--]?\d+·0*)([^0]\d)" $1| |$2
+"([--]?\d+·0*)([^0]\d\d)" $1| |$2
+"([--]?\d+·0*)([^0]\d\d)0*" $1| |$2
+
+"([--]?\d+·0*)(([^0]|[^0]\d*[^0]))0*" $1| $(read:\2)
+"([--]?\d+)·(\d*)(\d)" $(\1·\2)| |$3
+"([--]?\d+)·" $1| coma
+
+# used for decimal part
+#read:(\d*[^0])0*$ $(read:\1)
+read:(\d*[1-9])(00+)([1-9]\d*) $(read:\1)| |$(read:\2) |$(read:\3)
+read:(\d$) $1
+read:0(\d+) $(read:0)| |$(read:\1)
+read:([1-9]\d) $1
+read:([1-9]\d\d) $1
+read:(\d\d\d) $1
+read:(\d\d)((\d\d)+) $(read:\1)| |$(read:\2)
+read:(\d\d)((\d\d)*)(\d\d\d) $(read:\1)| |$(read:\2)| |$(read:\4)
+
+
+# convert masculine forms to feminine forms
+# it can be run after: standard number conversion; and after ordinal, partitive functions.
+## runned with feminine function.
+f:(.*iliard)(.*) \1$(f:\2) # convert only <1,000,000,000
+f:(.*ili)(.*) \1$(f:\2) # convert only <100,0000
+f:(.*d)o(s[^èé]*) $(f:\1ue\2) # 2 -> dos -> dues
+f:(.*cent)(s.*) $(f:\1e\2) # cents -> centes
+f:(((.*)[^a-zèé]|))u$ \1una # vint-i-u -> vint-i-una
+## runned after ord function.
+f:(.*[^0-9])n$ \1na # segon -> segona
+f:(.*[^0-9]r)$ \1a # tercer -> tercera
+f:(.*[^0-9]r)t$ \1ta # quart -> quarta
+f:(.*[^0-9])è$ \1ena # sisè -> sisena
+f:(.*[^0-9])é$ \1ena # sisé -> sisena
+## runned after ord2 function.
+f:(.*[0-9])[nrtè]$ \1a # 2n -> 2a
+## runnded after part function.
+f:(.*ter)ç$ \1cera # terç -> tercera
+f:(.*è[sc]i)m$ \1ma # milionèsim -> milionèsima
+f:(.*[^0-9]i)g$ \1tja # mig -> mitja
+
+
+no-centes:(.*)centes(.*) \1cents\2
+no-centes:(.*) \1
+
+# convert ordinal numbers (1st, 2nd, 3rd,... nth) to partitive (1, 1/2, 1/3, .... 1/n)
+p:(.*)primer$ \1unitat
+p:(.*)segon$ \1mig
+p:(.*)tercer$ \1terç
+p:(.*quart)$ \1
+p:(.*)des[èé]$ \1dècim
+p:((.*)cent)[èé]$ \1èsim
+p:((.*)mil)[èé]$ \1·lèsim
+p:((.*)ilion)[èé]$ \1èsim
+p:((.*)iliard)[èé]$ \1èsim
+
+
+# fallback, ignore 1-letter not-defined fuctions
+.:(.*) \1
+
+# runned after ordinal and partitive fuctions
+pl:(.*[^\d][nrtnec])$ \1s
+pl:(.*[^\d])ig$ \1igs # mig -> mitjos
+pl:(.*[^\d])ja$ \1ges
+pl:(.*[^\d])a$ \1es
+pl:(.*[^\d])[èé]$ \1ens
+# after ord2: 1r->1rs, 2n->2ns, 5è->5ns, ...
+pl:(\d+[rnrt])$ \1s # 1r -> 1rs, 2n -> 2ns, 4t -> 4ts
+pl:(\d+)[èé]$ \1ns # 5è -> 5ns
+pl:(\d+)a$ \1es # 2a -> 2es
+# after partitive
+pl:([^[0-9]*[sç])$ \1os # dos -> dosos, terç > terços
+pl:([^[0-9]*è[sc]im)$ \1s # dècim -> dècims
+#fallback
+pl:(.*) \1
+
+
+# unit/subunit singular/plural
+# million or greater part of the number name separated by "ili" pattern
+# before masculine to feminine conversion
+us(.).:([^,]*),([^,]*),([^,]*),([^,]*),([^,]*),([^,]*) $(\1:\7)| \2
+up(.).:([^,]*),([^,]*),([^,]*),([^,]*),([^,]*),([^,]*) $(\1:\7)| \3
+ud(.).:([^,]*),([^,]*),([^,]*),([^,]*),([^,]*),([^,]*) $(\1:\7)| \4
+ss.(.):([^,]*),([^,]*),([^,]*),([^,]*),([^,]*),([^,]*) $(\1:\7)| \5
+sp.(.):([^,]*),([^,]*),([^,]*),([^,]*),([^,]*),([^,]*) $(\1:\7)| \6
+
+# "mm" means masculine unit and masculine subunit
+# Usually used by Catalan users
+CHF:(.+),(.+) $(\2mm: franc suís, francs suïssos, de francs suïssos, cèntim, cèntims, \1)
+EUR:(.+),(.+) $(\2mm: euro, euros, d'euros, cèntim, cèntims, \1)
+GBP:(.+),(.+) $(\2fm: lliura esterlina, lliures esterlines, de lliures esterlines, penic, penics, \1)
+JPY:(.+),(.+) $(\2mm: ien, iens, de iens, sen, sen, \1)
+USD:(.+),(.+) $(\2mm: dòlar dels EUA, dòlars dels EUA, de dòlars dels EUA, centau, centaus, \1)
+# ACTIVE ISO 4217 CODES--A--
+AED:(.+),(.+) $(\2mm: dírham dels Emirats Àrabs Units, dírhams dels Emirats Àrabs Units, de dírhams dels Emirats Àrabs Units, fils, fulús, \1)
+AFN:(.+),(.+) $(\2mm: afgani, afganis, d'afganis, puli, puli, \1)
+ALL:(.+),(.+) $(\2mm: lek, lekë, de lekë, qindarka, qindarka, \1)
+AMD:(.+),(.+) $(\2mm: dram, drams, de drams, luma, luma, \1)
+ANG:(.+),(.+) $(\2mm: florí de les Antilles Neerlandeses, florins de les Antilles Neerlandeses, de florins de les Antilles Neerlandeses, cèntim, cèntims, \1)
+AOA:(.+),(.+) $(\2fm: kwanza, kwanzes, de kwanzes, cèntim cèntims, \1)
+ARS:(.+),(.+) $(\2mm: peso argentí, pesos argentins, de pesos argentins, centau, centaus, \1)
+AUD(.+),(.+) $(\2mm: dòlar australià, dòlars australians, de dòlars australians, centau, centaus, \1)
+AWG:(.+),(.+) $(\2mm: florí d'Aruba, florins d'Aruba, de florins d'Aruba, cèntim, cèntims, \1)
+AZN:(.+),(.+) $(\2mm: manat azerbaidjanès, manats azerbaidjanesos, de manats azerbaidjanesos, qəpik, qəpik, \1)
+# ACTIVE ISO 4217 CODES --X--
+#XAF Franc CFA emès pel BEAC (Banc dels Estats de l'Àfrica Central)
+XAG:(.+),(.+) $(\2fm: unça de plata, unces de plata, d'unces de plata, cèntim, cèntims, \1)
+XAU:(.+),(.+) $(\2fm: unça d'or, unces d'or, d'unces d'or, cèntim, cèntims, \1)
+#XBA Unitat compensatòria europea (EURCO) (unitat per al mercat d'obligacions)
+#XBB Unitat monetària europea (EMU-6) (unitat per al mercat d'obligacions)
+#XBC Unitat de compte europea 9 (EUA-9) (unitat per al mercat d'obligacions)
+#XBD Unitat de compte europea 17 (EUA-17) (unitat per al mercat d'obligacions)
+#XCD Dòlar del Carib Oriental
+#XDR Drets especials de gir (del Fons Monetari Internacional)
+#XFU Franc UIC (divisa especial)
+#XOF Franc CFA emès pel BCEAO (Banc Central dels Estats de l'Àfrica Occidental)
+XPD:(.+),(.+) $(\2fm: unça de pal·ladi, unces de pal·ladi, d'unces de pal·ladi, cèntim, cèntims, \1)
+#XPF Franc CFP (per als territoris francesos del Pacífic)
+XPT:(.+),(.+) $(\2fm: unça de platí, unces de platí, d'unces de platí, cèntim, cèntims, \1)
+#XTS Codi reservat per a proves
+#XXX Sense moneda, sense transacció monetària
+# OBSOLETE ISO 4217 CODES --Replaced by EUR--
+ADF:(.+),(.+) $(\2mm: franc andorrà, francs andorrans, de francs andorrans, cèntim, cèntims, \1)
+ADP:(.+),(.+) $(\2fm: pesseta andorrana, pessetes andorranes, de pessetes andorranes, cèntim, cèntims, \1)
+ATS:(.+),(.+) $(\2mm: xíling austríac, xílings austríacs, de xílings austríacs, groschen, groschen, \1)
+BEF:(.+),(.+) $(\2mm: franc belga, francs belgues, de francs belgues, cèntim, cèntims, \1)
+CYP:(.+),(.+) $(\2mm: lliura xipriota, lliures xipriotes, de lliures xipriotes, cèntim, cèntims, \1)
+DEM:(.+),(.+) $(\2mm: marc alemany, marcs alemanys, de marcs alemanys, penic, penics, \1)
+ESP:(.+),(.+) $(\2fm: pesseta, pessetes, de pessetes, cèntim, cèntims, \1)
+FIM:(.+),(.+) $(\2mm: marc finlandès, marcs finlandesos, de marcs finlandesos, penic, penics, \1)
+FRF:(.+),(.+) $(\2mm: franc francès, francs francesos, de francs francesos, cèntim, cèntims, \1)
+GRD:(.+),(.+) $(\2fm: dracma grega, dracmes gregues, leptó, leptà, \1)
+IEP:(.+),(.+) $(\2fm: lliura irlandesa, lliures irlandeses, de lliures irlandeses, penic, penics, \1)
+ITL:(.+),(.+) $(\2fm: lira italiana, lires italianes, de lires italianes, cèntim, cèntims, \1)
+LUF:(.+),(.+) $(\2mm: franc luxemburguès, francs luxemburguesos, de francs luxemburguesos, cèntim, cèntims, \1)
+MCF:(.+),(.+) $(\2mm: franc monegasc, francs monegascs, de francs monegascs, cèntim, cèntims, \1)
+MTL:(.+),(.+) $(\2fm: lira maltesa, lires malteses, de lires malteses, cèntim, cèntims, \1)
+NLG:(.+),(.+) $(\2mm: florí neerlandès, florins neerlandesos, de florins neerlandesos, cèntim, cèntims, \1)
+PTE:(.+),(.+) $(\2mm: escut portuguès, escuts portuguesos, de escuts portuguesos, centau, centaus, \1)
+SIT:(.+),(.+) $(\2mm: tolar eslovè, tolars eslovens, de tolars eslovens, stotin, stotinov, \1)
+SKK:(.+),(.+) $(\2fm: corona eslovaca, corones eslovaques, de corones eslovaques, halier, halierov, \1)
+SML:(.+),(.+) $(\2fm: lira de San Marino, lires de San Marino, de lires de San Marino, cèntim, cèntims, \1)
+VAL:(.+),(.+) $(\2fm: lira vaticana, lires vaticanes, de lires vaticanes, cèntim, cèntims, \1)
+XEU:(.+),(.+) $(\2mm: ecu, ecus, d'ecus, cèntim, cèntims, \1)
+
+#crypto-currencies
+XMR:(.+),(.+) $(\2mm: monero, moneros, de moneros, piconero, piconeros, \1) #TODO: 1,000,000,000,000 piconeros = 1 monero
+XBT:(.+),(.+) $(\2mm: bitcoin, bitcoins, de bitcoins, satoshi, satoshis, \1) # TODO: 10,000,000 satoshis = 1,000 millibitcoin = 1 bitcoin
+
+# unknow currency
+[A-Z]{3}:.* ""
+
+
+"([A-Z]{3}) ([-−]?1)([.,]00?)?"$(\1:|$2,us)
+"([A-Z]{3}) ([-−]?\d+0{6,})([.,]00?)?"$(\1:|$2,ud)
+"([A-Z]{3}) ([-−]?\d+)([.,]00?)?"$(\1:|$2,up)
+"(([A-Z]{3}) [-−]?\d+)[.,](01)" $1 amb$(\2:un,ss)
+"(([A-Z]{3}) [-−]?\d+)[.,](\d)" $1 amb$(\2:|$(\30),sp)
+"(([A-Z]{3}) [-−]?\d+)[.,](\d\d)" $1 amb$(\2:|$3,sp)
+
+
+# detects number followed by currency code
+"([-−]?\d+)([.,]\d+)? ([A-Z]{3})" $(\3 \1\2)
+
+
+# currency symbols
+"€[ ]?([^ ]*)" $(EUR \1)
+"£[ ]?([^ ]*)" $(GBP \1)
+"\$[ ]?([^ ]*)" $(USD \1)
+"¥[ ]?([^ ]*)" $(JPY \1)
+"₩[ ]?([^ ]*)" $(KRW \1)
+"₽[ ]?([^ ]*)" $(RUB \1)
+"ɱ[ ]?([^ ]*)" $(XMR \1)
+"₿[ ]?([^ ]*)" $(XBT \1)
+
+"([^ ]+)[ ]?€$" $(EUR \1)
+"([^ ]+)[ ]?£$" $(GBP \1)
+"([^ ]+)[ ]?\$$" $(USD \1)
+"([^ ]+)[ ]?¥$" $(JPY \1)
+"([^ ]+)[ ]?₩$" $(KRW \1)
+"([^ ]+)[ ]?₽$" $(RUB \1)
+"([^ ]+)[ ]?ɱ$" $(XMR \1)
+"([^ ]+)[ ]?₿$" $(XBT \1)
+
+== feminine ==
+
+1 una
+(.*) $(f:|$1)
+
+== masculine ==
+
+1 un
+(.*) $1
+
+== ordinal(-masculine)? ==
+
+([-−]\d+) ""
+\d+[,.] ""
+0 zeroé # [:ca-valencia:] [:ca-ES-valencia:]
+0 zeroè
+1 primer
+2 segon
+3 tercer
+4 quart
+(\d+)$ $(ordinal $2)
+"un ([^ ]*(ilió|iliard))$" $(ordinal \2)
+(.*li)ó$ \2oné # [:ca-valencia:] [:ca-ES-valencia:]
+(.*li)ó$ \2onè
+(.*(cent|mil|ion|iliard))s?$ \2é # [:ca-valencia:] [:ca-ES-valencia:]
+(.*(cent|mil|ion|iliard))s?$ \2è 
+"(.* )u$" \2uné # [:ca-valencia:] [:ca-ES-valencia:]
+"(.* )u$" \2unè
+(.*-)u$ \2uné # [:ca-valencia:] [:ca-ES-valencia:]
+(.*-)u$ \2unè
+"u" primer
+"un" primer
+"dos" segon
+"tres" terç
+"quatre" quart
+(.*)cinc$ \2cinqué # [:ca-valencia:] [:ca-ES-valencia:]
+(.*)cinc$ \2cinquè
+(.*)dènou$ \2denové # [:ca-valencia:] [:ca-ES-valencia:]
+(.*)nou$ \2nové # [:ca-valencia:] [:ca-ES-valencia:]
+(.*)nou$ \2novè
+(.*)deu$ \2desé # [:ca-valencia:] [:ca-ES-valencia:]
+(.*)deu$ \2desè
+(.*)dèsset$ \2desseté # [:ca-valencia:] [:ca-ES-valencia:]
+(.*)díhuit$ \2dihuité # [:ca-valencia:] [:ca-ES-valencia:]
+(.*)[ae]$ \2é # [:ca-valencia:] [:ca-ES-valencia:]
+(.*)[ae]$ \2è
+(.*\D)$ \2é # [:ca-valencia:] [:ca-ES-valencia:]
+(.*\D)$ \2è
+
+== ordinal-feminine ==
+([-−]\d+) ""
+\d+[,.] ""
+(\d+)$ $(no-centes:$(f:$(ordinal \1)))
+
+== ordinal-masculine-plural ==
+
+([-−]?\d+) $(ordinal-masculine-plural $(ordinal \1))
+primer primers
+segon segons
+(.*)è  \1ens
+(.*)er   \1ers
+
+== ordinal-feminine-plural ==
+
+([-−]?\d+) $(ordinal-feminine-plural $(ordinal-feminine \1))
+(.*)a  \1es
+
+== ordinal-number(-masculine)? ==
+
+#(\d+) $(o:\2)
+1$ 1r
+2$ 2n
+3$ 3r
+4$ 4t
+(\d+)$ \2é # [:ca-valencia:] [:ca-ES-valencia:]
+(\d+)$ \2è
+
+== ordinal-number-feminine ==
+(\d+)$ \1a
+
+== partitive(-masculine)? ==
+([--]?\d+) $(p:$(ordinal \2))
+
+== partitive-feminine ==
+([--]?\d+) $(no-centes:$(f:$(p:$(ordinal \1))))
+
+
+== partitive(-masculine)?-plural ==
+([--]?\d+) $(pl:$(p:$(ordinal $2)))
+
+== partitive-feminine-plural ==
+([--]?\d+) $(no-centes:$(pl:$(f:$(p:$(ordinal $1)))))
+
+== fraction(-masculine)? ==
+([--]?1)(/1)? $2
+([--]?1)/2 mig
+([--]?1)/([3-9]\d*) $(masculine \2)| $(partitive \3)
+([--]?\d+)(/1)? $2
+([--]?\d+)/([1-9]\d*) $2| $(partitive-plural \3)
+
+== fraction-feminine ==
+([--]?1)(/1)? $(f:$1)| unitat
+([--]?1)/([1-9]\d*) $(f:$1)| $(partitive-feminine \2)| part
+([--]?\d+)(/1)? $(f:$1)| unitats
+([--]?\d+)/([1-9]\d*) $(f:$1)| $(partitive-feminine-plural \2)| parts
+
+== collective ==
+2 parell, parella o duo
+3 tern, terna, tercet, trio, tríada o treset
+4 qüern, tètrada, quartet, quarteta o quàdruple
+5 quintern, quintet, cinquet o quíntuple
+6 sextet, siset o sèxtuple
+7 septet, setet o sèptuple
+8 octet o òctuple
+9 nònuple
+10 dècada o dècuple
+12 dotzena
+100 centenar
+144 grossa
+1000 miler
+10000 miríada
+
+== years ==
+2 bienni
+3 trienni
+4 quadrienni
+5 quinquenni o lustre
+6 sesenni
+7 septenni
+10 dècada o decenni
+12 duodecenni
+15 quindecenni
+20 vintenni o vicenni
+30 trentenni o tricenni
+40 quarantenni
+50 cinquantenni
+60 seixantenni
+70 setantenni
+80 huitantenni [:ca-valencia:] [:ca-ES-valencia:]
+80 vuitantenni
+90 norantenni
+100 segle o centenni
+1000 mil·lenni
+
+== multiplicative ==
+2 doble o duple
+3 triple
+4 quàdruple
+5 quíntuple
+6 sèxtuple
+7 sèptuple
+8 òctuple
+9 nònuple
+10 dècuple
+12 duodècuple
+100 cèntuple
+1/10 subdècuple
+1/2 súbduple
+
+== help ==
+
+"" $(1)|, $(2), $(3)\n$(help feminine)$(help masculine)$(help ordinal-number-masculine)$(help ordinal-number-feminine)$(help ordinal-feminine)$(help ordinal-masculine)
+(feminine|masculine|ordinal(-number)?(-feminine|-masculine)?) \1: $(\1 1), $(\1 2), $(\1 3)\n
+

text/cleaners.py

@@ -0,0 +1,150 @@
+""" from https://github.com/keithito/tacotron """
+
+'''
+Cleaners are transformations that run over the input text at both training and eval time.
+
+Cleaners can be selected by passing a comma-delimited list of cleaner names as the "cleaners"
+hyperparameter. Some cleaners are English-specific. You'll typically want to use:
+  1. "english_cleaners" for English text
+  2. "transliteration_cleaners" for non-English text that can be transliterated to ASCII using
+     the Unidecode library (https://pypi.python.org/pypi/Unidecode)
+  3. "basic_cleaners" if you do not want to transliterate (in this case, you should also update
+     the symbols in symbols.py to match your data).
+'''
+
+import re
+from unidecode import unidecode
+from text.numbers import normalize_numbers
+from text.numbers_ca import normalize_numbers_ca
+from text.symbols import symbols
+
+# Regular expression matching whitespace:
+_whitespace_re = re.compile(r'\s+')
+
+# List of (regular expression, replacement) pairs for abbreviations:
+_abbreviations_en = [(re.compile('\\b%s\\.' % x[0], re.IGNORECASE), x[1]) for x in [
+  ('mrs', 'misess'),
+  ('mr', 'mister'),
+  ('dr', 'doctor'),
+  ('st', 'saint'),
+  ('co', 'company'),
+  ('jr', 'junior'),
+  ('maj', 'major'),
+  ('gen', 'general'),
+  ('drs', 'doctors'),
+  ('rev', 'reverend'),
+  ('lt', 'lieutenant'),
+  ('hon', 'honorable'),
+  ('sgt', 'sergeant'),
+  ('capt', 'captain'),
+  ('esq', 'esquire'),
+  ('ltd', 'limited'),
+  ('col', 'colonel'),
+  ('ft', 'fort'),
+]]
+
+# List of (regular expression, replacement) pairs for catalan abbreviations:
+_abbreviations_ca = [(re.compile('\\b%s\\b' % x[0], re.IGNORECASE), x[1]) for x in [
+  ('tv3', 't v tres'),
+  ('8tv', 'vuit t v'),
+  ('pp', 'p p'),
+  ('psoe', 'p soe'),
+  ('sr.?', 'senyor'),
+  ('sra.?', 'senyora'),
+  ('srta.?', 'senyoreta')
+]]
+
+_replacements_ca = [(re.compile('%s' % x[0], re.IGNORECASE), x[1]) for x in [
+  (';', ','),
+  (':', '\.'),
+  ('\.\.\.,', ','),
+  ('\.\.\.', '…'),
+  ('ñ','ny')
+]]
+
+
+def expand_abbreviations(text, lang='ca'):
+  if lang == 'en':
+    _abbreviations = _abbreviations_en
+  elif lang == 'ca':
+    _abbreviations = _abbreviations_ca
+  else:
+    raise ValueError('no %s language for abbreviations'%lang)
+  for regex, replacement in _abbreviations:
+    text = re.sub(regex, replacement, text)
+  return text
+
+
+def convert_characters(text, lang='ca'):
+  if lang == 'ca':
+    _replacements = _replacements_ca
+  else:
+    raise ValueError('no %s language for punctuation conversion'%lang)
+  for regex, replacement in _replacements_ca:
+    text = re.sub(regex, replacement, text)
+  return text
+
+
+def expand_numbers(text, lang="ca"):
+  if lang == 'ca':
+    return normalize_numbers_ca(text)
+  else:
+    return normalize_numbers(text)
+
+
+def lowercase(text):
+  return text.lower()
+
+
+def collapse_whitespace(text):
+  return re.sub(_whitespace_re, ' ', text)
+
+
+def convert_to_ascii(text, lang="ca"):
+  if lang == 'en':
+    return unidecode(text)
+  elif lang == 'ca':
+    char_replace = []
+    for t in set(list(text)):
+      if t not in symbols:
+        char_replace.append([t, unidecode(t)])
+    for target, replace in char_replace:
+      text = text.replace(target, replace)
+    return text
+  else:
+    raise ValueError('no %s language for punctuation conversion'%lang)
+
+
+def basic_cleaners(text):
+  '''Basic pipeline that lowercases and collapses whitespace without transliteration.'''
+  text = lowercase(text)
+  text = collapse_whitespace(text)
+  return text
+
+
+def transliteration_cleaners(text):
+  '''Pipeline for non-English text that transliterates to ASCII.'''
+  text = convert_to_ascii(text)
+  text = lowercase(text)
+  text = collapse_whitespace(text)
+  return text
+
+
+def english_cleaners(text):
+  '''Pipeline for English text, including number and abbreviation expansion.'''
+  text = convert_to_ascii(text)
+  text = lowercase(text)
+  text = expand_numbers(text, lang='en')
+  text = expand_abbreviations(text, lang='en')
+  text = collapse_whitespace(text)
+  return text
+
+
+def catalan_cleaners(text):
+  text = lowercase(text)
+  text = expand_numbers(text, lang="ca")
+  text = convert_characters(text, lang="ca")
+  text = convert_to_ascii(text, lang="ca")
+  text = expand_abbreviations(text, lang="ca")
+  text = collapse_whitespace(text)
+  return text

text/cmudict.py

@@ -0,0 +1,65 @@
+""" from https://github.com/keithito/tacotron """
+
+import re
+
+
+valid_symbols = [
+  'AA', 'AA0', 'AA1', 'AA2', 'AE', 'AE0', 'AE1', 'AE2', 'AH', 'AH0', 'AH1', 'AH2',
+  'AO', 'AO0', 'AO1', 'AO2', 'AW', 'AW0', 'AW1', 'AW2', 'AY', 'AY0', 'AY1', 'AY2',
+  'B', 'CH', 'D', 'DH', 'EH', 'EH0', 'EH1', 'EH2', 'ER', 'ER0', 'ER1', 'ER2', 'EY',
+  'EY0', 'EY1', 'EY2', 'F', 'G', 'HH', 'IH', 'IH0', 'IH1', 'IH2', 'IY', 'IY0', 'IY1',
+  'IY2', 'JH', 'K', 'L', 'M', 'N', 'NG', 'OW', 'OW0', 'OW1', 'OW2', 'OY', 'OY0',
+  'OY1', 'OY2', 'P', 'R', 'S', 'SH', 'T', 'TH', 'UH', 'UH0', 'UH1', 'UH2', 'UW',
+  'UW0', 'UW1', 'UW2', 'V', 'W', 'Y', 'Z', 'ZH'
+]
+
+_valid_symbol_set = set(valid_symbols)
+
+
+class CMUDict:
+  '''Thin wrapper around CMUDict data. http://www.speech.cs.cmu.edu/cgi-bin/cmudict'''
+  def __init__(self, file_or_path, keep_ambiguous=True):
+    if isinstance(file_or_path, str):
+      with open(file_or_path, encoding='latin-1') as f:
+        entries = _parse_cmudict(f)
+    else:
+      entries = _parse_cmudict(file_or_path)
+    if not keep_ambiguous:
+      entries = {word: pron for word, pron in entries.items() if len(pron) == 1}
+    self._entries = entries
+
+
+  def __len__(self):
+    return len(self._entries)
+
+
+  def lookup(self, word):
+    '''Returns list of ARPAbet pronunciations of the given word.'''
+    return self._entries.get(word.upper())
+
+
+
+_alt_re = re.compile(r'\([0-9]+\)')
+
+
+def _parse_cmudict(file):
+  cmudict = {}
+  for line in file:
+    if len(line) and (line[0] >= 'A' and line[0] <= 'Z' or line[0] == "'"):
+      parts = line.split('  ')
+      word = re.sub(_alt_re, '', parts[0])
+      pronunciation = _get_pronunciation(parts[1])
+      if pronunciation:
+        if word in cmudict:
+          cmudict[word].append(pronunciation)
+        else:
+          cmudict[word] = [pronunciation]
+  return cmudict
+
+
+def _get_pronunciation(s):
+  parts = s.strip().split(' ')
+  for part in parts:
+    if part not in _valid_symbol_set:
+      return None
+  return ' '.join(parts)

text/numbers.py

@@ -0,0 +1,71 @@
+""" from https://github.com/keithito/tacotron """
+
+import inflect
+import re
+
+
+_inflect = inflect.engine()
+_comma_number_re = re.compile(r'([0-9][0-9\,]+[0-9])')
+_decimal_number_re = re.compile(r'([0-9]+\.[0-9]+)')
+_pounds_re = re.compile(r'£([0-9\,]*[0-9]+)')
+_dollars_re = re.compile(r'\$([0-9\.\,]*[0-9]+)')
+_ordinal_re = re.compile(r'[0-9]+(st|nd|rd|th)')
+_number_re = re.compile(r'[0-9]+')
+
+
+def _remove_commas(m):
+  return m.group(1).replace(',', '')
+
+
+def _expand_decimal_point(m):
+  return m.group(1).replace('.', ' point ')
+
+
+def _expand_dollars(m):
+  match = m.group(1)
+  parts = match.split('.')
+  if len(parts) > 2:
+    return match + ' dollars'  # Unexpected format
+  dollars = int(parts[0]) if parts[0] else 0
+  cents = int(parts[1]) if len(parts) > 1 and parts[1] else 0
+  if dollars and cents:
+    dollar_unit = 'dollar' if dollars == 1 else 'dollars'
+    cent_unit = 'cent' if cents == 1 else 'cents'
+    return '%s %s, %s %s' % (dollars, dollar_unit, cents, cent_unit)
+  elif dollars:
+    dollar_unit = 'dollar' if dollars == 1 else 'dollars'
+    return '%s %s' % (dollars, dollar_unit)
+  elif cents:
+    cent_unit = 'cent' if cents == 1 else 'cents'
+    return '%s %s' % (cents, cent_unit)
+  else:
+    return 'zero dollars'
+
+
+def _expand_ordinal(m):
+  return _inflect.number_to_words(m.group(0))
+
+
+def _expand_number(m):
+  num = int(m.group(0))
+  if num > 1000 and num < 3000:
+    if num == 2000:
+      return 'two thousand'
+    elif num > 2000 and num < 2010:
+      return 'two thousand ' + _inflect.number_to_words(num % 100)
+    elif num % 100 == 0:
+      return _inflect.number_to_words(num // 100) + ' hundred'
+    else:
+      return _inflect.number_to_words(num, andword='', zero='oh', group=2).replace(', ', ' ')
+  else:
+    return _inflect.number_to_words(num, andword='')
+
+
+def normalize_numbers(text):
+  text = re.sub(_comma_number_re, _remove_commas, text)
+  text = re.sub(_pounds_re, r'\1 pounds', text)
+  text = re.sub(_dollars_re, _expand_dollars, text)
+  text = re.sub(_decimal_number_re, _expand_decimal_point, text)
+  text = re.sub(_ordinal_re, _expand_ordinal, text)
+  text = re.sub(_number_re, _expand_number, text)
+  return text

text/numbers_ca.py

@@ -0,0 +1,54 @@
+import re
+import io
+import pathlib
+from text.soros import compile
+
+filepath = pathlib.Path(__file__).parent.absolute()
+with io.open(f"{filepath}/ca.sor", 'r', encoding="utf-8") as prg:
+    num2text = compile(prg.read(), 'ca')
+
+_separador_milers_re = re.compile(r'([0-9][0-9\.]+[0-9]{3})')
+_decimal_re = re.compile(r'([0-9]+\,[0-9]+)')
+_ordinal_ms_re = re.compile(r'([0-9]+)(r|er|n|on|t|rt|è|e|ne|nè)+(\b)')
+_ordinal_mp_re = re.compile(r'([0-9]+)(rs|ns|ts|ns)+(\b)')
+_ordinal_fs_re = re.compile(r'([0-9]+)(a|ra|na|ta)+(\b)')
+_ordinal_fp_re = re.compile(r'([0-9]+)(es)+(\b)')
+_cardinal_re = re.compile(r'[0-9]+')
+_fraccions_re = re.compile(r'(\b)([0-9]+\/[0-9]+)(\b)')
+_hores_re = re.compile(r'(\b)([0-9]{1,2}):([0-9]{2})(\b)')
+
+def _esborra_separador_milers(m):
+  return m.group(1).replace('.', '')
+
+def _num2text(m):
+  return num2text.run(m.group(0))
+
+def _ordinal_ms(m):
+  return num2text.run(f"ordinal {m.group(1)}") + m.group(3)
+
+def _ordinal_mp(m):
+  return num2text.run(f"ordinal-masculine-plural {m.group(1)}") + m.group(3)
+
+def _ordinal_fs(m):
+  return num2text.run(f"ordinal-feminine {m.group(1)}") + m.group(3)
+
+def _ordinal_fp(m):
+  return num2text.run(f"ordinal-feminine-plural {m.group(1)}") + m.group(3)
+
+def _fraccions(m):
+  return m.group(1) + num2text.run(f"fraction {m.group(2)}") + m.group(3)
+
+def _hores(m):
+  return m.group(1) + num2text.run(m.group(2)) + " i " + num2text.run(m.group(3))  + m.group(4)
+
+def normalize_numbers_ca(text):
+  text = re.sub(_separador_milers_re, _esborra_separador_milers, text)
+  text = re.sub(_decimal_re, _num2text, text)
+  text = re.sub(_ordinal_ms_re, _ordinal_ms, text)
+  text = re.sub(_ordinal_mp_re, _ordinal_mp, text)
+  text = re.sub(_ordinal_fs_re, _ordinal_fs, text)
+  text = re.sub(_ordinal_fp_re, _ordinal_fp, text)
+  text = re.sub(_fraccions_re, _fraccions, text)
+  text = re.sub(_hores_re, _hores, text)
+  text = re.sub(_cardinal_re, _num2text, text)
+  return text

text/numbers_ca_test.py

@@ -0,0 +1,106 @@
+import unittest
+
+from text.numbers_ca import normalize_numbers_ca
+
+
+class NumbersCa(unittest.TestCase):
+    def test_cardinals(self):
+        """
+        Converteix cardinals simples en una frase
+        """
+        self.assertEqual(normalize_numbers_ca("Va nèixer el 23 de desembre de 1988"), "Va nèixer el vint-i-tres de desembre de mil nou-cents vuitanta-vuit")
+        self.assertEqual(normalize_numbers_ca("tinc 3 preguntes"), "tinc tres preguntes")
+    
+    def test_separador_milers(self):
+        """
+        Ignora separadors de milers
+        """
+        self.assertEqual(normalize_numbers_ca("1.000"), "mil")
+        self.assertEqual(normalize_numbers_ca("323.400"), "tres-cents vint-i-tres mil quatre-cents")
+        self.assertEqual(normalize_numbers_ca("900.323.400"), "nou-cents milions tres-cents vint-i-tres mil quatre-cents")
+    
+    def test_decimals(self):
+        """
+        Converteix decimals
+        """
+        self.assertEqual(normalize_numbers_ca("1,33"), "u coma trenta-tres")
+        self.assertEqual(normalize_numbers_ca("75,5"), "setanta-cinc coma cinc")
+        self.assertEqual(normalize_numbers_ca("75,555"), "setanta-cinc coma cinc-cents cinquanta-cinc")
+        self.assertEqual(normalize_numbers_ca("999.999.999,99"), "nou-cents noranta-nou milions nou-cents noranta-nou mil nou-cents noranta-nou coma noranta-nou")
+        self.assertEqual(normalize_numbers_ca("1,12345678900"), "u coma dotze trenta-quatre cinquanta-sis set-cents vuitanta-nou")
+
+    def test_decimals_2(self):
+        """
+        Ignora comes que no pertànyen a un número decimal
+        """
+        self.assertEqual(normalize_numbers_ca("Va comprar pa, vi i llonganisses"), "Va comprar pa, vi i llonganisses")
+        self.assertEqual(normalize_numbers_ca("El número guanyador és 1, 23, 55, 34"), "El número guanyador és u, vint-i-tres, cinquanta-cinc, trenta-quatre")
+
+    def test_ordinals_ms(self):
+        """
+        Converteix ordinals masculins singulars
+        """
+        self.assertEqual(normalize_numbers_ca("Va arribar 4t de 5"), "Va arribar quart de cinc")
+        self.assertEqual(normalize_numbers_ca("el 1r va ser ell"), "el primer va ser ell")
+        self.assertEqual(normalize_numbers_ca("el 3er, no va aguantar"), "el tercer, no va aguantar")
+        self.assertEqual(normalize_numbers_ca("2n"), "segon")
+        self.assertEqual(normalize_numbers_ca("2on"), "segon")
+        self.assertEqual(normalize_numbers_ca("4t"), "quart")
+        self.assertEqual(normalize_numbers_ca("4rt"), "quart")
+        self.assertEqual(normalize_numbers_ca("5è: remogueu la barreja"), "cinquè: remogueu la barreja")
+        self.assertEqual(normalize_numbers_ca("6e"), "sisè")
+        self.assertEqual(normalize_numbers_ca("6e"), "sisè")
+        self.assertEqual(normalize_numbers_ca("21nè"), "vint-i-unè")
+        self.assertEqual(normalize_numbers_ca("un 81ne de Palamós"), "un vuitanta-unè de Palamós")
+
+    def test_ordinals_fs(self):
+        """
+        Converteix ordinals femenins singulars
+        """
+        self.assertEqual(normalize_numbers_ca("1a"), "primera")
+        self.assertEqual(normalize_numbers_ca("3ra"), "tercera")
+        self.assertEqual(normalize_numbers_ca("2a"), "segona")
+        self.assertEqual(normalize_numbers_ca("2na"), "segona")
+        self.assertEqual(normalize_numbers_ca("4a."), "quarta.")
+        self.assertEqual(normalize_numbers_ca("pugi a la 4ta, després giri a l'esquerra"), "pugi a la quarta, després giri a l'esquerra")
+        self.assertEqual(normalize_numbers_ca("va quedar 5a en la classificació"), "va quedar cinquena en la classificació")
+        self.assertEqual(normalize_numbers_ca("la 5na vegada"), "la cinquena vegada")
+
+    def test_ordinals_mp(self):
+        """
+        Converteix ordinals masculins plurals
+        """
+        self.assertEqual(normalize_numbers_ca("1rs"), "primers")
+        self.assertEqual(normalize_numbers_ca("van arribar 2ns"), "van arribar segons")
+    
+    def test_ordinals_fp(self):
+        """
+        Converteix ordinals femenins plurals
+        """
+        self.assertEqual(normalize_numbers_ca("1es"), "primeres")
+    
+    def test_fraccions_s(self):
+        """
+        Converteix fraccions singulars
+        """
+        self.assertEqual(normalize_numbers_ca("1/2 got de vi"), "mig got de vi")
+        self.assertEqual(normalize_numbers_ca("1/3 de farina"), "un terç de farina")
+        self.assertEqual(normalize_numbers_ca("1/8"), "un vuitè")
+        
+    def test_fraccions_p(self):
+        """
+        Converteix fraccions plurals
+        """
+        self.assertEqual(normalize_numbers_ca("4/2 gots de vi"), "quatre migs gots de vi")
+        self.assertEqual(normalize_numbers_ca("2/3 de farina"), "dos terços de farina")
+        self.assertEqual(normalize_numbers_ca("3/8"), "tres vuitens")
+
+    def test_hores(self):
+        """
+        Converteix hores de manera simplificada
+        """
+        self.assertEqual(normalize_numbers_ca("a les 11:45"), "a les onze i quaranta-cinc")
+        self.assertEqual(normalize_numbers_ca("a partir de les 23:12"), "a partir de les vint-i-tres i dotze")
+        
+if __name__ == '__main__':
+    unittest.main()

text/soros.py

@@ -0,0 +1,140 @@
+"Soros interpreter (see http://numbertext.org)"
+from __future__ import unicode_literals
+from __future__ import print_function
+import re
+import sys
+
+
+def run(program, data, lang):
+    return compile(program, lang).run(data)
+
+
+def compile(program, lang):
+    return _Soros(program, lang)
+
+# conversion function
+
+
+def _tr(text, chars, chars2, delim):
+    for i in range(0, len(chars)):
+        text = text.replace(delim + chars[i], chars2[i])
+    return text
+
+
+# string literals for metacharacter encoding
+_m = "\\\";#$()|[]"
+# Unicode private area
+_c = u"\uE000\uE001\uE002\uE003\uE004\uE005\uE006\uE007\uE008\uE009"
+_pipe = u"\uE003"
+# separator prefix = \uE00A
+
+# pattern to recognize function calls in the replacement string
+_func = re.compile(_tr(r"""(?:\|?(?:\$\()+)?  # optional nested calls
+                (\|?\$\(([^\(\)]*)\)\|?)      # inner call (2 subgroups)
+                (?:\)+\|?)?""",               # optional nested calls
+                       _m[4:8], _c[:4], "\\"), re.X)  # \$, \(, \), \| -> \uE000..\uE003
+
+
+class _Soros:
+    def __init__(self, prg, lang):
+        self.lines = []
+        if prg.find("__numbertext__") == -1:
+            prg = "__numbertext__;" + prg
+        # default left zero deletion
+        # and separator function (no separation, if subcall returns with empty string)
+        prg = prg.replace("__numbertext__", u"""0+(0|[1-9]\\d*) $1
+\"([a-z][-a-z]* )0+(0|[1-9]\\d*)\" $(\\1\\2)
+\"\uE00A(.*)\uE00A(.+)\uE00A(.*)\" \\1\\2\\3
+\"\uE00A.*\uE00A\uE00A.*\"
+""")
+        prg = _tr(prg, _m[:4], _c[:4],
+                  "\\")  # \\, \", \;, \# -> \uE000..\uE003
+        # switch off all country-dependent lines, and switch on the requested ones
+        prg = re.sub(
+            r"(^|[\n;])([^\n;#]*#[^\n]*[\[]:[^\n:\]]*:][^\n]*)", r"\1#\2", prg)
+        prg = re.sub(r"(^|[\n;])#([^\n;#]*#[^\n]*[\[]:" +
+                     lang.replace("_", "-") + r":][^\n]*)", r"\1\2", prg)
+        matchline = re.compile("^\s*(\"[^\"]*\"|[^\s]*)\s*(.*[^\s])?\s*$")
+        prefix = ""
+        for s in re.sub("(#[^\n]*)?(\n|$)", ";", prg).split(";"):
+            macro = re.match("== *(.*[^ ]?) ==", s)
+            if macro != None:
+                prefix = macro.group(1)
+                continue
+            m = matchline.match(s)
+            if prefix != "" and s != "" and m != None:
+                s = m.group(1).strip("\"")
+                space = " " if s != "" else ""
+                caret = ""
+                if s[0:1] == "^":
+                    s = s[1:]
+                    caret = "^"
+                s2 = m.group(2) if m.group(2) != None else ""
+                s = "\"" + caret + prefix + space + s + "\" " + s2
+                m = matchline.match(s)
+            if m != None:
+                s = _tr(m.group(1).strip("\""), _c[1:4], _m[1:4], "") \
+                    .replace(_c[_m.find("\\")], "\\\\")  # -> \\, ", ;, #
+                if m.group(2) != None:
+                    s2 = m.group(2).strip("\"")
+                else:
+                    s2 = ""
+                # \$, \(, \), \|, \[, \] -> \uE004..\uE009
+                s2 = _tr(s2, _m[4:], _c[4:], "\\")
+                # call inner separator: [ ... $1 ... ] -> $(\uE00A ... \uE00A$1\uE00A ... )
+                s2 = re.sub(r"[\[]\$(\d\d?|\([^\)]+\))",
+                            u"$(\uE00A\uE00A|$\\1\uE00A", s2)
+                s2 = re.sub(r"[\[]([^\$[\\]*)\$(\d\d?|\([^\)]+\))",
+                            u"$(\uE00A\\1\uE00A$\\2\uE00A", s2)
+                # add "|" in terminating position
+                s2 = re.sub(r"\uE00A]$", "|\uE00A)", s2)
+                s2 = re.sub(r"]", ")", s2)
+                s2 = re.sub(r"(\$\d|\))\|\$", r"\1||$",
+                            s2)  # $()|$() -> $()||$()
+                # \uE000..\uE003-> \, ", ;, #
+                s2 = _tr(s2, _c[:4], _m[:4], "")
+                # $, (, ), | -> \uE000..\uE003
+                s2 = _tr(s2, _m[4:8], _c[:4], "")
+                # \uE004..\uE009 -> $, (, ), |, [, ]
+                s2 = _tr(s2, _c[4:], _m[4:], "")
+                s2 = re.sub(r"\\(\d)", r"\\g<\1>",
+                            re.sub(r"\uE000(\d)", "\uE000\uE001\\\\g<\\1>\uE002", s2))
+                try:
+                    self.lines = self.lines + [[
+                        re.compile("^" + s.lstrip("^").rstrip("$") + "$"),
+                        s2, s[:1] == "^", s[-1:] == "$"]]
+                except:
+                    print("Error in following regex line: " + s, file=sys.stderr)
+                    raise
+
+    def run(self, data):
+        return self._run(data, True, True)
+
+    def _run(self, data, begin, end):
+        for i in self.lines:
+            if not ((begin == False and i[2]) or (end == False and i[3])):
+                m = i[0].match(data)
+                if m:
+                    try:
+                        s = m.expand(i[1])
+                    except:
+                        print("Error for the following input: " +
+                              data, file=sys.stderr)
+                        raise
+                    n = _func.search(s)
+                    while n:
+                        b = False
+                        e = False
+                        if n.group(1)[0:1] == _pipe or n.group()[0:1] == _pipe:
+                            b = True
+                        elif n.start() == 0:
+                            b = begin
+                        if n.group(1)[-1:] == _pipe or n.group()[-1:] == _pipe:
+                            e = True
+                        elif n.end() == len(s):
+                            e = end
+                        s = s[:n.start(1)] + self._run(n.group(2),
+                                                       b, e) + s[n.end(1):]
+                        n = _func.search(s)
+                    return s
+        return ""

text/symbols.py

@@ -0,0 +1,17 @@
+""" from https://github.com/keithito/tacotron """
+
+'''
+Defines the set of symbols used in text input to the model.
+
+The default is a set of ASCII characters that works well for English or text that has been run through Unidecode. For other data, you can modify _characters. See TRAINING_DATA.md for details. '''
+from text import cmudict
+
+_pad = '_' # in principle not used in tacotron2
+_punctuation = '\'!,.?…· '
+_letters = 'AÀÁBCÇDEÉÈFGHIÍÏJKLMNOÓÒPQRSTUÜÚVWXYZaàábcçdeéèfghiíïjklmnoóòpqrstuüúvwxyz'
+
+# Prepend "@" to ARPAbet symbols to ensure uniqueness (some are the same as uppercase letters):
+_arpabet = ['@' + s for s in cmudict.valid_symbols]
+
+# Export all symbols:
+symbols = [_pad] + list(_punctuation) + list(_letters) + _arpabet

text/symbols_en.py

@@ -0,0 +1,18 @@
+""" from https://github.com/keithito/tacotron """
+
+'''
+Defines the set of symbols used in text input to the model.
+
+The default is a set of ASCII characters that works well for English or text that has been run through Unidecode. For other data, you can modify _characters. See TRAINING_DATA.md for details. '''
+from text import cmudict
+
+_pad        = '_'
+_punctuation = '!\'(),.:;? '
+_special = '-'
+_letters = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz'
+
+# Prepend "@" to ARPAbet symbols to ensure uniqueness (some are the same as uppercase letters):
+_arpabet = ['@' + s for s in cmudict.valid_symbols]
+
+# Export all symbols:
+symbols = [_pad] + list(_special) + list(_punctuation) + list(_letters) + _arpabet

tokenizer_config.json

@@ -0,0 +1 @@
+{"unk_token": "[UNK]", "bos_token": "<s>", "eos_token": "</s>", "pad_token": "[PAD]", "do_lower_case": false, "word_delimiter_token": "|", "special_tokens_map_file": null, "tokenizer_file": null, "name_or_path": "PereLluis13/wav2vec2-xls-r-300m-ca", "tokenizer_class": "Wav2Vec2CTCTokenizer"}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:34861f6ec08a47ca474aba28cfc694e9303c20ab6b67ba041b792072f2a8e759
+size 3055

vocab.json

@@ -0,0 +1 @@
+{"#": 1, "'": 2, "-": 3, "a": 4, "b": 5, "c": 6, "d": 7, "e": 8, "f": 9, "g": 10, "h": 11, "i": 12, "j": 13, "k": 14, "l": 15, "m": 16, "n": 17, "o": 18, "p": 19, "q": 20, "r": 21, "s": 22, "t": 23, "u": 24, "v": 25, "w": 26, "x": 27, "y": 28, "z": 29, "·": 30, "à": 31, "ç": 32, "è": 33, "é": 34, "í": 35, "ï": 36, "ò": 37, "ó": 38, "ú": 39, "ü": 40, "ः": 41, "|": 0, "[UNK]": 42, "[PAD]": 43}