xinference 1.2.0__py3-none-any.whl → 1.2.2__py3-none-any.whl

xinference/thirdparty/melo/text/spanish.py

@@ -0,0 +1,122 @@
+import pickle
+import os
+import re
+
+from . import symbols
+from .es_phonemizer import cleaner as es_cleaner
+from .es_phonemizer import es_to_ipa
+from transformers import AutoTokenizer
+
+
+def distribute_phone(n_phone, n_word):
+    phones_per_word = [0] * n_word
+    for task in range(n_phone):
+        min_tasks = min(phones_per_word)
+        min_index = phones_per_word.index(min_tasks)
+        phones_per_word[min_index] += 1
+    return phones_per_word
+
+def text_normalize(text):
+    text = es_cleaner.spanish_cleaners(text)
+    return text
+
+def post_replace_ph(ph):
+    rep_map = {
+        "：": ",",
+        "；": ",",
+        "，": ",",
+        "。": ".",
+        "！": "!",
+        "？": "?",
+        "\n": ".",
+        "·": ",",
+        "、": ",",
+        "...": "…"
+    }
+    if ph in rep_map.keys():
+        ph = rep_map[ph]
+    if ph in symbols:
+        return ph
+    if ph not in symbols:
+        ph = "UNK"
+    return ph
+
+def refine_ph(phn):
+    tone = 0
+    if re.search(r"\d$", phn):
+        tone = int(phn[-1]) + 1
+        phn = phn[:-1]
+    return phn.lower(), tone
+
+
+def refine_syllables(syllables):
+    tones = []
+    phonemes = []
+    for phn_list in syllables:
+        for i in range(len(phn_list)):
+            phn = phn_list[i]
+            phn, tone = refine_ph(phn)
+            phonemes.append(phn)
+            tones.append(tone)
+    return phonemes, tones
+
+
+# model_id = 'bert-base-uncased'
+model_id = 'dccuchile/bert-base-spanish-wwm-uncased'
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+
+def g2p(text, pad_start_end=True, tokenized=None):
+    if tokenized is None:
+        tokenized = tokenizer.tokenize(text)
+    # import pdb; pdb.set_trace()
+    phs = []
+    ph_groups = []
+    for t in tokenized:
+        if not t.startswith("#"):
+            ph_groups.append([t])
+        else:
+            ph_groups[-1].append(t.replace("#", ""))
+    
+    phones = []
+    tones = []
+    word2ph = []
+    # print(ph_groups)
+    for group in ph_groups:
+        w = "".join(group)
+        phone_len = 0
+        word_len = len(group)
+        if w == '[UNK]':
+            phone_list = ['UNK']
+        else:
+            phone_list = list(filter(lambda p: p != " ", es_to_ipa.es2ipa(w)))
+        
+        for ph in phone_list:
+            phones.append(ph)
+            tones.append(0)
+            phone_len += 1
+        aaa = distribute_phone(phone_len, word_len)
+        word2ph += aaa
+        # print(phone_list, aaa)
+        # print('=' * 10)
+
+    if pad_start_end:
+        phones = ["_"] + phones + ["_"]
+        tones = [0] + tones + [0]
+        word2ph = [1] + word2ph + [1]
+    return phones, tones, word2ph
+
+def get_bert_feature(text, word2ph, device=None):
+    from text import spanish_bert
+    return spanish_bert.get_bert_feature(text, word2ph, device=device)
+
+if __name__ == "__main__":
+    text = "en nuestros tiempos estos dos pueblos ilustres empiezan a curarse, gracias sólo a la sana y vigorosa higiene de 1789."
+    # print(text)
+    text = text_normalize(text)
+    print(text)
+    phones, tones, word2ph = g2p(text)
+    bert = get_bert_feature(text, word2ph)
+    print(phones)
+    print(len(phones), tones, sum(word2ph), bert.shape)
+
+

xinference/thirdparty/melo/text/spanish_bert.py

@@ -0,0 +1,39 @@
+import torch
+from transformers import AutoTokenizer, AutoModelForMaskedLM
+import sys
+
+model_id = 'dccuchile/bert-base-spanish-wwm-uncased'
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = None
+
+def get_bert_feature(text, word2ph, device=None):
+    global model
+    if (
+        sys.platform == "darwin"
+        and torch.backends.mps.is_available()
+        and device == "cpu"
+    ):
+        device = "mps"
+    if not device:
+        device = "cuda"
+    if model is None:
+        model = AutoModelForMaskedLM.from_pretrained(model_id).to(
+            device
+        )
+    with torch.no_grad():
+        inputs = tokenizer(text, return_tensors="pt")
+        for i in inputs:
+            inputs[i] = inputs[i].to(device)
+        res = model(**inputs, output_hidden_states=True)
+        res = torch.cat(res["hidden_states"][-3:-2], -1)[0].cpu()
+        
+    assert inputs["input_ids"].shape[-1] == len(word2ph)
+    word2phone = word2ph
+    phone_level_feature = []
+    for i in range(len(word2phone)):
+        repeat_feature = res[i].repeat(word2phone[i], 1)
+        phone_level_feature.append(repeat_feature)
+
+    phone_level_feature = torch.cat(phone_level_feature, dim=0)
+
+    return phone_level_feature.T

xinference/thirdparty/melo/text/symbols.py

@@ -0,0 +1,290 @@
+# punctuation = ["!", "?", "…", ",", ".", "'", "-"]
+punctuation = ["!", "?", "…", ",", ".", "'", "-", "¿", "¡"]
+pu_symbols = punctuation + ["SP", "UNK"]
+pad = "_"
+
+# chinese
+zh_symbols = [
+    "E",
+    "En",
+    "a",
+    "ai",
+    "an",
+    "ang",
+    "ao",
+    "b",
+    "c",
+    "ch",
+    "d",
+    "e",
+    "ei",
+    "en",
+    "eng",
+    "er",
+    "f",
+    "g",
+    "h",
+    "i",
+    "i0",
+    "ia",
+    "ian",
+    "iang",
+    "iao",
+    "ie",
+    "in",
+    "ing",
+    "iong",
+    "ir",
+    "iu",
+    "j",
+    "k",
+    "l",
+    "m",
+    "n",
+    "o",
+    "ong",
+    "ou",
+    "p",
+    "q",
+    "r",
+    "s",
+    "sh",
+    "t",
+    "u",
+    "ua",
+    "uai",
+    "uan",
+    "uang",
+    "ui",
+    "un",
+    "uo",
+    "v",
+    "van",
+    "ve",
+    "vn",
+    "w",
+    "x",
+    "y",
+    "z",
+    "zh",
+    "AA",
+    "EE",
+    "OO",
+]
+num_zh_tones = 6
+
+# japanese
+ja_symbols = [
+    "N",
+    "a",
+    "a:",
+    "b",
+    "by",
+    "ch",
+    "d",
+    "dy",
+    "e",
+    "e:",
+    "f",
+    "g",
+    "gy",
+    "h",
+    "hy",
+    "i",
+    "i:",
+    "j",
+    "k",
+    "ky",
+    "m",
+    "my",
+    "n",
+    "ny",
+    "o",
+    "o:",
+    "p",
+    "py",
+    "q",
+    "r",
+    "ry",
+    "s",
+    "sh",
+    "t",
+    "ts",
+    "ty",
+    "u",
+    "u:",
+    "w",
+    "y",
+    "z",
+    "zy",
+]
+num_ja_tones = 1
+
+# English
+en_symbols = [
+    "aa",
+    "ae",
+    "ah",
+    "ao",
+    "aw",
+    "ay",
+    "b",
+    "ch",
+    "d",
+    "dh",
+    "eh",
+    "er",
+    "ey",
+    "f",
+    "g",
+    "hh",
+    "ih",
+    "iy",
+    "jh",
+    "k",
+    "l",
+    "m",
+    "n",
+    "ng",
+    "ow",
+    "oy",
+    "p",
+    "r",
+    "s",
+    "sh",
+    "t",
+    "th",
+    "uh",
+    "uw",
+    "V",
+    "w",
+    "y",
+    "z",
+    "zh",
+]
+num_en_tones = 4
+
+# Korean
+kr_symbols = ['ᄌ', 'ᅥ', 'ᆫ', 'ᅦ', 'ᄋ', 'ᅵ', 'ᄅ', 'ᅴ', 'ᄀ', 'ᅡ', 'ᄎ', 'ᅪ', 'ᄑ', 'ᅩ', 'ᄐ', 'ᄃ', 'ᅢ', 'ᅮ', 'ᆼ', 'ᅳ', 'ᄒ', 'ᄆ', 'ᆯ', 'ᆷ', 'ᄂ', 'ᄇ', 'ᄉ', 'ᆮ', 'ᄁ', 'ᅬ', 'ᅣ', 'ᄄ', 'ᆨ', 'ᄍ', 'ᅧ', 'ᄏ', 'ᆸ', 'ᅭ', '(', 'ᄊ', ')', 'ᅲ', 'ᅨ', 'ᄈ', 'ᅱ', 'ᅯ', 'ᅫ', 'ᅰ', 'ᅤ', '~', '\\', '[', ']', '/', '^', ':', 'ㄸ', '*']
+num_kr_tones = 1
+
+# Spanish
+es_symbols = [
+        "N",
+        "Q",
+        "a",
+        "b",
+        "d",
+        "e",
+        "f",
+        "g",
+        "h",
+        "i",
+        "j",
+        "k",
+        "l",
+        "m",
+        "n",
+        "o",
+        "p",
+        "s",
+        "t",
+        "u",
+        "v",
+        "w",
+        "x",
+        "y",
+        "z",
+        "ɑ",
+        "æ",
+        "ʃ",
+        "ʑ",
+        "ç",
+        "ɯ",
+        "ɪ",
+        "ɔ",
+        "ɛ",
+        "ɹ",
+        "ð",
+        "ə",
+        "ɫ",
+        "ɥ",
+        "ɸ",
+        "ʊ",
+        "ɾ",
+        "ʒ",
+        "θ",
+        "β",
+        "ŋ",
+        "ɦ",
+        "ɡ",
+        "r",
+        "ɲ",
+        "ʝ",
+        "ɣ",
+        "ʎ",
+        "ˈ",
+        "ˌ",
+        "ː"
+    ]
+num_es_tones = 1
+
+# French 
+fr_symbols = [
+    "\u0303",
+    "œ",
+    "ø",
+    "ʁ",
+    "ɒ",
+    "ʌ",
+    "ɜ",
+    "ɐ"
+]
+num_fr_tones = 1
+
+# German 
+de_symbols = [
+    "ʏ",
+    "̩"
+  ]
+num_de_tones = 1
+
+# Russian 
+ru_symbols = [
+    "ɭ",
+    "ʲ",
+    "ɕ",
+    "\"",
+    "ɵ",
+    "^",
+    "ɬ"
+]
+num_ru_tones = 1
+
+# combine all symbols
+normal_symbols = sorted(set(zh_symbols + ja_symbols + en_symbols + kr_symbols + es_symbols + fr_symbols + de_symbols + ru_symbols))
+symbols = [pad] + normal_symbols + pu_symbols
+sil_phonemes_ids = [symbols.index(i) for i in pu_symbols]
+
+# combine all tones
+num_tones = num_zh_tones + num_ja_tones + num_en_tones + num_kr_tones + num_es_tones + num_fr_tones + num_de_tones + num_ru_tones
+
+# language maps
+language_id_map = {"ZH": 0, "JP": 1, "EN": 2, "ZH_MIX_EN": 3, 'KR': 4, 'ES': 5, 'SP': 5 ,'FR': 6}
+num_languages = len(language_id_map.keys())
+
+language_tone_start_map = {
+    "ZH": 0,
+    "ZH_MIX_EN": 0,
+    "JP": num_zh_tones,
+    "EN": num_zh_tones + num_ja_tones,
+    'KR': num_zh_tones + num_ja_tones + num_en_tones,
+    "ES": num_zh_tones + num_ja_tones + num_en_tones + num_kr_tones,
+    "SP": num_zh_tones + num_ja_tones + num_en_tones + num_kr_tones,
+    "FR": num_zh_tones + num_ja_tones + num_en_tones + num_kr_tones + num_es_tones,
+}
+
+if __name__ == "__main__":
+    a = set(zh_symbols)
+    b = set(en_symbols)
+    print(sorted(a & b))