DeepMorphy

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+best_checkpoint/-44.data-00000-of-00001 filter=lfs diff=lfs merge=lfs -text
+best_checkpoint/-44.meta filter=lfs diff=lfs merge=lfs -text

10_release_dict.py

@@ -0,0 +1,244 @@
+import re
+import os
+import gzip
+import pickle
+from utils import CONFIG, get_dict_path, load_datasets
+
+
+NAR_REG = re.compile("\d+-.*")
+RANDOM_SEED = 1917
+VECT_PATH = CONFIG['vect_words_path']
+DATASET_PATH = CONFIG['dataset_path']
+REZ_PATHS = CONFIG['publish_dictionary_paths']
+DICT_WORDS_PATH = CONFIG['dict_words_path']
+NOT_DICT_WORDS_PATH = CONFIG['dataset_words_path']
+MAX_WORD_SIZE = CONFIG['max_word_size']
+DICT_POST_TYPES = CONFIG['dict_post_types']
+GRAMMEMES_TYPES = CONFIG['grammemes_types']
+IGNORE_AD_TAGS = CONFIG['dict_ignore_tags']
+REPLACE_WORD_DICT_ID = 1
+
+CLASSES_INDEX_DICT = {
+    cls: GRAMMEMES_TYPES[gram]['classes'][cls]['index']
+    for gram in sorted(GRAMMEMES_TYPES, key=lambda x: GRAMMEMES_TYPES[x]['index'])
+    for cls in GRAMMEMES_TYPES[gram]['classes']
+}
+
+POST_POWER_DICT = {}
+for key in DICT_POST_TYPES:
+    POST_POWER_DICT[key] = DICT_POST_TYPES[key]['power'] if 'power' in DICT_POST_TYPES[key] else 1
+
+p_dic = GRAMMEMES_TYPES['post']['classes']
+for key in p_dic:
+    POST_POWER_DICT[key] = p_dic[key]['power'] if 'power' in p_dic[key] else 1
+
+with open(CONFIG['inflect_templates_path'], 'rb') as f:
+    inflect_templates = pickle.load(f)
+
+with open(CONFIG['tags_path'], 'rb') as f:
+    tpl_cls_dict = pickle.load(f)
+
+lemma_cls_dict = {}
+for lemma_tpl in inflect_templates:
+    lemma_id = tpl_cls_dict[lemma_tpl]['i']
+    for tpl in inflect_templates[lemma_tpl]:
+        lemma_cls_dict[tpl_cls_dict[tpl]['i']] = lemma_id
+
+lemma_dict = {}
+for item in load_datasets('inflect', 'test', 'train', 'valid'):
+    if item['id'] not in lemma_dict:
+        lemma_dict[item['id']] = (item['x_src'], item['x_cls'])
+
+ad_tags_dict = {}
+with open(VECT_PATH, 'rb') as f:
+    vec_words = pickle.load(f)
+    for word in vec_words:
+        item = vec_words[word]
+        for form in item['forms']:
+            lexeme_id_key = 'inflect_id' if 'inflect_id' in form else 'id'
+            lexeme_id = form[lexeme_id_key]
+
+            if 'ad_tags' not in form:
+                continue
+
+            if lexeme_id not in ad_tags_dict:
+                ad_tags_dict[lexeme_id] = set()
+
+            ad_tags_dict[lexeme_id].add(form['ad_tags'])
+
+
+def build_index(words_dics):
+    text_forms_dict = {}
+    for id in words_dics:
+        for item in words_dics[id]:
+            text = item['text']
+            if text not in text_forms_dict:
+                text_forms_dict[text] = []
+            text_forms_dict[text].append(item)
+
+    index = []
+    for text in text_forms_dict:
+        lexemes = [str(item['id']) for item in text_forms_dict[text]]
+        lexemes = ','.join(lexemes)
+        index.append(f"{text}:{lexemes}")
+
+    index = list(set(index))
+    index = '\n'.join(index)
+    return index
+
+
+def create_dictionary(words_dics):
+    index = build_index(words_dics)
+
+    lexeme = []
+    for id in words_dics:
+        cur_lexeme = [id, '\t']
+
+        order = []
+        cur_forms_dict = {}
+        for item in words_dics[id]:
+            if item['text'] not in cur_forms_dict:
+                cur_forms_dict[item['text']] = {}
+
+            replace_other = item['replace_other'] if 'replace_other' in item else False
+            cur_form_dic = cur_forms_dict[item['text']]
+            if item['main'] not in cur_forms_dict or not cur_forms_dict[item['main']]:
+                cur_form_dic[item['main']] = replace_other
+
+            if item['text'] not in order:
+                order.append(item['text'])
+
+        for text in order:
+            cur_lexeme.append(text)
+            cur_lexeme.append(':')
+            for cls in cur_forms_dict[text]:
+                replace_other = cur_forms_dict[text][cls]
+                cur_lexeme.append(str(cls))
+                if replace_other:
+                    cur_lexeme.append('!')
+
+                cur_lexeme.append(",")
+
+            del cur_lexeme[-1]
+            cur_lexeme.append(';')
+
+        del cur_lexeme[-1]
+        lexeme.append(''.join(cur_lexeme))
+
+    lexeme = '\n'.join(lexeme)
+    return index, lexeme
+
+
+def save_dictionary(index, lexeme, paths, file_name):
+    for path in paths:
+       path = os.path.join(path, f"{file_name}_index.txt.gz")
+       with gzip.open(path, 'wb+') as f:
+           f.write(index.encode('utf-8'))
+
+    for path in paths:
+       path = os.path.join(path, f"{file_name}.txt.gz")
+       with gzip.open(path, 'wb+') as f:
+           f.write(lexeme.encode('utf-8'))
+
+
+def release_dict_items():
+    with open(DICT_WORDS_PATH, 'rb') as f:
+        words = pickle.load(f)
+
+    words = [word for word in words if word['post'] != 'numb']
+    dict_words = {}
+    for word in words:
+        if word['id'] not in dict_words:
+            dict_words[word['id']] = []
+        dict_words[word['id']].append(word)
+
+    for id in dict_words:
+        un_id_dict = {}
+        rez_list = []
+        for word in sorted(dict_words[id], key=lambda x: x['index']):
+            if word['main'] not in un_id_dict:
+                un_id_dict[word['main']] = word
+            else:
+                un_id_dict[word['main']]['replace_other'] = True
+            rez_list.append(word)
+        dict_words[id] = rez_list
+
+    index, lexeme = create_dictionary(dict_words)
+    save_dictionary(index, lexeme, REZ_PATHS, 'dict')
+
+
+def release_correction_items():
+    dict_words = {}
+
+    with open(get_dict_path('lemma'), 'rb') as f:
+        items = pickle.load(f)
+    for word in items:
+        lexeme_id = word['id']
+        if lexeme_id not in lemma_dict \
+            or (lexeme_id in ad_tags_dict and any([key in ad_tags_dict[lexeme_id] for key in IGNORE_AD_TAGS])):
+            continue
+
+        if lexeme_id not in dict_words:
+            dict_words[lexeme_id] = []
+
+        dict_words[lexeme_id].append(word)
+        dict_words[lexeme_id].append({
+            'id': lexeme_id,
+            'text': lemma_dict[lexeme_id][0],
+            'main': lemma_dict[lexeme_id][1],
+        })
+
+    with open(os.path.join(CONFIG['bad_path'], "bad_lemma.pkl"), 'rb') as f:
+        items = pickle.load(f)
+    for word in items:
+        word = word[0]
+        lexeme_id = word['id']
+        if lexeme_id not in lemma_dict \
+            or (lexeme_id in ad_tags_dict and any([key in ad_tags_dict[lexeme_id] for key in IGNORE_AD_TAGS])):
+            continue
+
+        if lexeme_id not in dict_words:
+            dict_words[lexeme_id] = []
+
+        lemma, lemma_cls = lemma_dict[lexeme_id]
+        dict_words[lexeme_id].append(dict(id=lexeme_id, main=word['main_cls'], text=word['x_src'], replace_other=True))
+        dict_words[lexeme_id].append(dict(id=lexeme_id, main=lemma_cls, text=lemma, replace_other=True))
+
+    with open(os.path.join(CONFIG['bad_path'], "bad_inflect.pkl"), 'rb') as f:
+        items = pickle.load(f)
+    for word in items:
+        word = word[0]
+        lexeme_id = word['id']
+        if lexeme_id not in lemma_dict \
+            or (lexeme_id in ad_tags_dict and any([key in ad_tags_dict[lexeme_id] for key in IGNORE_AD_TAGS])):
+            continue
+
+        if lexeme_id not in dict_words:
+            dict_words[lexeme_id] = []
+        dict_words[lexeme_id].append(dict(id=lexeme_id, main=word['x_cls'], text=word['x_src'], replace_other=True))
+        dict_words[lexeme_id].append(dict(id=lexeme_id, main=word['y_cls'], text=word['y_src'], replace_other=True))
+
+    with open(os.path.join(CONFIG['bad_path'], "bad_main.pkl"), 'rb') as f:
+        items = pickle.load(f)
+    for bad_item in items:
+        text = bad_item[0]['src']
+        for word in vec_words[text]['forms']:
+            lexeme_id = word['id']
+            if lexeme_id not in lemma_dict \
+                or (lexeme_id in ad_tags_dict and any([key in ad_tags_dict[lexeme_id] for key in IGNORE_AD_TAGS])):
+                continue
+
+            if lexeme_id not in dict_words:
+                dict_words[lexeme_id] = []
+
+            cls_id = tpl_cls_dict[word['main']]['i']
+            dict_words[lexeme_id].append(dict(id=lexeme_id, main=cls_id, text=text, replace_other=True))
+            dict_words[lexeme_id].append(dict(id=lexeme_id, main=lemma_dict[lexeme_id][1], text=lemma_dict[lexeme_id][0], replace_other=True))
+
+    index, lexeme = create_dictionary(dict_words)
+    save_dictionary(index, lexeme, REZ_PATHS, 'dict_correction')
+
+
+release_correction_items()
+release_dict_items()
+

11_release_tests.py

@@ -0,0 +1,331 @@
+import os
+import pickle
+import numpy as np
+from utils import CONFIG, RANDOM, load_datasets
+
+from lxml import etree
+from xml.etree.ElementTree import ElementTree
+
+
+GRAM_TYPES = CONFIG['grammemes_types']
+ROOT = CONFIG['publish_tests_path']
+DICT_WORDS_PATH = CONFIG['dict_words_path']
+NMB_DATA_PATH = CONFIG['numb_data_path']
+with open(NMB_DATA_PATH, 'rb') as f:
+    numb_data = pickle.load(f)
+
+with open(CONFIG['tags_path'], 'rb') as f:
+    tags = pickle.load(f)
+    is_lemma_dict = {tags[key]['i']: tags[key]['l'] for key in tags}
+    tag_index_order = {tags[tag]['i']: tags[tag]['o'] for tag in tags}
+
+
+def release_gram_tests(items, key, cls_dic, result_path):
+    root = etree.Element('Tests')
+    for word in items:
+        y = np.argwhere(word['y'] == 1).ravel()
+        y = ';'.join([cls_dic[index] for index in y])
+        test = etree.Element("T")
+        test.set('x', word['src'])
+        test.set('y', y)
+        root.append(test)
+
+    rez_path = os.path.join(result_path, f'{key}.xml')
+    tree = ElementTree(root)
+    with open(rez_path, 'wb+') as f:
+        tree.write(f, xml_declaration=True, encoding='utf-8')
+
+
+def release_main_tests(items, result_path, y_is_index=True):
+    root = etree.Element('Tests')
+    for word in items:
+        if y_is_index:
+            y = word['y']
+        else:
+            y = np.argwhere(word['y'] == 1).ravel()
+        y = ';'.join([str(index) for index in y])
+        test = etree.Element("T")
+        test.set('x', word['src'])
+        test.set('y', y)
+        root.append(test)
+
+    rez_path = os.path.join(result_path, f'main.xml')
+    tree = ElementTree(root)
+    with open(rez_path, 'wb+') as f:
+        tree.write(f, xml_declaration=True, encoding='utf-8')
+
+
+def release_lemma_tests(items, result_path):
+    root = etree.Element('Tests')
+    for word in items:
+        test = etree.Element("T")
+        test.set('x', word['x_src'])
+        test.set('x_c', str(word['main_cls']))
+        test.set('y', word['y_src'])
+        root.append(test)
+
+    rez_path = os.path.join(result_path, 'lemma.xml')
+    tree = ElementTree(root)
+    with open(rez_path, 'wb+') as f:
+        tree.write(f, xml_declaration=True, encoding='utf-8')
+
+
+def release_inflect_tests(items, result_path):
+    root = etree.Element('Tests')
+    for word in items:
+        test = etree.Element("T")
+        test.set('x', word['x_src'])
+        test.set('x_c', str(word['x_cls']))
+        test.set('y', word['y_src'])
+        test.set('y_c', str(word['y_cls']))
+        root.append(test)
+
+    rez_path = os.path.join(result_path, 'inflect.xml')
+    tree = ElementTree(root)
+    with open(rez_path, 'wb+') as f:
+        tree.write(f, xml_declaration=True, encoding='utf-8')
+
+
+def release_nn_tests():
+    res_path = os.path.join(ROOT, 'Network')
+    for gram in GRAM_TYPES:
+        items = load_datasets(gram, 'test')
+        cls = GRAM_TYPES[gram]['classes']
+        dic = {cls[g_key]['index']: g_key for g_key in cls}
+        release_gram_tests(items, gram, dic, res_path)
+
+    release_main_tests(load_datasets('main', 'test'), res_path, False)
+    release_lemma_tests(load_datasets('lemma', 'test'), res_path)
+    release_inflect_tests(load_datasets('inflect', 'test'), res_path)
+
+
+def merge_same_main(items):
+    rez_dict = {}
+    for item in items:
+        if item['src'] not in rez_dict:
+            rez_dict[item['src']] = []
+
+        rez_dict[item['src']].append(item)
+
+    rez_list = []
+    for text in rez_dict:
+        ys = [item['y'] for item in rez_dict[text]]
+        rez_list.append(dict(src=text, y=ys))
+    return rez_list
+
+
+def release_dictionary_tests():
+    res_path = os.path.join(ROOT, 'Dict')
+    with open(DICT_WORDS_PATH, 'rb') as f:
+        words = pickle.load(f)
+
+    lexeme_dict = {}
+    for word in words:
+        if word['post'] == 'numb':
+            continue
+
+        if word['id'] not in lexeme_dict:
+            lexeme_dict[word['id']] = []
+
+        lexeme_dict[word['id']].append(word)
+
+    main = []
+    inflect = []
+    lemmas = []
+    for word_id in lexeme_dict:
+        lexeme_words = lexeme_dict[word_id]
+        for item in lexeme_words:
+            main.append(dict(src=item['text'], y=item['main']))
+
+        for word in lexeme_words:
+            lemma = word['lemma'] if 'lemma' in word else word['text']
+            if is_lemma_dict[word['main']]:
+                lemma = word['text']
+
+            lemmas.append(dict(
+                x_src=word['text'],
+                main_cls=word['main'],
+                y_src=lemma
+            ))
+
+        un_cls_ids = []
+        rez_items = []
+        for item in sorted(lexeme_words, key=lambda x: x['index']):
+            if item['main'] in un_cls_ids:
+                continue
+
+            un_cls_ids.append(item['main'])
+            rez_items.append(item)
+        lexeme_words = rez_items
+
+        for i in range(0, len(lexeme_words) - 2):
+            main_word = lexeme_words[i]
+            for j in range(i, len(lexeme_words) - 1):
+                to_word = lexeme_words[j]
+                if to_word['main'] == main_word['main']:
+                    continue
+
+                inflect.append(dict(
+                    x_src=main_word['text'],
+                    x_cls=main_word['main'],
+                    y_src=to_word['text'],
+                    y_cls=to_word['main'],
+                    id=word_id
+                ))
+
+    main = merge_same_main(main)
+    release_main_tests(main, res_path)
+    release_lemma_tests(lemmas, res_path)
+    release_inflect_tests(inflect, res_path)
+
+
+def release_numb_tests():
+    res_path = os.path.join(ROOT, 'Numb')
+    main = []
+    inflect = []
+    lemmas = []
+    for val in numb_data['numbers']:
+        n_el = etree.Element("N")
+        n_el.set('v', str(val))
+        for tp in numb_data['numbers'][val]:
+            if tp == 'nar_end' or tp == 'lemma':
+                continue
+
+            items = numb_data['numbers'][val][tp]
+            lemma, _ = items[0]
+            for text, index in items:
+                main.append(dict(src=text, y=index))
+                lemmas.append(dict(
+                    x_src=text,
+                    main_cls=index,
+                    y_src=lemma
+                ))
+
+            un_cls_ids = []
+            rez_items = []
+            for item in items:
+                if item[1] in un_cls_ids:
+                    continue
+
+                un_cls_ids.append(item[1])
+                rez_items.append(item)
+            items = rez_items
+
+            for i in range(0, len(items) - 2):
+                main_text, main_index = items[i]
+                for j in range(i, len(items) - 1):
+                    to_text, to_index = items[j]
+                    inflect.append(dict(
+                        x_src=main_text,
+                        x_cls=main_index,
+                        y_src=to_text,
+                        y_cls=to_index,
+                        id=f"{val}{tp}"
+                    ))
+
+    main = merge_same_main(main)
+    release_main_tests(main, res_path)
+    release_lemma_tests(lemmas, res_path)
+    release_inflect_tests(inflect, res_path)
+
+
+def release_nar_numb_tests():
+    res_path = os.path.join(ROOT, 'NarNumb')
+    main = []
+    inflect = []
+    lemmas = []
+    for val in numb_data['numbers']:
+        n_el = etree.Element("N")
+        n_el.set('v', str(val))
+        items = numb_data['numbers'][val]['nar_end']
+        lemma_id = numb_data['numbers'][val]['p'][0][1]
+        lemma = f"{val}-{items[lemma_id]}"
+        for index in items:
+            text = f"{val}-{items[index]}"
+            main.append(dict(src=text, y=index))
+            lemmas.append(dict(
+                x_src=text,
+                main_cls=index,
+                y_src=lemma
+            ))
+
+        ids = list(items.keys())
+        for i in range(0, len(ids) - 2):
+            main_index = ids[i]
+            main_text = items[main_index]
+            main_text = f"{val}-{main_text}"
+            for j in range(i, len(items) - 1):
+                to_index = ids[j]
+                to_text = items[to_index]
+                to_text = f"{val}-{to_text}"
+                inflect.append(dict(
+                    x_src=main_text,
+                    x_cls=main_index,
+                    y_src=to_text,
+                    y_cls=to_index,
+                    id=f"{val}nar"
+                ))
+
+    main = merge_same_main(main)
+    release_main_tests(main, res_path)
+    release_lemma_tests(lemmas, res_path)
+    release_inflect_tests(inflect, res_path)
+
+
+def release_reg_tests():
+    res_path = os.path.join(ROOT, 'Reg')
+    int_tag = None
+    romn_tag = None
+    unkn_tag = None
+    punct_tag = None
+
+    for tag in tags:
+        if 'int' in tag:
+            int_tag = tags[tag]['i']
+        elif 'romn' in tag:
+            romn_tag = tags[tag]['i']
+        elif 'unkn' in tag:
+            unkn_tag = tags[tag]['i']
+        elif 'punct' in tag:
+            punct_tag = tags[tag]['i']
+
+    main = []
+    puncts = ['.', ',', '?', '!', '_', '"', '(', ')', ':', ';', '-']
+    for p in puncts:
+        main.append(dict(src=p, y=punct_tag))
+
+        text = [p]
+        for _ in range(1, RANDOM.randint(2, 5)):
+            text.append(puncts[RANDOM.randint(0, len(puncts)-1)])
+
+        text = ''.join(text)
+        main.append(dict(src=text, y=punct_tag))
+
+    for i in range(100):
+        val = RANDOM.randint(0, 1000000)
+        main.append(dict(src=str(val), y=int_tag))
+
+    roms = ['i', 'iii', 'iv', 'iv', 'c', 'd', 'm', 'md', 'mi']
+    for rom in roms:
+        main.append(dict(src=rom, y=romn_tag))
+        main.append(dict(src=rom.upper(), y=romn_tag))
+
+    unkn = ['test', 'sdasdas', 'home']
+    for v in unkn:
+        main.append(dict(src=v, y=unkn_tag))
+        main.append(dict(src=v.upper(), y=unkn_tag))
+
+    lemmas = [dict(x_src=item['src'], main_cls=item['y'], y_src=item['src']) for item in main]
+    inflect = [dict(x_src=item['src'], x_cls=item['y'], y_src=item['src'], y_cls=item['y'], id=item['src']) for item in main]
+
+    main = merge_same_main(main)
+    release_main_tests(main, res_path)
+    release_lemma_tests(lemmas, res_path)
+    release_inflect_tests(inflect, res_path)
+
+
+release_reg_tests()
+release_nar_numb_tests()
+release_numb_tests()
+release_dictionary_tests()
+release_nn_tests()

12_release_model.py

@@ -0,0 +1,316 @@
+import logging
+import os
+import pickle
+from shutil import copyfile
+
+from xml.etree.ElementTree import ElementTree
+from lxml import etree
+
+from model import RNN
+from utils import CONFIG
+from tester import Tester
+
+
+class Releaser:
+    def __init__(self):
+        self.config = CONFIG
+        self.dataset_path = self.config['dict_path']
+        self.model_key = self.config['model_key']
+        self.chars = self.config['chars']
+        self.gram_types = self.config['grammemes_types']
+        self.rnn = RNN(True)
+        self.tester = Tester()
+        self.pd_publish_paths = [
+            os.path.join(path, f"frozen_model_{self.model_key}.pb")
+            for path in self.config['publish_net_paths']
+        ]
+        self.xml_publish_paths = [
+            os.path.join(path, f"release_{self.model_key}.xml")
+            for path in self.config['publish_net_paths']
+        ]
+        self.xml_gram_paths = [
+            os.path.join(path, "grams.xml")
+            for path in self.config['publish_gramm_paths']
+        ]
+        self.xml_numbers_paths = [
+            os.path.join(path, "numbers.xml")
+            for path in self.config['publish_numbers_paths']
+        ]
+        self.xml_tags_paths = [
+            os.path.join(path, "tags.xml")
+            for path in self.config['publish_tags_paths']
+        ]
+        self.test_result_paths = [
+            os.path.join(path, "test_info.txt")
+            for path in self.config['test_results_paths']
+        ]
+        self.publish_dataset_info_paths = [
+            os.path.join(path, "dataset_info.txt")
+            for path in self.config['publish_dataset_info_paths']
+        ]
+        self.public_inflect_templates_paths = [
+            os.path.join(path, "inflect_templates.xml")
+            for path in self.config['public_inflect_templates_paths']
+        ]
+        self.classes_dic = self.config['main_classes']
+        self.rev_classes_dic = {
+            self.classes_dic[key]: ",".join([key for key in list(key) if key is not None])
+            for key in self.classes_dic
+        }
+        with open(CONFIG['tags_path'], 'rb') as f:
+            self.tags = pickle.load(f)
+
+        with open(CONFIG['numb_data_path'], 'rb') as f:
+            self.numb_data = pickle.load(f)
+
+        with open(self.config['inflect_templates_path'], 'rb') as f:
+            self.inflect_templates = pickle.load(f)
+
+    def release_model(self):
+        pd_release_path, gram_ops, out_ops = self.rnn.release()
+        for path in self.pd_publish_paths:
+            copyfile(pd_release_path, path)
+
+        self.__release_test_metrics__()
+        self.__release_numbers_xml__()
+        self.__release_gramm_docs__()
+        self.__release_inflect_docs__()
+        self.__release_grams_xml__()
+        self.__release_tags_xml__()
+        self.__release_dataset_info__()
+        self.__release_model_xml__(out_ops, gram_ops)
+
+    def __release_test_metrics__(self):
+        results = self.tester.test()
+        for path in self.test_result_paths:
+            with open(path, 'w+') as f:
+                f.write(results)
+
+    def __release_model_xml__(self, out_ops, gram_ops):
+        root = etree.Element('Root')
+        for key in out_ops:
+            root.set(key, out_ops[key])
+
+        chars_el = etree.Element('Chars')
+        chars_el.set("start_char", str(self.config['start_token']))
+        chars_el.set("end_char", str(self.config['end_token']))
+        for index, value in enumerate(self.chars):
+            char_el = etree.Element("Char")
+            char_el.set('index', str(index))
+            char_el.set('value', value)
+            chars_el.append(char_el)
+        root.append(chars_el)
+
+        grams_el = etree.Element('Grams')
+        for gram in self.gram_types:
+            gram_el = etree.Element("G")
+            gram_el.set('key', gram)
+            gram_el.set('op', gram_ops[gram]['prob'])
+            grams_el.append(gram_el)
+        root.append(grams_el)
+
+        inflect_el = etree.Element("Inflect")
+        for main_key in self.inflect_templates:
+            temp_el = etree.Element("Im")
+            temp_el.set('i', str(self.classes_dic[main_key]))
+            inflect_el.append(temp_el)
+            for form in self.inflect_templates[main_key]:
+                form_el = etree.Element("I")
+                form_el.set('i', str(self.classes_dic[form]))
+                temp_el.append(form_el)
+        root.append(inflect_el)
+
+        tree = ElementTree(root)
+        for path in self.xml_publish_paths:
+            with open(path, 'wb+') as f:
+                tree.write(f, xml_declaration=True, encoding='utf-8')
+
+        logging.info("Model released")
+
+    def __release_grams_xml__(self):
+        nn_types = self.config['grammemes_types']
+        dict_post_types = self.config['dict_post_types']
+        other_types = self.config['other_post_types']
+        root = etree.Element('Grams')
+        for gram in nn_types:
+            gram_el = etree.Element("G")
+            gram_el.set('index', str(nn_types[gram]['index']))
+            gram_el.set('key_en', gram)
+            gram_el.set('key_ru', nn_types[gram]['key_ru'])
+            root.append(gram_el)
+            gr_dic = nn_types[gram]['classes']
+            for key_en in gr_dic:
+                item = gr_dic[key_en]
+                cls_el = etree.Element("C")
+                cls_el.set('key_en', key_en)
+                cls_el.set('key_ru', str(item['key_ru']))
+                cls_el.set('nn_index', str(item['index']))
+                gram_el.append(cls_el)
+
+            if gram == "post":
+                for key_en in dict_post_types:
+                    item = dict_post_types[key_en]
+                    cls_el = etree.Element("C")
+                    cls_el.set('key_en', key_en)
+                    cls_el.set('key_ru', str(item['key_ru']))
+                    gram_el.append(cls_el)
+
+                for key_en in other_types:
+                    item = other_types[key_en]
+                    cls_el = etree.Element("C")
+                    cls_el.set('key_en', key_en)
+                    cls_el.set('key_ru', str(item['key_ru']))
+                    gram_el.append(cls_el)
+
+        tree = ElementTree(root)
+        for path in self.xml_gram_paths:
+            with open(path, 'wb+') as f:
+                tree.write(f, xml_declaration=True, encoding='utf-8')
+
+    def __release_numbers_xml__(self):
+        root = etree.Element('NumbData')
+        root.set("reg", self.numb_data['regex'])
+        root.set("l", ','.join([str(i) for i in self.numb_data['lemma_cls_ids']]))
+        for val in self.numb_data['numbers']:
+            n_el = etree.Element("N")
+            n_el.set('v', str(val))
+            for tp in self.numb_data['numbers'][val]:
+                if tp == 'nar_end' or tp == 'lemma':
+                    continue
+
+                for tpl in self.numb_data['numbers'][val][tp]:
+                    w_el = etree.Element("W")
+                    w_el.set('t', tpl[0])
+                    w_el.set('i', str(tpl[1]))
+                    w_el.set('k', tp)
+                    n_el.append(w_el)
+
+            nar_ends = self.numb_data['numbers'][val]['nar_end']
+            for cls in nar_ends:
+                w_el = etree.Element("E")
+                w_el.set('t', nar_ends[cls])
+                w_el.set('i', str(cls))
+                n_el.append(w_el)
+
+            root.append(n_el)
+
+        tree = ElementTree(root)
+        for path in self.xml_numbers_paths:
+            with open(path, 'wb+') as f:
+                tree.write(f, xml_declaration=True, encoding='utf-8')
+
+    def __release_tags_xml__(self):
+        root = etree.Element('Tags')
+        for tag in self.tags:
+            val = self.tags[tag]
+            cls_el = etree.Element("T")
+            cls_el.set('i', str(val['i']))
+            cls_el.set('v', ",".join([key if key is not None else '' for key in tag]))
+            cls_el.set('p', val['p'])
+            cls_el.set('o', str(val['o']))
+
+            if val['l']:
+                cls_el.set('l', '1')
+            root.append(cls_el)
+
+        tree = ElementTree(root)
+        for path in self.xml_tags_paths:
+            with open(path, 'wb+') as f:
+                tree.write(f, xml_declaration=True, encoding='utf-8')
+
+    def __release_dataset_info__(self):
+        doc = etree.iterparse(self.dataset_path, events=('start', 'end'))
+        itr = iter(doc)
+        event, element = next(itr)
+        while not (event == 'start' and element.tag == 'dictionary'):
+            pass
+
+        version = element.attrib['version']
+        revision = element.attrib['revision']
+        for path in self.publish_dataset_info_paths:
+            with open(path, 'w+') as f:
+                f.write(f"dictionary\nversion={version}\nrevision={revision}")
+
+    def __release_gramm_docs__(self):
+        mds = [
+            "# Поддерживамые грамматические категории и граммемы",
+            "В DeepMorphy используется слегка измененное подмножество граммем и грамматичеких категорий из словарей [OpenCorpora](http://opencorpora.org/dict.php?act=gram)."
+        ]
+
+        for gram_cat_key in self.gram_types:
+            gram_cat = self.gram_types[gram_cat_key]
+            mds.append(f"- **{gram_cat['name'].capitalize()}** (ru='{gram_cat['key_ru']}', en='{gram_cat_key}') :")
+
+            classes = dict(gram_cat['classes'])
+            if gram_cat_key == 'post':
+                classes.update(self.config['dict_post_types'])
+                classes.update(self.config['other_post_types'])
+
+            for gram in classes:
+                gram_obj = classes[gram]
+                mds.append(f"    - {gram_obj['name_ru']} (ru='{gram_obj['key_ru']}',en='{gram}')")
+
+        mds = "\n".join(mds)
+        with open(self.config['publish_gram_doc_path'], 'w+') as f:
+            f.write(mds)
+
+    def __release_inflect_docs__(self):
+        mds = [
+            "# ��писок поддерживаемых словоизменений",
+            "Словоизменение возможно только в рамках выделенных жирным категорий:"
+        ]
+
+        post_index = self.gram_types['post']['index']
+        gndr_index = self.gram_types['gndr']['index']
+
+        en_ru_dict = {}
+        for gram_cat in self.gram_types:
+            for cls in self.gram_types[gram_cat]['classes']:
+                cls_data = self.gram_types[gram_cat]['classes'][cls]
+                en_ru_dict[cls] = cls_data['key_ru']
+
+        def create_tag_text(tag):
+            tag_text = [en_ru_dict[key] for key in list(tag) if key is not None]
+            tag_text = ",".join(tag_text)
+            return f"    - {tag_text}"
+
+        for main_tpl in sorted(self.inflect_templates):
+            post = main_tpl[post_index]
+            gndr = main_tpl[gndr_index]
+            if post == "infn":
+                header_text = "Глаголы и глагольные формы"
+            elif post == "adjf":
+                header_text = "Прилагательные"
+            elif post == "noun" and gndr == 'masc':
+                header_text = "Существительные мужского рода"
+            elif post == "noun" and gndr == 'femn':
+                header_text = "Существительные женского рода"
+            elif post == "noun" and gndr == 'neut':
+                header_text = "Существительные среднего рода"
+            elif post == "noun" and gndr == 'msf':
+                header_text = "Существительные общего рода"
+            else:
+                raise NotImplemented()
+
+            mds.append(f"- **{header_text}**:")
+            items = [(item, self.tags[item]['o']) for item in self.inflect_templates[main_tpl]]
+            items.append((main_tpl, self.tags[main_tpl]['o']))
+            tags = sorted(items, key=lambda x: x[1], reverse=True)
+            for tag in tags:
+                mds.append(create_tag_text(tag[0]))
+
+        mds = "\n".join(mds)
+        with open(self.config['publish_inflect_doc_path'], 'w+') as f:
+            f.write(mds)
+
+    @staticmethod
+    def __build_bad_words__(tester):
+        words = tester.get_bad_words()
+        logging.info(f"Wrong words count {len(words)}")
+        with open(os.path.join("wrong_words.pkl"), 'wb+') as f:
+            pickle.dump(words, f)
+
+
+if __name__ == "__main__":
+    tester = Releaser()
+    tester.release_model()

1_load_dictionary.py

@@ -0,0 +1,217 @@
+import logging
+import pickle
+from tqdm import tqdm
+from lxml import etree
+from utils import get_grams_info, CONFIG
+
+
+DIC_PATH = CONFIG['dict_path']
+MAX_WORD_SIZE = CONFIG['max_word_size']
+DATASET_WORDS_PATH = CONFIG['dataset_words_path']
+DICTS_WORDS_PATH = CONFIG['dict_words_path']
+DICT_POST_TYPES = CONFIG['dict_post_types']
+LEMMAS_PROPS = CONFIG['lemma_same_word']
+AD_TAGS = CONFIG['ad_tags']
+SRC_CONVERT, _ = get_grams_info(CONFIG)
+i = 0
+
+
+def parse_words(itr):
+    global i
+
+    cur_word = None
+    cur_item = None
+    event, element = next(itr)
+
+    while not (event == 'end' and element.tag == 'lemmata'):
+
+        if event == 'start' and element.tag == 'lemma':
+            cur_word = {
+                'id': element.attrib['id'],
+                'lemma': None,
+                'forms': []
+            }
+
+        if event == 'start' and (element.tag == 'l' or element.tag == 'f'):
+            cur_item = {'text': None, 'index': i}
+            i += 1
+
+        if event == 'end' and element.tag == 'l':
+            cur_item['text'] = element.attrib['t']
+            cur_word['lemma'] = cur_item
+            cur_item = None
+
+        if event == 'end' and element.tag == 'g' and element.attrib['v'].lower() in SRC_CONVERT:
+            src_key = element.attrib['v'].lower()
+            gram_type, gram = SRC_CONVERT[src_key]
+            cur_item[gram_type] = gram
+        elif event == 'end' and element.tag == 'g' \
+                and element.attrib['v'].lower() in AD_TAGS \
+                and 'ad_tags' not in cur_item:
+            cur_item['ad_tags'] = [element.attrib['v'].lower()]
+        elif event == 'end' and element.tag == 'g' and element.attrib['v'].lower() in AD_TAGS:
+            cur_item['ad_tags'].append(element.attrib['v'].lower())
+
+        if event == 'end' and element.tag == 'f':
+            cur_item['text'] = element.attrib['t']
+            cur_word['forms'].append(cur_item)
+            cur_item = None
+
+        if event == 'end' and element.tag == 'lemma':
+            yield cur_word
+            cur_word = None
+
+        event, element = next(itr)
+
+
+def get_flat_words(words):
+    for item in words:
+        lemma = item['lemma']
+        lemma['lemma'] = lemma['text']
+        lemma['id'] = item['id']
+        for form in item['forms']:
+            word = dict(lemma)
+            for key in form:
+                word[key] = form[key]
+
+            if 'ad_tags' in word:
+                word['ad_tags'] = ','.join(word['ad_tags'])
+
+            yield word
+
+
+def parse_link_types(itr):
+    event, element = next(itr)
+    link_types = {}
+
+    while not (event == 'start' and element.tag == 'link_types'):
+        event, element = next(itr)
+
+    while not (event == 'end' and element.tag == 'link_types'):
+
+        if event == 'end' and element.tag == 'type':
+            link_types[element.text] = element.attrib['id']
+
+        event, element = next(itr)
+    return link_types
+
+
+def parse_links(itr):
+    event, element = next(itr)
+    while not (event == 'end' and element.tag == 'links'):
+        if event == 'end' and element.tag == 'link':
+            yield {
+                'from': element.attrib['from'],
+                'to': element.attrib['to'],
+                'type': element.attrib['type']
+            }
+        event, element = next(itr)
+
+
+def set_lemma_and_inflect_id(words, link_types, links):
+    same_inflect_id_post = ['noun', 'adjf', 'infn']
+    lemmas_dict = {}
+    for word in words:
+        for norm_f in LEMMAS_PROPS:
+            is_lemma = True
+            for key in norm_f:
+                if key in word and word[key] != norm_f[key]:
+                    is_lemma = False
+                    break
+
+            if is_lemma:
+                lemmas_dict[word['id']] = word['text']
+                del word['lemma']
+                break
+
+        if 'lemma' in word and word['post'] in same_inflect_id_post:
+            word['inflect_id'] = word['id']
+
+    inv_link_type_dict = {
+        link_types[key]: key
+        for key in link_types
+    }
+
+    links = {
+        (link['to'], inv_link_type_dict[link['type']]): link['from']
+        for link in links
+    }
+
+    prtf_dict = {}
+    for word in words:
+        link_type = None
+        if word['post'] == 'verb':
+            link_type = 'INFN-VERB'
+        elif word['post'] == 'prtf':
+            link_type = 'INFN-PRTF'
+        elif word['post'] == 'grnd':
+            link_type = 'INFN-GRND'
+        elif word['post'] == 'adjs':
+            link_type = 'ADJF-ADJS'
+        elif word['post'] == 'comp':
+            link_type = 'ADJF-COMP'
+
+        if not link_type:
+            continue
+
+        key = (word['id'], link_type)
+        if key in links and links[key] in lemmas_dict:
+            lemma = lemmas_dict[links[key]]
+            word['lemma'] = lemma
+            lemmas_dict[word['id']] = lemma
+        else:
+            del word['lemma']
+
+        if key in links:
+            word['inflect_id'] = links[key]
+
+        if key in links and word['post'] == 'prtf':
+            prtf_dict[word['id']] = word['inflect_id']
+
+    for word in words:
+        if word['post'] != 'prts':
+            continue
+
+        key = (word['id'], 'PRTF-PRTS')
+        if key in links:
+            lemma = lemmas_dict[links[key]]
+            word['lemma'] = lemma
+            word['inflect_id'] = prtf_dict[links[key]]
+            lemmas_dict[word['id']] = lemma
+        else:
+            del word['lemma']
+
+
+doc = etree.iterparse(DIC_PATH, events=('start', 'end'))
+itr = iter(doc)
+event, element = next(itr)
+logging.info("Parsing dictionary xml")
+while not (event == 'start' and element.tag == 'lemmata'):
+    event, element = next(itr)
+
+words = list(parse_words(itr))
+link_types = parse_link_types(itr)
+links = list(parse_links(itr))
+words = list(get_flat_words(words))
+set_lemma_and_inflect_id(words, link_types, links)
+words = [dict(t) for t in {tuple(sorted(d.items())) for d in words}]
+dict_words = [word for word in words if word['post'] in DICT_POST_TYPES]
+dataset_words = [word for word in words if word['post'] not in DICT_POST_TYPES]
+dataset_words_dic = {}
+for word in tqdm(dataset_words):
+    w_len = len(word['text'])
+    if w_len > MAX_WORD_SIZE:
+        continue
+
+    if word['text'] not in dataset_words_dic:
+        dataset_words_dic[word['text']] = []
+    dataset_words_dic[word['text']].append(word)
+
+logging.info(f"Dict words: {len(dict_words)}")
+logging.info(f"Dataset words: {len(dataset_words_dic)}")
+
+with open(DATASET_WORDS_PATH, 'wb+') as f:
+    pickle.dump(dataset_words_dic, f)
+
+with open(DICTS_WORDS_PATH, 'wb+') as f:
+    pickle.dump(dict_words, f)

2_vectorize.py

@@ -0,0 +1,43 @@
+import pickle
+import numpy as np
+from tqdm import tqdm
+from utils import CONFIG
+
+CHARS = CONFIG['chars']
+END_TOKEN = CONFIG['end_token']
+MAX_WORD_SIZE = CONFIG['max_word_size']
+WORDS_PATH = CONFIG['dataset_words_path']
+VECT_PATH = CONFIG['vect_words_path']
+CHARS_INDEXES = {c: index for index, c in enumerate(CHARS)}
+
+
+def vectorize_text(text):
+    word_vect = np.full((MAX_WORD_SIZE,), END_TOKEN, dtype=np.int32)
+    for index, c in enumerate(text):
+        if c in CHARS:
+            word_vect[index] = CHARS_INDEXES[c]
+        else:
+            word_vect[index] = CHARS_INDEXES["UNDEFINED"]
+
+    seq_len = len(text)
+    return word_vect, seq_len
+
+
+def vectorize_words(words_dic):
+    vect_dic = {}
+    for word in tqdm(words_dic, desc="Vectorizing words"):
+        vect_dic[word] = {
+            'vect': vectorize_text(word),
+            'forms': words_dic[word]
+        }
+
+    return vect_dic
+
+
+with open(WORDS_PATH, 'rb') as f:
+    words_dic = pickle.load(f)
+
+vec_words = vectorize_words(words_dic)
+
+with open(VECT_PATH, 'wb+') as f:
+    pickle.dump(vec_words, f)

3_cls_dataset.py

@@ -0,0 +1,85 @@
+import os
+import pickle
+import numpy as np
+from tqdm import tqdm
+from sklearn.preprocessing import normalize
+from collections import defaultdict
+from utils import get_grams_info, CONFIG, save_dataset
+
+
+VECT_PATH = CONFIG['vect_words_path']
+CLS_CLASSES_PATH = CONFIG['cls_classes_path']
+GRAMMEMES_TYPES = CONFIG['grammemes_types']
+SRC_CONVERT, CLASSES_INDEXES = get_grams_info(CONFIG)
+
+
+def generate_dataset(vec_words, cls_type, cls_dic):
+    ordered_keys = [cls for cls in sorted(cls_dic, key=lambda cls: cls_dic[cls])]
+    weights = [0 for key in ordered_keys]
+    for word in tqdm(vec_words, desc=f"Calculating {cls_type} weights"):
+        for form in vec_words[word]['forms']:
+            if cls_type in form:
+                i = cls_dic[form[cls_type]]
+                weights[i] = weights[i] + 1
+
+    weights = normalize(np.asarray(weights).reshape(1, -1))
+    weights = np.ones((len(ordered_keys),)) - weights
+
+    rez_items = defaultdict(list)
+    cur_cls = None
+    for word in tqdm(vec_words, desc=f"Generating classification {cls_type} dataset"):
+        y = np.zeros((len(ordered_keys),), dtype=np.int)
+        has_classes = False
+        for form in vec_words[word]['forms']:
+            if cls_type in form:
+                cur_cls = form[cls_type]
+                index = cls_dic[cur_cls]
+                y[index] = 1
+                has_classes = True
+
+        if has_classes:
+            items = rez_items[cur_cls]
+            items.append({
+                'src': word,
+                'x': vec_words[word]['vect'],
+                'y': y,
+                'weight': weights.reshape(-1, 1)[y == 1].max()
+            })
+            rez_items[cur_cls] = items
+
+    save_dataset(rez_items, cls_type)
+
+
+def generate_all(vec_words):
+    for cls_type in CLASSES_INDEXES:
+        cls_dic = CLASSES_INDEXES[cls_type]
+        generate_dataset(vec_words, cls_type, cls_dic)
+
+    un_classes = []
+    for word in tqdm(vec_words, desc="Setting main class"):
+        for form in vec_words[word]['forms']:
+            tpl = tuple(
+                form[key] if key in form else None
+                for key in GRAMMEMES_TYPES
+            )
+            if tpl not in un_classes:
+                un_classes.append(tpl)
+            form['main'] = tpl
+
+    with open(VECT_PATH, 'wb+') as f:
+        pickle.dump(vec_words, f)
+
+    cls_dic = {
+        tpl: index
+        for index, tpl in enumerate(un_classes)
+    }
+    generate_dataset(vec_words, 'main', cls_dic)
+    print(f"Main classes count: {len(cls_dic)}")
+    with open(CLS_CLASSES_PATH, 'wb+') as f:
+        pickle.dump(cls_dic, f)
+
+
+with open(VECT_PATH, 'rb') as f:
+    vwords = pickle.load(f)
+
+generate_all(vwords)

4_lemma_dataset.py

@@ -0,0 +1,70 @@
+import pickle
+from tqdm import tqdm
+from collections import defaultdict
+from utils import CONFIG, save_dataset, save_dictionary_items
+
+
+MIN_WORD_SIZE = CONFIG['min_word_size']
+PREFIX_FILTER_LENGTH = CONFIG['prefix_filter_length']
+VECT_PATH = CONFIG['vect_words_path']
+CLS_CLASSES_PATH = CONFIG['cls_classes_path']
+DICT_WORDS_PATH = CONFIG['dics_path']
+
+
+def generate(vec_words, main_cls_dic):
+    dict_words = []
+    rez_dict = defaultdict(list)
+    for word in tqdm(vec_words, desc="Generating lemma dataset"):
+        dic = vec_words[word]
+        x_vec = dic['vect']
+
+        for form in dic['forms']:
+            main_cls = main_cls_dic[form['main']]
+
+            if 'lemma' in form:
+                word_y = form['lemma']
+            else:
+                continue
+
+            if word_y not in vec_words \
+               or MIN_WORD_SIZE > len(word_y) \
+               or MIN_WORD_SIZE > len(word):
+                continue
+
+            if word_y[:PREFIX_FILTER_LENGTH] != word[:PREFIX_FILTER_LENGTH] \
+                    and word_y[:PREFIX_FILTER_LENGTH].replace('ё', 'е') != word[:PREFIX_FILTER_LENGTH].replace('ё', 'е')\
+                    and form['post'] != 'comp':
+                #tqdm.write('Word to dictionary: {0} -> {1}'.format(word, word_y))
+                dict_words.append(dict(
+                    text=word,
+                    text_y=word_y,
+                    main=main_cls,
+                    id=form['inflect_id']
+                ))
+                continue
+
+            y_vec = vec_words[word_y]['vect']
+            items = rez_dict[main_cls]
+            items.append({
+                'id': form['inflect_id'],
+                'x_src': word,
+                'x': x_vec[0],
+                'x_len': x_vec[1],
+                'y_src': word_y,
+                'y': y_vec[0],
+                'y_len': y_vec[1],
+                'main_cls': main_cls
+            })
+            rez_dict[main_cls] = items
+
+    save_dataset(rez_dict, 'lemma')
+    save_dictionary_items(dict_words, 'lemma')
+
+
+with open(VECT_PATH, 'rb') as f:
+    vwords = pickle.load(f)
+
+with open(CLS_CLASSES_PATH, 'rb') as f:
+    cls_dic = pickle.load(f)
+
+generate(vwords, cls_dic)

5_inflect_dataset.py

@@ -0,0 +1,118 @@
+import pickle
+from tqdm import tqdm
+
+from utils import CONFIG, save_dataset
+
+
+MIN_WORD_SIZE = CONFIG['min_word_size']
+PREFIX_FILTER_LENGTH = CONFIG['prefix_filter_length']
+VECT_PATH = CONFIG['vect_words_path']
+CLS_CLASSES_PATH = CONFIG['cls_classes_path']
+GRAMMEMES_TYPES = CONFIG['grammemes_types']
+TEMPLATES_PATH = CONFIG['inflect_templates_path']
+IGNORE_TAGS = CONFIG['inflect_ignore_tags']
+
+
+def create_forms_dict(vect_words):
+    root_posts = ['noun', 'infn', 'adjf']
+    forms_dict = {}
+    for word in vect_words:
+        for form in vect_words[word]['forms']:
+            is_main = 'inflect_id' not in form
+            key = form['id'] if is_main else form['inflect_id']
+            if key not in forms_dict:
+                forms_dict[key] = dict(root=None, items=[])
+
+            form_dict = forms_dict[key]
+            if is_main:
+                form_dict['root'] = form
+            else:
+                form_dict['items'].append(form)
+
+    forms_dict = {
+        key: forms_dict[key]
+        for key in forms_dict if forms_dict[key]['root'] is not None and forms_dict[key]['root']['post'] in root_posts
+    }
+    return forms_dict
+
+
+def create_templates(forms_dict):
+    templates = dict()
+    for key in forms_dict:
+        item = forms_dict[key]
+        root = item['root']
+        if root['main'] not in templates:
+            templates[root['main']] = set()
+
+        for form in item['items']:
+            templates[root['main']].add(form['main'])
+
+    return templates
+
+
+def generate_dataset(forms_dict, vect_words, cls_dic):
+    rez_dict = {}
+    for key in tqdm(forms_dict, desc="Generating dataset"):
+        item = forms_dict[key]
+        root = item['root']
+        x_cls = cls_dic[root['main']]
+        x, x_len = vect_words[root['text']]['vect']
+        prefix_filter = root['text'][:PREFIX_FILTER_LENGTH]
+        prefix_filter_e = prefix_filter.replace('ё', 'е')
+        if MIN_WORD_SIZE > len(root['text']):
+            continue
+
+        form_dict = {}
+        for form in item['items']:
+            if MIN_WORD_SIZE > len(form['text']):
+                continue
+
+            if 'ad_tags' in form and any([tag for tag in IGNORE_TAGS if tag in form['ad_tags']]):
+                #tqdm.write("Ignore form {0} for {1} by tags {2}".format(form['text'], root['text'], form['ad_tags']))
+                continue
+
+            if not (form['text'].startswith(prefix_filter) or
+                    form['text'].replace('ё', 'е').startswith(prefix_filter_e)):
+                #tqdm.write("Ignore form {0} for {1}".format(form['text'], root['text']))
+                continue
+
+            y_cls = cls_dic[form['main']]
+            if y_cls in form_dict and form_dict[y_cls]['index'] < form['index']:
+                #tqdm.write("Ignore duplicate form {0} [{1}] for {2} ".format(form['text'], form_dict[y_cls]['text'], root['text']))
+                continue
+
+            form_dict[y_cls] = form
+
+        for y_cls in form_dict:
+            form = form_dict[y_cls]
+            y, y_len = vect_words[form['text']]['vect']
+            if y_cls not in rez_dict:
+                rez_dict[y_cls] = []
+
+            rez_dict[y_cls].append(dict(
+                id=form['inflect_id'],
+                x_src=root['text'],
+                x=x,
+                x_cls=x_cls,
+                x_len=x_len,
+                y_src=form['text'],
+                y=y,
+                y_cls=y_cls,
+                y_len=y_len
+            ))
+
+    save_dataset(rez_dict, 'inflect')
+
+
+with open(VECT_PATH, 'rb') as f:
+    vwords = pickle.load(f)
+
+with open(CLS_CLASSES_PATH, 'rb') as f:
+    cls_dic = pickle.load(f)
+
+forms_dict = create_forms_dict(vwords)
+templates = create_templates(forms_dict)
+with open(TEMPLATES_PATH, 'wb+') as f:
+    pickle.dump(templates, f)
+
+generate_dataset(forms_dict, vwords, cls_dic)

6_train.py

@@ -0,0 +1,3 @@
+from model import RNN
+rnn = RNN(False)
+rnn.train()

7_build_numbers.py

@@ -0,0 +1,88 @@
+import yaml
+import pickle
+from utils import CONFIG, create_cls_tuple
+
+
+VECT_PATH = CONFIG['vect_words_path']
+CLS_CLASSES_PATH = CONFIG['cls_classes_path']
+NMB_CLASSES_PATH = CONFIG['numb_classes_path']
+NMB_DATA_PATH = CONFIG['numb_data_path']
+SOGL_CHARS = ['б', 'в', 'г', 'д', 'ж', 'з', 'й', 'к', 'л', 'м', 'н', 'п', 'р', 'с', 'т', 'ф', 'х', 'ц', 'ч', 'ш', 'щ']
+GLASN_CHARS = ['а', 'о', 'и', 'е', 'ё', 'э', 'ы', 'у', 'ю', 'я']
+
+
+with open(CLS_CLASSES_PATH, 'rb') as f:
+    cur_classes_count = len(pickle.load(f)) + 1
+
+with open('numb.yml') as f:
+    numbr_src_dic = yaml.load(f)
+
+
+def get_nar_end(text):
+    end = text[-1]
+    if text[-2] in SOGL_CHARS and text[-1] in GLASN_CHARS:
+        end = text[-2:]
+
+    return end
+
+
+lemma_cls_ids = set()
+res_dict = {}
+numb_cls_dict = {}
+for n_key in numbr_src_dic:
+    n_key_data = {
+        'nar_end': {}
+    }
+    res_dict[n_key] = n_key_data
+    for t in numbr_src_dic[n_key]:
+        lemma_text = None
+        lemma_number_text = None
+
+        for index, item in enumerate(numbr_src_dic[n_key][t]):
+            item['post'] = 'numb'
+            cls_tpl = create_cls_tuple(item)
+            if cls_tpl not in numb_cls_dict:
+                numb_cls_dict[cls_tpl] = cur_classes_count
+                cur_classes_count += 1
+
+            cur_class = numb_cls_dict[cls_tpl]
+            if index == 0 and cls_tpl not in lemma_cls_ids:
+                lemma_cls_ids.add(cur_class)
+
+            if t not in n_key_data:
+                n_key_data[t] = []
+
+            items = n_key_data[t]
+            items.append((item['text'], cur_class))
+            if t == 'p' and cur_class not in n_key_data['nar_end']:
+                end = get_nar_end(item['text'])
+                n_key_data['nar_end'][cur_class] = end
+
+regex = []
+for val in res_dict:
+    cur_group = []
+    for key in res_dict[val]:
+        if key == 'nar_end' or key == 'lemma':
+            continue
+
+        for tpl in res_dict[val][key]:
+            cur_group.append(tpl[0])
+
+    cur_group = list(set(cur_group))
+    cur_group = '|'.join(cur_group)
+    cur_group = f'(?<_{val}>{cur_group})'
+    regex.insert(0, cur_group)
+
+regex = '|'.join(regex)
+regex = f"^({regex})+$"
+
+
+with open(NMB_CLASSES_PATH, 'wb+') as f:
+    pickle.dump(numb_cls_dict, f)
+
+with open(NMB_DATA_PATH, 'wb+') as f:
+    pickle.dump({
+        'regex': regex,
+        'lemma_cls_ids': list(lemma_cls_ids),
+        'numbers': res_dict
+    }, f)

8_generate_tags.py

@@ -0,0 +1,102 @@
+import pickle
+from utils import CONFIG, create_cls_tuple
+
+
+def set_order(items):
+    index = 0
+    order_dict = {}
+    for post in CONFIG['dict_post_types']:
+        order_dict[post] = index
+        index += 1
+
+    for post in CONFIG['other_post_types']:
+        order_dict[post] = index
+        index += 1
+
+    for gram_cat in GRAM_TYPES:
+        for gram in GRAM_TYPES[gram_cat]['classes']:
+            order_dict[gram] = index
+            index += 1
+
+    order_items = []
+    for tpl in items:
+        order_key = tuple([order_dict[gram] if gram is not None else 1024 for gram in list(tpl)])
+        order_items.append((tpl, order_key))
+
+    for index, tpl in enumerate(sorted(order_items, key=lambda x: x[1], reverse=False)):
+        items[tpl[0]]['o'] = index
+
+
+GRAM_TYPES = CONFIG['grammemes_types']
+with open(CONFIG['numb_classes_path'], 'rb') as f:
+    numb_classes_dic = pickle.load(f)
+with open(CONFIG['numb_data_path'], 'rb') as f:
+    numb_data = pickle.load(f)
+
+lemma_same_words = []
+for cls in CONFIG['lemma_same_word']:
+    lemma_same_words.append(create_cls_tuple(cls))
+
+items = {}
+classes_dic = CONFIG['main_classes']
+for tpl in classes_dic:
+    items[tpl] = {
+        'i': classes_dic[tpl],
+        'p': 'nn',
+        'l':  tpl in lemma_same_words
+    }
+
+for tpl in numb_classes_dic:
+    cls_index = numb_classes_dic[tpl]
+    items[tpl] = {
+        'i': cls_index,
+        'p': 'numb',
+        'l': cls_index in numb_data['lemma_cls_ids']
+    }
+
+max_cls_id = max([numb_classes_dic[key] for key in numb_classes_dic])
+max_cls_id += 1
+post_index = GRAM_TYPES['post']['index']
+for key in CONFIG['other_post_types']:
+    tpl = [None for item in CONFIG['grammemes_types']]
+    tpl[post_index] = key
+    tpl = tuple(tpl)
+    items[tpl] = {
+        'i': max_cls_id,
+        'p': 'reg',
+        'l':  max_cls_id
+    }
+    max_cls_id += 1
+
+tpl = [None for item in CONFIG['grammemes_types']]
+tpl[post_index] = 'unkn'
+tpl = tuple(tpl)
+items[tpl] = {
+    'i': max_cls_id,
+    'p': 'reg',
+    'l':  max_cls_id
+}
+max_cls_id += 1
+
+with open(CONFIG['dict_words_path'], 'rb') as f:
+    dic_words = pickle.load(f)
+
+for dic_item in dic_words:
+    cls_tpl = create_cls_tuple(dic_item)
+    if cls_tpl not in items:
+        items[cls_tpl] = {
+            'i': max_cls_id,
+            'p': 'dict',
+            'l': ('npro' in cls_tpl and 'nomn' in cls_tpl) or cls_tpl in lemma_same_words
+        }
+        max_cls_id += 1
+
+    dic_item['main'] = items[cls_tpl]['i']
+
+with open(CONFIG['dict_words_path'], 'wb+') as f:
+    pickle.dump(dic_words, f)
+
+set_order(items)
+
+with open(CONFIG['tags_path'], 'wb+') as f:
+    pickle.dump(items, f)

9_build_bad_words.py

@@ -0,0 +1,4 @@
+from tester import Tester
+
+tester = Tester()
+tester.build_bad()

README.md

@@ -0,0 +1,15 @@
+# Порядок запуска
+Требуется python 3.6.8 (на более ранних версиях 3.6 падает pickle при сериализации больших файлов)
+- load_dictionary - загрузка словаря opencorpora во внутренний формат. Извлекает словарные слова и слова для обучения
+- vectorize - векторизует слова для обучения
+- cls_dataset - генерирует датасеты для задач классификации, расставляет во всех векторизованных словах главный класс
+- lemma_dataset - генерирует датасет для лемматизации
+- inflect_dataset - генерирует датасет для постановки слов в форму
+- train - тренирует модель
+- build_numbers - формирует данные по числительным
+- generate_tags - генерирует объединенную нумерацию для тегов
+- build_bad_words - собирает слова, в которых сеть делает ошибки
+- release_dict - опубликовывает новый словарь
+- release_tests - публикует интеграционные тесты
+- release_dict - публикует релизный вариант словаря
+- release_dict - публикует модель

best_checkpoint/-44.data-00000-of-00001

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0b29dd603d51ab98ce3bcb1132a7a124189eb8c21b7cdc4854b884edae081c5
+size 9443076

best_checkpoint/-44.meta

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d84c8ba242dc09eb81ad1c26f9470895f26da1dbc0b80117d7d35efed5f0f55
+size 36330278

best_checkpoint/checkpoint

@@ -0,0 +1,46 @@
+model_checkpoint_path: "-44"
+all_model_checkpoint_paths: "-0"
+all_model_checkpoint_paths: "-1"
+all_model_checkpoint_paths: "-2"
+all_model_checkpoint_paths: "-3"
+all_model_checkpoint_paths: "-4"
+all_model_checkpoint_paths: "-5"
+all_model_checkpoint_paths: "-6"
+all_model_checkpoint_paths: "-7"
+all_model_checkpoint_paths: "-8"
+all_model_checkpoint_paths: "-9"
+all_model_checkpoint_paths: "-10"
+all_model_checkpoint_paths: "-11"
+all_model_checkpoint_paths: "-12"
+all_model_checkpoint_paths: "-13"
+all_model_checkpoint_paths: "-14"
+all_model_checkpoint_paths: "-15"
+all_model_checkpoint_paths: "-16"
+all_model_checkpoint_paths: "-17"
+all_model_checkpoint_paths: "-18"
+all_model_checkpoint_paths: "-19"
+all_model_checkpoint_paths: "-20"
+all_model_checkpoint_paths: "-21"
+all_model_checkpoint_paths: "-22"
+all_model_checkpoint_paths: "-23"
+all_model_checkpoint_paths: "-24"
+all_model_checkpoint_paths: "-25"
+all_model_checkpoint_paths: "-26"
+all_model_checkpoint_paths: "-27"
+all_model_checkpoint_paths: "-28"
+all_model_checkpoint_paths: "-29"
+all_model_checkpoint_paths: "-30"
+all_model_checkpoint_paths: "-31"
+all_model_checkpoint_paths: "-32"
+all_model_checkpoint_paths: "-33"
+all_model_checkpoint_paths: "-34"
+all_model_checkpoint_paths: "-35"
+all_model_checkpoint_paths: "-36"
+all_model_checkpoint_paths: "-37"
+all_model_checkpoint_paths: "-38"
+all_model_checkpoint_paths: "-39"
+all_model_checkpoint_paths: "-40"
+all_model_checkpoint_paths: "-41"
+all_model_checkpoint_paths: "-42"
+all_model_checkpoint_paths: "-43"
+all_model_checkpoint_paths: "-44"

config.yml

@@ -0,0 +1,681 @@
+random_seed: 1917
+validation_persent: 2
+test_persent: 1
+min_word_size: 2
+max_word_size: 32
+model_key: small #big
+prefix_filter_length: 1
+
+
+dict_path: /media/alex/hybrid/Projects/Resources/dict.opcorpora.xml
+dataset_words_path: data/dataset_words.pkl
+dict_words_path: data/dict_words.pkl
+vect_words_path: data/vec_words.pkl
+cls_classes_path: data/classification_classes.pkl
+numb_classes_path: data/numb_classes.pkl
+numb_data_path: data/numb.pkl
+tags_path: data/tags.pkl
+dics_path: data/
+dataset_path: data/dataset
+inflect_templates_path: data/inflect_templates.pkl
+export_path: export
+save_path: checkpoints/
+bad_path: data/
+
+test_results_paths:
+  - latest_release
+
+publish_net_paths:
+  - ../../cs/DeepMorphy/NeuralNet/
+  - latest_release
+
+publish_gramm_paths:
+  - ../../cs/DeepMorphy/
+  - latest_release
+
+publish_tags_paths:
+  - ../../cs/DeepMorphy/
+  - latest_release
+
+publish_numbers_paths:
+  - ../../cs/DeepMorphy/Numb
+  - latest_release
+
+publish_dictionary_paths:
+  - ../../cs/DeepMorphy/WordDict
+  - latest_release
+
+publish_dataset_info_paths:
+  - latest_release
+
+public_inflect_templates_paths:
+  - latest_release
+  - ../../cs/DeepMorphy/
+
+publish_tests_path: ../../cs/IntegrationTester/
+publish_gram_doc_path: ../../../gram.md
+publish_inflect_doc_path: ../../../inflect.md
+
+train_devices:
+#  - /cpu:0
+  - /gpu:0
+  - /gpu:1
+  - /gpu:2
+filler: "############################################################################"
+start_token: 36
+end_token: 35
+main_class_k: 4
+
+train_steps:
+  #- mood
+  #- voic
+  #- post
+  #- gndr
+  #- nmbr
+  #- case
+  #- tens
+  #- pers
+  #- main
+  #- lemm
+  - inflect
+
+
+ignore_restore: []
+  #- lemm
+  #- inflect
+
+graph_part_configs:
+  default:
+    #learn params
+    stop_main_metric_delta: 0.1
+    learn_rate: 0.00001
+    learn_rate_decay_step: 0.1
+    min_learn_rate: 0.00000001
+    return_step: 2
+    keep_drop: 1
+    use_weights: true
+    clip_grads: true
+    test_batch_size:  2048
+    train_batch_size: 4096
+
+    # common params
+    char_vector_size: 32
+    rnn_state_size: 32
+    rnn_layers_count: 2
+    rnn_bidirectional: True
+    use_residual: false
+    main_metric_type: Loss
+
+    #seq2seq params
+    max_length: 33
+    encoder:
+      gram_vector_size: 7
+      ad_cls_vector_size: 8
+      char_vector_size: 64
+      rnn_state_size: 64
+      rnn_layers_count: 2
+      rnn_bidirectional: true
+      use_residual: false
+    decoder:
+      gram_vector_size: 14
+      ad_cls_vector_size: 16
+      rnn_state_size: 128
+      rnn_layers_count: 1
+      keep_drop: 1.0
+      use_residual: false
+
+  pers:
+    test_batch_size:  256
+    train_batch_size: 1024
+    char_vector_size: 8
+    rnn_state_size: 8
+
+  mood:
+    char_vector_size: 16
+    rnn_state_size: 16
+
+  main:
+    char_vector_size: 64
+    rnn_state_size: 64
+    test_batch_size: 4096
+    train_batch_size: 4096
+
+  lemm:
+    char_vector_size: 64
+    batch_size: 8192
+    main_metric_type: Accuracy
+    stop_main_metric_delta: 0.00001
+
+  inflect:
+    transfer_init: false
+    char_vector_size: 64
+    batch_size: 8192
+    main_metric_type: Accuracy
+    stop_main_metric_delta: 0.000001
+
+chars:
+  - UNDEFINED # неизвестный символ
+  - а
+  - б
+  - в
+  - г
+  - д
+  - е
+  - ё
+  - ж
+  - з
+  - и
+  - й
+  - к
+  - л
+  - м
+  - н
+  - о
+  - п
+  - р
+  - с
+  - т
+  - у
+  - ф
+  - х
+  - ч
+  - ц
+  - ш
+  - щ
+  - ъ
+  - ы
+  - ь
+  - э
+  - ю
+  - я
+  - "-"
+
+lemma_same_word:
+  - post: noun
+    nmbr: sing
+    case: nomn
+    gndr: masc
+
+  - post: noun
+    nmbr: sing
+    case: nomn
+    gndr: neut
+
+  - post: noun
+    nmbr: sing
+    case: nomn
+    gndr: femn
+
+  - post: noun
+    nmbr: sing
+    case: nomn
+    gndr: msf
+
+  - post: adjf
+    nmbr: sing
+    gndr: masc
+    case: nomn
+
+  - post: infn
+
+  - post: advb
+
+  - post: prep
+
+  - post: conj
+
+  - post: prcl
+
+  - post: intj
+
+  - post: prep
+
+  - post: pred
+    tens: pres
+
+
+
+other_post_types:
+  punct:
+    key_ru: пункт
+    name_ru: пунктуция
+    desc_ru: пунктуация
+
+  int:
+    key_ru: цифра
+    name_ru: цифра
+    desc_ru: цифра
+
+  romn:
+    key_ru: рим_цифр
+    name_ru: римская цифра
+    desc_ru: римская цифра
+
+  unkn:
+    key_ru: неизв
+    name_ru: неизвестный токен
+    desc_ru: неизвестный токен
+
+
+dict_post_types:
+  npro:
+    key_ru: мест
+    power: 2
+    name_ru: местоимение
+    desc_ru: местоимение
+    keys:
+      - NPRO
+
+  prep:
+    key_ru: предл
+    power: 2
+    name_ru: предлог
+    desc_ru: предлог
+    keys:
+      - PREP
+
+  conj:
+    key_ru: союз
+    power: 2
+    name_ru: союз
+    desc_ru: союз
+    keys:
+      - CONJ
+
+  prcl:
+    key_ru: част
+    power: 0
+    name_ru: частица
+    desc_ru: частица
+    keys:
+      - PRCL
+
+  intj:
+    key_ru: межд
+    power: 0.5
+    name_ru: междометие
+    desc_ru: междометие
+    keys:
+      - INTJ
+
+  numb:
+    key_ru: числ
+    power: 2
+    name_ru: числительное
+    desc_ru: числительное
+    keys:
+      - NUMR
+
+  pred:
+    key_ru: предик
+    power: 0
+    name_ru: предикатив
+    desc_ru: предикатив
+    keys:
+      - PRED
+
+grammemes_types:
+  post:
+    index: 0
+    key_ru: чр
+    name: часть речи
+    classes:
+      noun:
+        index: 0
+        key_ru: сущ
+        name_ru: существительное
+        desc_ru: существительное
+        keys:
+          - NOUN
+
+      adjf:
+        index: 1
+        key_ru: прил
+        name_ru: полное прилагательное
+        desc_ru: полное прилагательное
+        keys:
+          - ADJF
+
+      adjs:
+        index: 2
+        key_ru: кр_прил
+        name_ru: краткое прилагательное
+        desc_ru: полное прилагательное
+        keys:
+          - ADJS
+
+      verb:
+        index: 3
+        key_ru: гл
+        name_ru: личная форма глагола
+        desc_ru: личная форма глагола
+        keys:
+          - VERB
+
+      infn:
+        index: 4
+        key_ru: инф_гл
+        name_ru: инфинитив глагола
+        desc_ru: инфинитив глагола
+        keys:
+          - INFN
+
+      comp:
+        index: 5
+        key_ru: комп
+        name_ru: компаратив
+        desc_ru: компаратив
+        keys:
+          - COMP
+
+      prtf:
+        index: 6
+        key_ru: прич
+        name_ru: причастие
+        desc_ru: причастие
+        keys:
+          - PRTF
+
+      prts:
+        index: 7
+        key_ru: кр_прич
+        name_ru: краткое причастие
+        desc_ru: краткое причастие
+        keys:
+          - PRTS
+
+      grnd:
+        index: 8
+        key_ru: деепр
+        name_ru: деепричастие
+        desc_ru: деепричастие
+        keys:
+          - GRND
+
+      advb:
+        index: 9
+        key_ru: нареч
+        name_ru: наречие
+        desc_ru: наречие
+        keys:
+          - ADVB
+
+  gndr:
+    index: 1
+    key_ru: род
+    name: род
+    classes:
+      masc:
+        index: 0
+        key_ru: муж
+        name_ru: мужской
+        desc_ru: мужской род
+        keys:
+          - masc
+
+      femn:
+        index: 1
+        key_ru: жен
+        name_ru: женский
+        desc_ru: женский род
+        keys:
+          - femn
+
+      neut:
+        index: 2
+        key_ru: ср
+        name_ru: средний
+        desc_ru: средний род
+        keys:
+          - neut
+
+      msf:
+        index: 3
+        key_ru: общ
+        name_ru: общий
+        desc_ru: общий род
+        keys:
+          - ms-f
+
+  nmbr:
+    index: 2
+    key_ru: число
+    name: число
+    classes:
+      sing:
+        index: 0
+        key_ru: ед
+        name_ru: единственное
+        desc_ru: единственное число
+        keys:
+          - sing
+
+      plur:
+        index: 1
+        key_ru: мн
+        name_ru: множественное
+        desc_ru: множественное число
+        keys:
+          - plur
+
+  case:
+   index: 3
+   key_ru: падеж
+   name: падеж
+   classes:
+     nomn:
+       index: 0
+       key_ru: им
+       name_ru: именительный
+       desc_ru: именительный падеж
+       keys:
+         - nomn
+
+     gent:
+       index: 1
+       key_ru: рд
+       name_ru: родительный
+       desc_ru: родительный падеж
+       keys:
+         - gent
+
+     datv:
+       index: 2
+       key_ru: дт
+       name_ru: дательный
+       desc_ru: дательный падеж
+       keys:
+         - datv
+
+     accs:
+       index: 3
+       key_ru: вн
+       name_ru: винительный
+       desc_ru: винительный падеж
+       keys:
+         - accs
+
+     ablt:
+       index: 4
+       key_ru: тв
+       name_ru: творительный
+       desc_ru: творительный падеж
+       keys:
+         - ablt
+
+     loct:
+       index: 5
+       key_ru: пр
+       name_ru: предложный
+       desc_ru: предложный падеж
+       keys:
+         - loct
+
+     gent1:
+       index: 6
+       key_ru: рд1
+       name_ru: первый родительный
+       desc_ru: первый родительный падеж
+       keys:
+         - gen1
+
+     gent2:
+       index: 7
+       key_ru: рд2
+       name_ru: второй родительный
+       desc_ru: второй родительный падеж
+       keys:
+         - gen2
+
+     voct:
+       index: 8
+       key_ru: зв
+       name_ru: звательный
+       desc_ru: звательный падеж
+       keys:
+         - voct
+
+     accs2:
+       index: 9
+       key_ru: вн2
+       name_ru: второй винительный
+       desc_ru: второй винительный падеж
+       keys:
+         - acc2
+
+     loct1:
+       index: 10
+       key_ru: пр1
+       name_ru: первый предложный
+       desc_ru: первый предложный падеж
+       keys:
+         - loc1
+
+     loct2:
+       index: 11
+       key_ru: пр2
+       name_ru: второй предложный
+       desc_ru: второй предложный падеж
+       keys:
+         - loc2
+
+  pers:
+    index: 4
+    key_ru: лицо
+    name: лицо
+    classes:
+      1per:
+        index: 0
+        key_ru: 1л
+        name_ru: 1 лицо
+        desc_ru: 1 лицо
+        keys:
+          - 1per
+
+      2per:
+        index: 1
+        key_ru: 2л
+        name_ru: 2 лицо
+        desc_ru: 2 лицо
+        keys:
+          - 2per
+
+      3per:
+        index: 2
+        key_ru: 3л
+        name_ru: 3 лицо
+        desc_ru: 3 лицо
+        keys:
+          - 3per
+
+  tens:
+    index: 5
+    key_ru: время
+    name: время
+    classes:
+      pres:
+        index: 0
+        key_ru: наст
+        name_ru: настоящее
+        desc_ru: настоящее время
+        keys:
+          - pres
+
+      past:
+        index: 1
+        key_ru: прош
+        name_ru: прошедшее
+        desc_ru: прошедшее время
+        keys:
+          - past
+
+      futr:
+        index: 2
+        key_ru: буд
+        name_ru: будущее
+        desc_ru: будущее время
+        keys:
+          - futr
+
+  mood:
+    index: 6
+    key_ru: накл
+    name: наклонение
+    classes:
+      indc:
+        index: 0
+        key_ru: изъяв
+        name_ru: изъявительное
+        desc_ru: изъявительное наклонение
+        keys:
+          - indc
+
+      impr:
+        index: 1
+        key_ru: повел
+        name_ru: повелительное
+        desc_ru: повелительное наклонение
+        keys:
+          - impr
+
+  voic:
+    index: 7
+    key_ru: залог
+    name: залог
+    classes:
+      actv:
+        index: 0
+        key_ru: дейст
+        name_ru: действительный
+        desc_ru: действительный залог
+        keys:
+          - actv
+
+      pssv:
+        index: 1
+        key_ru: страд
+        name_ru: страдательный
+        desc_ru: страдательный залог
+        keys:
+          - pssv
+
+ad_tags:
+  - incl
+  - excl
+  - abbr
+  - surn
+  - patr
+  - orgn
+  - trad
+  - cmp2
+  - geox
+
+inflect_ignore_tags:
+  - incl
+  - cmp2
+
+dict_ignore_tags:
+  - patr
+  - orgn
+  - surn
+

graph/base.py

@@ -0,0 +1,248 @@
+import os
+import tensorflow as tf
+import tf_utils as tfu
+from tqdm import tqdm
+from abc import ABC, abstractmethod
+from utils import RANDOM
+
+
+class TfContext:
+    def __init__(self,
+                 sess,
+                 saver,
+                 learn_rate_op):
+        self.sess = sess
+        self.saver = saver
+        self.learn_rate_op = learn_rate_op
+        self.epoch = 0
+
+
+class GraphPartBase(ABC):
+    def __init__(self, for_usage, global_settings, current_settings, optimiser, reset_optimiser, key, metric_names):
+        self.key = key
+        self.global_settings = global_settings
+        self.filler = global_settings['filler']
+        self.main_metric_name = current_settings['main_metric_type']
+        self.settings = current_settings
+        self.for_usage = for_usage
+        self.optimiser = optimiser
+        self.reset_optimiser = reset_optimiser
+        self.metric_names = metric_names
+        self.max_word_size = global_settings['max_word_size']
+        self.checkpoints_keep = 10000
+        self.chars_count = len(global_settings['chars']) + 1
+        self.dataset_path = global_settings['dataset_path']
+        self.grammemes_count = len(global_settings['grammemes_types'])
+        self.main_classes = global_settings['main_classes']
+        self.main_classes_count = len(self.main_classes)
+        self.metrics_reset = []
+        self.metrics_update = []
+        self.devices_metrics = {metr: [] for metr in self.metric_names}
+        self.main_scope_name = key.title()
+        self.save_path = global_settings['save_path']
+        self.dev_grads = []
+        self.losses = []
+        self.devices = global_settings['train_devices']
+        self.devices_count = len(self.devices)
+        self.dataset_path = global_settings['dataset_path']
+        self.xs = []
+        self.x_seq_lens = []
+        self.prints = []
+        self.main_cls_dic = self.global_settings['main_classes']
+        self.learn_rate_val = self.settings['learn_rate']
+        self.best_model_metric = None
+        self.best_epoch = None
+        self.init_checkpoint = None
+
+    def train(self, tc):
+        return_step = 0
+        trains = self.__load_dataset__('train')
+        valids = self.__load_dataset__('valid')
+        self.best_model_metric = self.__valid_loop__(tc, valids)
+        self.best_epoch = -1
+        while True:
+            tqdm.write(self.filler)
+            tqdm.write(self.filler)
+            tqdm.write(self.main_scope_name)
+
+            train_main_metric = self.__train_loop__(tc, trains)
+            valid_main_metric = self.__valid_loop__(tc, valids)
+
+            tqdm.write(f"Epoch {tc.epoch} Train {self.main_metric_name}: {train_main_metric} Validation {self.main_metric_name}: {valid_main_metric}")
+            need_decay = False
+            delta = self.__calc_metric_delta__(self.best_model_metric, valid_main_metric)
+
+            if delta > 0:
+                if delta < self.settings['stop_main_metric_delta']:
+                    tqdm.write(f"{self.main_metric_name} delta is less then min value")
+                    need_decay = True
+                else:
+                    return_step = 0
+                self.best_model_metric = valid_main_metric
+                self.best_epoch = tc.epoch
+                tc.saver.save(tc.sess, self.save_path, tc.epoch)
+                tc.epoch += 1
+            else:
+                tqdm.write("Best epoch is better then current")
+                tc.sess.run(self.reset_optimiser)
+                need_decay = True
+                self.__restore_best_epoch__(tc)
+
+            if not need_decay:
+                continue
+
+            if return_step == self.settings['return_step']:
+                self.__decay_params__()
+                if self.learn_rate_val < self.settings['min_learn_rate']:
+                    tqdm.write(f"Learning rate {self.learn_rate_val} is less then min learning rate")
+                    finish = self.__before_finish__()
+                    if finish:
+                        break
+                return_step = 0
+            else:
+                RANDOM.shuffle(trains)
+                tqdm.write(f"Return step increased")
+                return_step += 1
+
+        return self.best_epoch, self.best_model_metric
+
+    def __train_loop__(self, tc, trains):
+        tc.sess.run(self.metrics_reset)
+        for item in tqdm(trains, desc=f"Train, epoch {tc.epoch}"):
+            launch = [self.optimize]
+            launch.extend(self.metrics_update)
+            if len(self.prints):
+                launch.extend(self.prints)
+
+            feed_dic = self.__create_feed_dict__('train', item)
+            feed_dic[tc.learn_rate_op] = self.learn_rate_val
+            tc.sess.run(launch, feed_dic)
+
+        train_main_metric = self.__write_metrics_report__(tc.sess, "Train")
+        return train_main_metric
+
+    def __valid_loop__(self, tc, valids):
+        tc.sess.run(self.metrics_reset)
+        for item in tqdm(valids, desc=f"Validation, epoch {tc.epoch}"):
+            launch = []
+            launch.extend(self.metrics_update)
+            if len(self.prints):
+                launch.extend(self.prints)
+            feed_dic = self.__create_feed_dict__('valid', item)
+            tc.sess.run(launch, feed_dic)
+
+        valid_main_metric = self.__write_metrics_report__(tc.sess, "Valid")
+        return valid_main_metric
+
+    def __calc_metric_delta__(self, best_model_metric, cur_model_metric):
+        delta = best_model_metric - cur_model_metric
+        if self.main_metric_name != "Loss":
+            delta = -delta
+        return delta
+
+    def __before_finish__(self):
+        return True
+
+    def __decay_params__(self):
+        self.learn_rate_val = self.learn_rate_val * self.settings['learn_rate_decay_step']
+        tqdm.write(f"Learning rate decayed. New value: {self.learn_rate_val}")
+
+    def __restore_best_epoch__(self, tc):
+        if self.best_epoch == -1 and tc.epoch == 0:
+            tqdm.write(f"Restoring from init_checkpoint {self.best_epoch}")
+            self.restore(tc.sess, self.init_checkpoint)
+        elif self.best_epoch == -1:
+            tqdm.write(f"Restoring best epoch {tc.epoch}")
+            self.restore(tc.sess, os.path.join(self.save_path, f"-{tc.epoch}"))
+        else:
+            tqdm.write(f"Restoring best epoch {self.best_epoch}")
+            self.restore(tc.sess, os.path.join(self.save_path, f"-{self.best_epoch}"))
+
+    def build_graph_end(self):
+        with tf.variable_scope(self.main_scope_name, reuse=tf.AUTO_REUSE) as scope:
+            self.metrics = {
+                metr: tf.reduce_mean(self.devices_metrics[metr], name=metr)
+                for metr in self.devices_metrics
+            }
+            if not self.for_usage:
+                self.grads = tfu.average_gradients(self.dev_grads)
+                if self.settings['clip_grads']:
+                    self.grads = [(tf.clip_by_value(grad, -1., 1.), var) for grad, var in self.grads]
+
+                self.optimize = self.optimiser.apply_gradients(self.grads, name='Optimize')
+                self.loss = tf.reduce_sum(self.losses, name='GlobalLoss')
+
+    def build_graph_for_device(self, *args):
+        with tf.variable_scope(self.main_scope_name, reuse=tf.AUTO_REUSE) as scope:
+            self.__build_graph_for_device__(*args)
+
+    def restore(self, sess, check_point):
+        try:
+            vars = [
+                var
+                for var in tf.global_variables(f"{self.main_scope_name}/")
+                if "Adam" not in var.name
+            ]
+            saver = tf.train.Saver(var_list=vars)
+            saver.restore(sess, check_point)
+            self.init_checkpoint = check_point
+            tqdm.write(f"Restoration for graph part '{self.key}', scope {self.main_scope_name} success")
+        except Exception as ex:
+            tqdm.write(f"Restoration for graph part '{self.key}', scope {self.main_scope_name} failed. Error: {ex}")
+
+    def __write_metrics_report__(self, sess, step_name):
+        tqdm.write('')
+        launch_results = sess.run(self.metrics)
+        result = [f"{step_name} metrics: "]
+
+        for index, metr in enumerate(self.metrics):
+            result.append('{:>8}'.format(self.metric_names[index]))
+            result.append("=")
+            result.append("{0:.7f}".format(launch_results[metr]))
+            result.append(" ")
+
+        result = "".join(result)
+        tqdm.write(result)
+        return launch_results[self.main_metric_name]
+
+    def create_mean_metric(self, metric_index, values):
+        metr_epoch_loss, metr_update, metr_reset = tfu.create_reset_metric(
+            tf.metrics.mean,
+            self.metric_names[metric_index],
+            values
+        )
+        self.metrics_reset.append(metr_reset)
+        self.metrics_update.append(metr_update)
+        self.devices_metrics[self.metric_names[metric_index]].append(metr_epoch_loss)
+
+    def create_accuracy_metric(self, metric_index, labels, predictions):
+        metr_epoch_loss, metr_update, metr_reset = tfu.create_reset_metric(
+            tf.metrics.accuracy,
+            self.metric_names[metric_index],
+            labels=labels,
+            predictions=predictions
+        )
+        self.metrics_reset.append(metr_reset)
+        self.metrics_update.append(metr_update)
+        self.devices_metrics[self.metric_names[metric_index]].append(metr_epoch_loss)
+
+    def __create_feed_dict__(self, op_name, item):
+        feed_dic = {}
+        for dev_num, batch in enumerate(item):
+            feed_dic[self.xs[dev_num]] = batch['x']
+            feed_dic[self.x_seq_lens[dev_num]] = batch['x_seq_len']
+            self.__update_feed_dict__(op_name, feed_dic, batch, dev_num)
+
+        return feed_dic
+
+    @abstractmethod
+    def __update_feed_dict__(self, op_name, feed_dict, batch, dev_num):
+        pass
+
+    @abstractmethod
+    def __build_graph_for_device__(self, *args):
+        pass
+
+    @abstractmethod
+    def __load_dataset__(self, operation_name):
+        return []

graph/gram_cls.py

@@ -0,0 +1,112 @@
+import tensorflow as tf
+import tf_utils as tfu
+from graph.base import GraphPartBase
+
+
+class GramCls(GraphPartBase):
+    def __init__(self, key, for_usage, global_settings, current_settings, optimiser, reset_optimiser):
+        super().__init__(for_usage,
+                         global_settings,
+                         current_settings,
+                         optimiser,
+                         reset_optimiser,
+                         key,
+                         ['Loss', 'Accuracy'])
+        self.gram = key
+        self.grammemes = global_settings['grammemes_types']
+        self.classes = self.grammemes[key]['classes']
+        self.classes_count = len(self.classes)
+        self.checks = []
+        self.weights = []
+        self.keep_drops = []
+        self.dev_grads = []
+        self.probs = []
+        self.results = []
+        self.ys = []
+        self.losses = []
+
+    def __build_graph_for_device__(self, x, seq_len):
+        self.xs.append(x)
+        self.x_seq_lens.append(seq_len)
+
+        if self.for_usage:
+            keep_drop = tf.constant(1, dtype=tf.float32, name='KeepDrop')
+        else:
+            keep_drop = tf.placeholder(dtype=tf.float32, name='KeepDrop')
+
+        self.keep_drops.append(keep_drop)
+
+        y = tf.placeholder(dtype=tf.int32,
+                                    shape=(None, self.classes_count),
+                                    name='Y')
+        weights = tf.placeholder(dtype=tf.float32, shape=(None,), name='Weight')
+
+        x_emd_init = tf.random_normal((self.chars_count, self.settings['char_vector_size']))
+        x_emb = tf.get_variable("Embeddings", initializer=x_emd_init)
+
+        rnn_input = tf.nn.embedding_lookup(x_emb, x)
+        rnn_logits = tfu.build_rnn(rnn_input,
+                                   keep_drop,
+                                   seq_len,
+                                   self.settings,
+                                   for_usage=self.for_usage)
+
+        if not self.for_usage:
+            self.checks.append(tf.check_numerics(rnn_logits, "RnnLogitsNullCheck"))
+
+        logits = tfu.rnn_top('RnnTop',
+                                      rnn_logits,
+                                      self.settings,
+                                      self.classes_count)
+
+        if not self.for_usage:
+            self.checks.append(tf.check_numerics(logits, "LogitsNullCheck"))
+
+        float_y = tf.cast(y, tf.float32)
+        errors = tf.nn.sigmoid_cross_entropy_with_logits(logits=logits,
+                                                         labels=float_y)
+        errors = tf.reduce_sum(errors, axis=1)
+        errors = errors * weights
+
+        if not self.for_usage:
+            self.checks.append(tf.check_numerics(errors, "ErrorNullCheck"))
+
+        probs = tf.nn.softmax(logits)
+        result = tf.argmax(probs, axis=1, name="Results")
+        loss = tf.reduce_sum(errors)
+
+        if not self.for_usage:
+            self.checks.append(tf.check_numerics(errors, "LossNullCheck"))
+
+        grads = self.optimiser.compute_gradients(loss)
+        self.losses.append(loss)
+        self.probs.append(probs)
+        self.results.append(result)
+        self.ys.append(y)
+        self.weights.append(weights)
+        self.dev_grads.append(grads)
+
+        # metrics
+        self.create_mean_metric(0, loss)
+        labels = tf.math.argmax(y, axis=1)
+        predictions = tf.math.argmax(probs, axis=1)
+        self.create_accuracy_metric(1, labels, predictions)
+
+    def __update_feed_dict__(self, op_name, feed_dict, batch, dev_num):
+        feed_dict[self.keep_drops[dev_num]] = 1 if op_name == 'test' else self.settings['keep_drop']
+        feed_dict[self.ys[dev_num]] = batch['y']
+        feed_dict[self.weights[dev_num]] = batch['weight']
+
+    def __load_dataset__(self, operation_name):
+        return list(
+            tfu.load_cls_dataset(
+                self.dataset_path,
+                self.devices_count,
+                operation_name,
+                self.settings['train_batch_size'],
+                self.settings['use_weights'],
+                self.gram
+            )
+        )
+
+

graph/inflect.py

@@ -0,0 +1,85 @@
+import tensorflow as tf
+import tf_utils as tfu
+from graph.base import GraphPartBase
+
+
+class Inflect(GraphPartBase):
+    def __init__(self, for_usage, global_settings, current_settings, optimiser, reset_optimiser):
+        super().__init__(for_usage, global_settings, current_settings, optimiser, reset_optimiser, 'inflect', ["Loss", "AccuracyByChar", "Accuracy"])
+        self.chars_count = self.chars_count + 1
+        self.start_char_index = global_settings['start_token']
+        self.end_char_index = global_settings['end_token']
+        self.results = []
+        self.x_cls = []
+        self.ys = []
+        self.y_seq_lens = []
+        self.y_cls = []
+        self.keep_drops = []
+        self.decoder_keep_drops = []
+
+    def __build_graph_for_device__(self, x, x_seq_len, batch_size, cls=None):
+        self.xs.append(x)
+        self.x_seq_lens.append(x_seq_len)
+
+        x_cls = tf.placeholder(dtype=tf.int32, shape=(None,), name='XClass')
+        self.x_cls.append(x_cls)
+
+        if batch_size is None:
+            batch_size = self.settings['batch_size']
+
+        y = tf.placeholder(dtype=tf.int32, shape=(None, None), name='Y')
+        self.ys.append(y)
+
+        y_seq_len = tf.placeholder(dtype=tf.int32, shape=(None,), name='YSeqLen')
+        self.y_seq_lens.append(y_seq_len)
+
+        y_cls = tf.placeholder(dtype=tf.int32, shape=(None,), name='YClass')
+        self.y_cls.append(y_cls)
+
+        tfu.seq2seq(self,
+                    batch_size,
+                    x,
+                    y_cls,
+                    x_seq_len,
+                    y,
+                    x_cls,
+                    y_seq_len)
+
+    def __update_feed_dict__(self, op_name, feed_dict, batch, dev_num):
+        feed_dict[self.x_cls[dev_num]] = batch['x_cls']
+        feed_dict[self.y_cls[dev_num]] = batch['y_cls']
+        feed_dict[self.ys[dev_num]] = batch['y']
+        feed_dict[self.y_seq_lens[dev_num]] = batch['y_seq_len']
+        feed_dict[self.keep_drops[dev_num]] = self.settings['keep_drop']
+        feed_dict[self.decoder_keep_drops[dev_num]] = self.settings['decoder']['keep_drop']
+
+    def __load_dataset__(self, operation_name):
+        items = list(tfu.load_inflect_dataset(
+            self.dataset_path,
+            self.devices_count,
+            operation_name,
+            self.settings['batch_size']
+        ))
+        return items
+
+    def transfer_learning_init(self, sess):
+        my_prefix = f"{self.main_scope_name}/"
+        vars = {
+            var.name[len(my_prefix):]: var
+            for var in tf.global_variables(my_prefix)
+            if "Adam" not in var.name
+        }
+
+        lem_prefix = f"Lemm/"
+        lem_vars = {
+            var.name[len(lem_prefix):]: var
+            for var in tf.global_variables(lem_prefix)
+            if "Adam" not in var.name
+        }
+
+        for key in vars:
+            my_var = vars[key]
+            lem_var = lem_vars[key]
+            value = sess.run(lem_var)
+            sess.run(my_var.assign(value))
+        print()

graph/lemm.py

@@ -0,0 +1,61 @@
+
+import tensorflow as tf
+import tf_utils as tfu
+from tqdm import tqdm
+from graph.base import GraphPartBase
+
+
+class Lemm(GraphPartBase):
+    def __init__(self, for_usage, global_settings, current_settings, optimiser, reset_optimiser):
+        super().__init__(for_usage, global_settings, current_settings, optimiser, reset_optimiser, 'lemm', ["Loss", "AccuracyByChar", "Accuracy"])
+        self.chars_count = self.chars_count + 1
+        self.start_char_index = global_settings['start_token']
+        self.end_char_index = global_settings['end_token']
+        self.results = []
+        self.ys = []
+        self.y_seq_lens = []
+        self.cls = []
+        self.keep_drops = []
+        self.decoder_keep_drops = []
+
+    def __build_graph_for_device__(self, x, x_seq_len, batch_size, x_cls=None):
+        self.xs.append(x)
+        self.x_seq_lens.append(x_seq_len)
+
+        if x_cls is None:
+            x_cls = tf.placeholder(dtype=tf.int32, shape=(None,), name='XClass')
+            self.cls.append(x_cls)
+
+        if batch_size is None:
+            batch_size = self.settings['batch_size']
+
+        y = tf.placeholder(dtype=tf.int32, shape=(None, None), name='Y')
+        self.ys.append(y)
+
+        y_seq_len = tf.placeholder(dtype=tf.int32, shape=(None,), name='YSeqLen')
+        self.y_seq_lens.append(y_seq_len)
+
+        tfu.seq2seq(self,
+                    batch_size,
+                    x,
+                    x_cls,
+                    x_seq_len,
+                    y,
+                    x_cls,
+                    y_seq_len)
+
+    def __update_feed_dict__(self, op_name, feed_dict, batch, dev_num):
+        feed_dict[self.cls[dev_num]] = batch['x_cls']
+        feed_dict[self.ys[dev_num]] = batch['y']
+        feed_dict[self.y_seq_lens[dev_num]] = batch['y_seq_len']
+        feed_dict[self.keep_drops[dev_num]] = self.settings['keep_drop']
+        feed_dict[self.decoder_keep_drops[dev_num]] = self.settings['decoder']['keep_drop']
+
+    def __load_dataset__(self, operation_name):
+        items = list(tfu.load_lemma_dataset(
+            self.dataset_path,
+            self.devices_count,
+            operation_name,
+            self.settings['batch_size']
+        ))
+        return items

graph/main_cls.py

@@ -0,0 +1,103 @@
+import tensorflow as tf
+import tf_utils as tfu
+from graph.base import GraphPartBase
+
+
+class MainCls(GraphPartBase):
+    def __init__(self, for_usage, global_settings, current_settings, optimiser, reset_optimiser):
+        super().__init__(for_usage, global_settings, current_settings, optimiser, reset_optimiser, 'main', ['Loss', 'Accuracy'])
+        self.checks = []
+        self.weights = []
+        self.keep_drops = []
+        self.losses = []
+        self.probs = []
+        self.results = []
+        self.ys = []
+        self.drops = []
+        self.top_k = global_settings['main_class_k']
+
+    def __build_graph_for_device__(self, x, seq_len, gram_probs, gram_drop):
+        self.xs.append(x)
+        self.x_seq_lens.append(seq_len)
+        self.drops.append(gram_drop)
+
+        y = tf.placeholder(dtype=tf.int32, shape=(None, self.main_classes_count), name='Y')
+        weights = tf.placeholder(dtype=tf.float32, shape=(None,), name='Weight')
+        x_emd_init = tf.random_normal((self.chars_count, self.settings['char_vector_size']))
+        x_emb = tf.get_variable("Embeddings", initializer=x_emd_init)
+        rnn_input = tf.nn.embedding_lookup(x_emb, x)
+
+        if self.for_usage:
+            cls_keep_drop = tf.constant(1, dtype=tf.float32, name='KeepDrop')
+        else:
+            cls_keep_drop = tf.placeholder(dtype=tf.float32, name='KeepDrop')
+        self.keep_drops.append(cls_keep_drop)
+
+        init_state = tf.concat(gram_probs, 1)
+        with tf.variable_scope("InitRnnState", reuse=tf.AUTO_REUSE) as scope:
+            rez_size = self.settings['rnn_state_size']
+            w_softmax = tf.get_variable("W", (init_state.shape[1], rez_size))
+            b_softmax = tf.get_variable("b", [rez_size])
+            init_state = tf.matmul(init_state, w_softmax) + b_softmax
+
+        rnn_logits = tfu.build_rnn(rnn_input,
+                                   cls_keep_drop,
+                                   seq_len,
+                                   self.settings,
+                                   init_state,
+                                   init_state,
+                                   self.for_usage)
+
+        logits = tfu.rnn_top('RnnTop',
+                             rnn_logits,
+                             self.settings,
+                             self.main_classes_count)
+
+        float_y = tf.cast(y, tf.float32)
+        errors = tf.nn.sigmoid_cross_entropy_with_logits(logits=logits,
+                                                         labels=float_y)
+        errors = tf.reduce_sum(errors, axis=1)
+        errors = errors * weights
+        if not self.for_usage:
+            self.checks.append(tf.check_numerics(errors, "ErrorNullCheck"))
+
+        probs = tf.nn.softmax(logits)
+        result = tf.math.top_k(probs, self.top_k, name="Results")
+        loss = tf.reduce_sum(errors)
+
+        if not self.for_usage:
+            self.checks.append(tf.check_numerics(errors, "LossNullCheck"))
+
+        vars = tf.get_collection(tf.GraphKeys.GLOBAL_VARIABLES, scope=self.main_scope_name)
+        grads = self.optimiser.compute_gradients(loss, var_list=vars)
+
+        self.losses.append(loss)
+        self.probs.append(probs)
+        self.results.append(result)
+        self.ys.append(y)
+        self.dev_grads.append(grads)
+        self.weights.append(weights)
+
+        # metrics
+        self.create_mean_metric(0, loss)
+        labels = tf.math.argmax(y, axis=1)
+        predictions = tf.math.argmax(probs, axis=1)
+        self.create_accuracy_metric(1, labels, predictions)
+
+    def __update_feed_dict__(self, op_name, feed_dict, batch, dev_num):
+        for gram_drop in self.drops[dev_num]:
+            feed_dict[gram_drop] = 1
+        feed_dict[self.keep_drops[dev_num]] = 1 if op_name == 'test' else self.settings['keep_drop']
+        feed_dict[self.ys[dev_num]] = batch['y']
+        feed_dict[self.weights[dev_num]] = batch['weight']
+
+    def __load_dataset__(self, operation_name):
+        return list(
+            tfu.load_cls_dataset(
+                self.dataset_path,
+                self.devices_count,
+                operation_name,
+                self.settings['train_batch_size'],
+                self.settings['use_weights']
+            )
+        )

latest_release/dataset_info.txt

@@ -0,0 +1,3 @@
+dictionary
+version=0.92
+revision=412889

latest_release/dict.txt.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1decc4fb2b25659511d1700be8d7c84358ee15c638ffb7c4c0bf00ba832a9d90
+size 8530

latest_release/dict_correction.txt.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e730df5fcbcdc077eb17cf78bad394944ac01f3b36c90b5952cd7da75512c03d
+size 1015441

latest_release/dict_correction_index.txt.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c490b6acc83ef744164af6fc297110e5ed3986e0921099bc4f862a7e2ba29de0
+size 1662293

latest_release/dict_index.txt.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:851335c946f2d2606942a6a7a2b75d40d10138acf3a369925a1272aa1ebecb4f
+size 7699

latest_release/frozen_model_small.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32ee280767da16f99400f7c401b434fcd20613abc140a2fe9a88c44c4e5fa8a5
+size 4423913

latest_release/grams.xml

@@ -0,0 +1,2 @@
+<?xml version='1.0' encoding='utf-8'?>
+<Grams><G index="0" key_en="post" key_ru="чр"><C key_en="noun" key_ru="сущ" nn_index="0" /><C key_en="adjf" key_ru="прил" nn_index="1" /><C key_en="adjs" key_ru="кр_прил" nn_index="2" /><C key_en="verb" key_ru="гл" nn_index="3" /><C key_en="infn" key_ru="инф_гл" nn_index="4" /><C key_en="comp" key_ru="комп" nn_index="5" /><C key_en="prtf" key_ru="прич" nn_index="6" /><C key_en="prts" key_ru="кр_прич" nn_index="7" /><C key_en="grnd" key_ru="деепр" nn_index="8" /><C key_en="advb" key_ru="нареч" nn_index="9" /><C key_en="npro" key_ru="мест" /><C key_en="prep" key_ru="предл" /><C key_en="conj" key_ru="союз" /><C key_en="prcl" key_ru="част" /><C key_en="intj" key_ru="межд" /><C key_en="numb" key_ru="числ" /><C key_en="pred" key_ru="предик" /><C key_en="punct" key_ru="пункт" /><C key_en="int" key_ru="цифра" /><C key_en="romn" key_ru="рим_цифр" /><C key_en="unkn" key_ru="неизв" /></G><G index="1" key_en="gndr" key_ru="род"><C key_en="masc" key_ru="муж" nn_index="0" /><C key_en="femn" key_ru="жен" nn_index="1" /><C key_en="neut" key_ru="ср" nn_index="2" /><C key_en="msf" key_ru="общ" nn_index="3" /></G><G index="2" key_en="nmbr" key_ru="число"><C key_en="sing" key_ru="ед" nn_index="0" /><C key_en="plur" key_ru="мн" nn_index="1" /></G><G index="3" key_en="case" key_ru="падеж"><C key_en="nomn" key_ru="им" nn_index="0" /><C key_en="gent" key_ru="рд" nn_index="1" /><C key_en="datv" key_ru="дт" nn_index="2" /><C key_en="accs" key_ru="вн" nn_index="3" /><C key_en="ablt" key_ru="тв" nn_index="4" /><C key_en="loct" key_ru="пр" nn_index="5" /><C key_en="gent1" key_ru="рд1" nn_index="6" /><C key_en="gent2" key_ru="рд2" nn_index="7" /><C key_en="voct" key_ru="зв" nn_index="8" /><C key_en="accs2" key_ru="вн2" nn_index="9" /><C key_en="loct1" key_ru="пр1" nn_index="10" /><C key_en="loct2" key_ru="пр2" nn_index="11" /></G><G index="4" key_en="pers" key_ru="лицо"><C key_en="1per" key_ru="1л" nn_index="0" /><C key_en="2per" key_ru="2л" nn_index="1" /><C key_en="3per" key_ru="3л" nn_index="2" /></G><G index="5" key_en="tens" key_ru="время"><C key_en="pres" key_ru="наст" nn_index="0" /><C key_en="past" key_ru="прош" nn_index="1" /><C key_en="futr" key_ru="буд" nn_index="2" /></G><G index="6" key_en="mood" key_ru="накл"><C key_en="indc" key_ru="изъяв" nn_index="0" /><C key_en="impr" key_ru="повел" nn_index="1" /></G><G index="7" key_en="voic" key_ru="залог"><C key_en="actv" key_ru="дейст" nn_index="0" /><C key_en="pssv" key_ru="страд" nn_index="1" /></G></Grams>

latest_release/numbers.xml

@@ -0,0 +1,2 @@
+<?xml version='1.0' encoding='utf-8'?>
+<NumbData l="261,230,231,232,279" reg="^((?&lt;_1000000000&gt;миллиард|миллиардах|миллиардная|миллиардною|миллиардном|миллиарда|миллиардами|миллиардной|миллиардов|миллиарду|миллиардное|миллиардные|миллиардом|миллиардный|миллиардам|миллиарде|миллиардную|миллиарды|миллиардных|миллиардными|миллиардного|миллиардному|миллиардным)|(?&lt;_1000000&gt;миллиона|миллионную|миллионные|миллионного|миллионными|миллион|миллионам|миллионных|миллионами|миллионном|миллионный|миллиону|миллионе|миллионной|миллионному|миллионом|миллионным|миллионное|миллионною|миллионах|миллионная|миллионов)|(?&lt;_1000&gt;тысячные|тысяч|тысячными|тысячном|тысячью|тысячных|тысячи|тысячу|тысячная|тысячного|тысячную|тысячами|тысяче|тысячей|тысячным|тысячный|тысячной|тысячам|тысячною|тысячах|тысячному|тысяча|тысячное)|(?&lt;_900&gt;девятисотая|девятисотого|девятисотом|девятисот|девятисотых|девятисотыми|девятистах|девятисотые|девятисотому|девятисотым|девятисотое|девятьюстами|девятисотую|девятисотой|девятистам|девятьсот|девятисотый)|(?&lt;_800&gt;восьмисотым|восьмистах|восьмисотая|восьмисотых|восьмисотое|восьмисотую|восьмисотые|восьмистам|восьмисотому|восьмисотой|восьмисот|восьмисотый|восемьсот|восемьюстами|восьмисотыми|восьмисотом|восьмисотого)|(?&lt;_700&gt;семисотого|семисотое|семисотая|семисотому|семисотой|семисотые|семисот|семистах|семисотую|семисотым|семьюстами|семисотыми|семисотый|семисотом|семистам|семисотых|семьсот)|(?&lt;_600&gt;шестисотые|шестисотой|шестьсот|шестистах|шестисотое|шестисотая|шестисотый|шестисотых|шестисотую|шестисотому|шестисот|шестистам|шестьюстами|шестисотого|шестисотым|шестисотом|шестисотыми)|(?&lt;_500&gt;пятьюстами|пятисотого|пятисотый|пятьсот|пятисотом|пятистах|пятисотой|пятисотое|пятисотая|пятисотые|пятисотых|пятисотыми|пятистам|пятисотому|пятисотым|пятисотую|пятисот)|(?&lt;_400&gt;четырьмястами|четыремстам|четырёхсотый|четырёхсотая|четырёхсотой|четырёхсотых|четырёхсотом|четырёхсотым|четыреста|четырёхсотого|четырехсот|четырёхсотому|четырёхсотое|четырёхсотыми|четырехстах|четырёхсотую|четырёхсотые)|(?&lt;_300&gt;тремястами|триста|трёхсотую|трёхсотые|трёхсотой|трехсот|тремстам|трёхсотый|трехстах|трёхсотому|трёхсотого|трёхсотом|трёхсотая|трёхсотым|трёхсотыми|трёхсотое|трёхсотых)|(?&lt;_200&gt;двухсотым|двухстах|двухсотое|двумстам|двухсотой|двухсотыми|двухсотого|двухсотый|двухсотом|двухсотую|двухсотая|двухсотые|двухсот|двухсотому|двухсотых|двести|двумястами)|(?&lt;_100&gt;сотням|сотней|сотый|сотнею|сотом|сотнями|сот|сотых|сотни|сотая|сотого|сотому|сотые|сотню|сотыми|сотым|сотнях|сотую|ста|сто|сотня|сотен|сотое|сотой|сотне)|(?&lt;_90&gt;девяностом|девяностое|девяностого|девяностый|девяностым|девяностые|девяностой|девяностую|девяностыми|девяностых|девяностая|девяностому)|(?&lt;_80&gt;восьмидесятую|восьмидесятое|восьмидесятая|восьмидесятых|восьмидесятом|восьмидесяти|восьмидесятому|восьмидесятой|восемьюдесятью|восьмидесятого|восьмьюдесятью|восемьдесят|восьмидесятый|восьмидесятым|восьмидесятыми|восьмидесятые)|(?&lt;_70&gt;семидесятые|семидесятого|семидесятый|семидесятой|семидесятым|семидесяти|семидесятыми|семидесятом|семидесятых|семидесятую|семидесятое|семидесятому|семьдесят|семьюдесятью|семидесятая)|(?&lt;_60&gt;шестидесятое|шестидесятого|шестидесятый|шестидесятых|шестидесятом|шестидесяти|шестидесятому|шестидесятую|шестидесятые|шестидесятыми|шестидесятой|шестидесятая|шестьдесят|шестьюдесятью|шестидесятым)|(?&lt;_50&gt;пятидесятом|пятидесятые|пятидесяти|пятьдесят|пятидесятого|пятидесятая|пятидесятыми|пятидесятых|пятидесятым|полсотни|пятьюдесятью|пятидесятую|пятидесятое|пятидесятой|пятидесятому|пятидесятый)|(?&lt;_40&gt;сороковым|сороковому|сороковые|сорок|сорока|сорокового|сороковых|сороковое|сороковом|сороковую|сороковой|сороковая|сороковыми)|(?&lt;_30&gt;тридцатой|тридцатый|тридцатых|тридцати|тридцатому|тридцатую|тридцать|тридцатью|тридцатыми|тридцатое|тридцатые|тридцатая|тридцатом|тридцатым|тридцатого)|(?&lt;_20&gt;двадцатое|двадцатый|двадцатого|двадцатых|двадцать|двадцатому|двадцатым|двадцатую|двадцатью|двадцатыми|двадцатые|двадцати|двадцатой|двадцатая|двадцатом)|(?&lt;_19&gt;девятнадцатой|девятнадцатым|девятнадцатый|девятнадцатые|девятнадцатому|девятнадцатью|девятнадцатого|девятнадцатых|девятнадцатыми|девятнадцати|девятнадцать|девятнадцатую|девятнадцатом|девятнадцатая|девятнадцатое)|(?&lt;_18&gt;восемнадцатом|восемнадцатый|восемнадцатую|восемнадцатому|восемнадцатая|восемнадцать|восемнадцатого|восемнадцатые|восемнадцати|восемнадцатыми|восемнадцатым|восемнадцатое|восемнадцатых|восемнадцатью|восемнадцатой)|(?&lt;_17&gt;семнадцатые|семнадцатому|семнадцатый|семнадцатой|семнадцатом|семнадцатая|семнадцатым|семнадцатое|семнадцатью|семнадцатыми|семнадцати|семнадцатую|семнадцатых|семнадцать|семнадцатого)|(?&lt;_16&gt;шестнадцатую|шестнадцатыми|шестнадцатые|шестнадцать|шестнадцатом|шестнадцатым|шестнадцатое|шестнадцатью|шестнадцатого|шестнадцатый|шестнадцатых|шестнадцати|шестнадцатой|шестнадцатая|шестнадцатому)|(?&lt;_15&gt;пятнадцати|пятнадцатом|пятнадцатого|пятнадцатью|пятнадцатое|пятнадцатую|пятнадцатым|пятнадцатый|пятнадцатая|пятнадцатых|пятнадцатые|пятнадцатой|пятнадцать|пятнадцатыми|пятнадцатому)|(?&lt;_14&gt;четырнадцатью|четырнадцатый|четырнадцатые|четырнадцатой|четырнадцатого|четырнадцатых|четырнадцатыми|четырнадцатом|четырнадцатому|четырнадцатую|четырнадцати|четырнадцать|четырнадцатое|четырнадцатая|четырнадцатым)|(?&lt;_13&gt;тринадцатого|тринадцатое|тринадцатому|тринадцатым|тринадцатью|тринадцатых|тринадцатыми|тринадцатые|тринадцатую|тринадцать|тринадцатой|тринадцати|тринадцатая|тринадцатый|тринадцатом)|(?&lt;_12&gt;дюжину|двенадцатому|двенадцати|дюжин|дюжины|двенадцатым|дюжинах|двенадцатое|двенадцатые|дюжинами|дюжине|двенадцатых|двенадцатыми|двенадцатом|двенадцатую|дюжиной|двенадцатью|двенадцатого|дюжина|двенадцатая|двенадцатой|дюжиною|двенадцать|дюжинам|двенадцатый)|(?&lt;_11&gt;одиннадцати|одиннадцатому|одиннадцатом|одиннадцатью|одиннадцатых|одиннадцатая|одиннадцатые|одиннадцатый|одиннадцатыми|одиннадцатым|одиннадцатой|одиннадцать|одиннадцатую|одиннадцатого|одиннадцатое)|(?&lt;_10&gt;десятый|десятая|десятыми|десяти|десятое|десятым|десятью|десятому|десять|десятого|десятую|десятые|десятых|десятой|десятом)|(?&lt;_9&gt;девятое|девять|девятого|девятая|девятому|девятую|девятый|девятом|девятых|девяти|девятью|девятым|девятые|девятой|девятыми)|(?&lt;_8&gt;восьмыми|восьмую|восьмью|восьмой|восьмом|восьми|восьмое|восьмому|восемь|восемью|восьмых|восьмого|восьмые|восьмая|восьмым)|(?&lt;_7&gt;седьмом|седьмая|седьмое|седьмую|семь|седьмых|седьмые|седьмой|седьмым|седьмыми|семью|семи|седьмому|седьмого)|(?&lt;_6&gt;шестых|шестое|шестые|шестую|шестью|шестого|шестым|шестом|шести|шестому|шестой|шестыми|шесть|шестая)|(?&lt;_5&gt;пятых|пятой|пяти|пятую|пятью|пятое|пятая|пятого|пятым|пятому|пятыми|пять|пятом|пятый|пятые)|(?&lt;_4&gt;четвёртом|четвёртого|четвёртые|четвёртым|четвёртая|четырех|четвёртый|четвёртую|четвёртому|четвёртой|четыре|четырем|четвёртое|четырьмя|четвёртых|четвёртыми)|(?&lt;_3&gt;третьего|третьему|третьим|третьих|третье|третий|третьем|третьей|третью|три|трем|трех|третьи|третьею|тремя|третьими|третья)|(?&lt;_2&gt;двух|вторыми|вторых|вторую|двумя|второму|две|двум|вторая|второй|два|вторым|второе|вторые|второго|втором)|(?&lt;_1&gt;одна|одною|первыми|первая|первому|одну|первом|первым|одному|первого|одно|одним|одни|первых|одном|первое|одного|одной|одних|первой|одними|первый|первые|первую|один)|(?&lt;_0&gt;нулём|нулевой|нулевую|нолями|нулевыми|нулевого|нулю|нулевое|нулевая|нолям|нолю|нули|нулям|нулевых|ноле|ноль|нулевым|нолём|нолей|нулями|нулевом|нуле|нуль|нулевые|нуля|нулей|ноля|нолях|ноли|нулевому|нулях|нулевою))+$"><N v="0"><W i="230" k="p" t="нулевой" /><W i="231" k="p" t="нулевая" /><W i="232" k="p" t="нулевое" /><W i="233" k="p" t="нулевые" /><W i="234" k="p" t="нулевого" /><W i="235" k="p" t="нулевой" /><W i="236" k="p" t="нулевого" /><W i="237" k="p" t="нулевых" /><W i="238" k="p" t="нулевому" /><W i="239" k="p" t="нулевой" /><W i="240" k="p" t="нулевому" /><W i="241" k="p" t="нулевым" /><W i="242" k="p" t="нулевого" /><W i="242" k="p" t="нулевой" /><W i="243" k="p" t="нулевую" /><W i="244" k="p" t="нулевое" /><W i="245" k="p" t="нулевых" /><W i="245" k="p" t="нулевые" /><W i="246" k="p" t="нулевым" /><W i="247" k="p" t="нулевой" /><W i="247" k="p" t="нулевою" /><W i="248" k="p" t="нулевым" /><W i="249" k="p" t="нулевыми" /><W i="250" k="p" t="нулевом" /><W i="251" k="p" t="нулевой" /><W i="252" k="p" t="нулевом" /><W i="253" k="p" t="нулевых" /><W i="230" k="o" t="ноль" /><W i="254" k="o" t="ноли" /><W i="234" k="o" t="ноля" /><W i="255" k="o" t="нолей" /><W i="238" k="o" t="нолю" /><W i="256" k="o" t="нолям" /><W i="242" k="o" t="ноль" /><W i="257" k="o" t="ноли" /><W i="246" k="o" t="нолём" /><W i="258" k="o" t="нолями" /><W i="250" k="o" t="ноле" /><W i="259" k="o" t="нолях" /><W i="230" k="o1" t="нуль" /><W i="254" k="o1" t="нули" /><W i="234" k="o1" t="нул��" /><W i="255" k="o1" t="нулей" /><W i="238" k="o1" t="нулю" /><W i="256" k="o1" t="нулям" /><W i="242" k="o1" t="нуль" /><W i="257" k="o1" t="нули" /><W i="246" k="o1" t="нулём" /><W i="258" k="o1" t="нулями" /><W i="250" k="o1" t="нуле" /><W i="259" k="o1" t="нулях" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="х" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="1"><W i="230" k="p" t="первый" /><W i="231" k="p" t="первая" /><W i="232" k="p" t="первое" /><W i="233" k="p" t="первые" /><W i="234" k="p" t="первого" /><W i="235" k="p" t="первой" /><W i="236" k="p" t="первого" /><W i="237" k="p" t="первых" /><W i="238" k="p" t="первому" /><W i="239" k="p" t="первой" /><W i="240" k="p" t="первому" /><W i="241" k="p" t="первым" /><W i="242" k="p" t="первого" /><W i="242" k="p" t="первый" /><W i="243" k="p" t="первую" /><W i="244" k="p" t="первое" /><W i="245" k="p" t="первые" /><W i="245" k="p" t="первых" /><W i="246" k="p" t="первым" /><W i="247" k="p" t="первой" /><W i="248" k="p" t="первым" /><W i="249" k="p" t="первыми" /><W i="250" k="p" t="первом" /><W i="251" k="p" t="первой" /><W i="252" k="p" t="первом" /><W i="253" k="p" t="первых" /><W i="230" k="o" t="один" /><W i="231" k="o" t="одна" /><W i="232" k="o" t="одно" /><W i="233" k="o" t="одни" /><W i="234" k="o" t="одного" /><W i="235" k="o" t="одной" /><W i="236" k="o" t="одного" /><W i="237" k="o" t="одних" /><W i="238" k="o" t="одному" /><W i="239" k="o" t="одной" /><W i="240" k="o" t="одному" /><W i="241" k="o" t="одним" /><W i="243" k="o" t="одного" /><W i="242" k="o" t="один" /><W i="243" k="o" t="одну" /><W i="244" k="o" t="одно" /><W i="245" k="o" t="одни" /><W i="245" k="o" t="одних" /><W i="246" k="o" t="одним" /><W i="247" k="o" t="одной" /><W i="248" k="o" t="одним" /><W i="249" k="o" t="одними" /><W i="250" k="o" t="одном" /><W i="251" k="o" t="одной" /><W i="252" k="o" t="одном" /><W i="253" k="o" t="одних" /><W i="260" k="o" t="один" /><W i="261" k="o" t="одни" /><W i="260" k="o" t="одна" /><W i="262" k="o" t="одно" /><W i="263" k="o" t="одних" /><W i="264" k="o" t="одного" /><W i="264" k="o" t="одной" /><W i="265" k="o" t="одного" /><W i="266" k="o" t="одним" /><W i="267" k="o" t="одному" /><W i="267" k="o" t="одной" /><W i="268" k="o" t="одному" /><W i="269" k="o" t="одни" /><W i="269" k="o" t="одних" /><W i="270" k="o" t="один" /><W i="270" k="o" t="одного" /><W i="271" k="o" t="одну" /><W i="272" k="o" t="одно" /><W i="273" k="o" t="одними" /><W i="274" k="o" t="одним" /><W i="275" k="o" t="одной" /><W i="275" k="o" t="одною" /><W i="276" k="o" t="одним" /><W i="277" k="o" t="одном" /><W i="277" k="o" t="одной" /><W i="278" k="o" t="одном" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="2"><W i="230" k="p" t="второй" /><W i="231" k="p" t="вторая" /><W i="232" k="p" t="второе" /><W i="233" k="p" t="вторые" /><W i="234" k="p" t="второго" /><W i="235" k="p" t="второй" /><W i="236" k="p" t="второго" /><W i="237" k="p" t="вторых" /><W i="238" k="p" t="второму" /><W i="239" k="p" t="второй" /><W i="240" k="p" t="второму" /><W i="241" k="p" t="вторым" /><W i="243" k="p" t="второго" /><W i="242" k="p" t="второй" /><W i="243" k="p" t="вторую" /><W i="244" k="p" t="второе" /><W i="245" k="p" t="вторые" /><W i="245" k="p" t="вторых" /><W i="246" k="p" t="вторым" /><W i="247" k="p" t="второй" /><W i="248" k="p" t="вторым" /><W i="249" k="p" t="вторыми" /><W i="250" k="p" t="втором" /><W i="251" k="p" t="второй" /><W i="252" k="p" t="втором" /><W i="253" k="p" t="вторых" /><W i="279" k="o" t="два" /><W i="260" k="o" t="две" /><W i="262" k="o" t="два" /><W i="280" k="o" t="двух" /><W i="264" k="o" t="двух" /><W i="265" k="o" t="двух" /><W i="281" k="o" t="двум" /><W i="267" k="o" t="двум" /><W i="268" k="o" t="двум" /><W i="270" k="o" t="два" /><W i="270" k="o" t="двух" /><W i="271" k="o" t="две" /><W i="271" k="o" t="двух" /><W i="272" k="o" t="два" /><W i="274" k="o" t="двумя" /><W i="275" k="o" t="двумя" /><W i="276" k="o" t="двумя" /><W i="282" k="o" t="двух" /><W i="277" k="o" t="двух" /><W i="278" k="o" t="двух" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="243" t="го" /><E i="242" t="й" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="3"><W i="230" k="p" t="третий" /><W i="231" k="p" t="третья" /><W i="232" k="p" t="третье" /><W i="233" k="p" t="третьи" /><W i="234" k="p" t="третьего" /><W i="235" k="p" t="третьей" /><W i="236" k="p" t="третьего" /><W i="237" k="p" t="третьих" /><W i="238" k="p" t="третьему" /><W i="239" k="p" t="третьей" /><W i="240" k="p" t="третьему" /><W i="241" k="p" t="третьим" /><W i="242" k="p" t="третьего" /><W i="242" k="p" t="третий" /><W i="243" k="p" t="третью" /><W i="244" k="p" t="третье" /><W i="245" k="p" t="третьи" /><W i="245" k="p" t="третьих" /><W i="246" k="p" t="третьим" /><W i="247" k="p" t="третьей" /><W i="247" k="p" t="третьею" /><W i="248" k="p" t="третьим" /><W i="249" k="p" t="третьими" /><W i="250" k="p" t="третьем" /><W i="251" k="p" t="третьей" /><W i="252" k="p" t="третьем" /><W i="253" k="p" t="третьих" /><W i="261" k="o" t="три" /><W i="263" k="o" t="трех" /><W i="266" k="o" t="трем" /><W i="269" k="o" t="трех" /><W i="269" k="o" t="три" /><W i="273" k="o" t="тремя" /><W i="283" k="o" t="трех" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="и" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="и" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="4"><W i="230" k="p" t="четвёртый" /><W i="231" k="p" t="четвёртая" /><W i="232" k="p" t="четвёртое" /><W i="233" k="p" t="четвёртые" /><W i="234" k="p" t="четвёртого" /><W i="235" k="p" t="четвёртой" /><W i="236" k="p" t="четвёртого" /><W i="237" k="p" t="четвёртых" /><W i="238" k="p" t="четвёртому" /><W i="239" k="p" t="четвёртой" /><W i="240" k="p" t="четвёртому" /><W i="241" k="p" t="четвёртым" /><W i="242" k="p" t="четвёртого" /><W i="242" k="p" t="четвёртый" /><W i="243" k="p" t="четвёртую" /><W i="244" k="p" t="четвёртое" /><W i="245" k="p" t="четвёртые" /><W i="245" k="p" t="четвёртых" /><W i="246" k="p" t="четвёртым" /><W i="247" k="p" t="четвёртой" /><W i="248" k="p" t="четвёртым" /><W i="249" k="p" t="четвёртыми" /><W i="250" k="p" t="четвёртом" /><W i="251" k="p" t="четвёртой" /><W i="252" k="p" t="четвёртом" /><W i="253" k="p" t="четвёртых" /><W i="261" k="o" t="четыре" /><W i="263" k="o" t="четырех" /><W i="266" k="o" t="четырем" /><W i="269" k="o" t="четыре" /><W i="269" k="o" t="четырех" /><W i="273" k="o" t="четырьмя" /><W i="283" k="o" t="четырех" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="5"><W i="230" k="p" t="пятый" /><W i="231" k="p" t="пятая" /><W i="232" k="p" t="пятое" /><W i="233" k="p" t="пятые" /><W i="234" k="p" t="пятого" /><W i="235" k="p" t="пятой" /><W i="236" k="p" t="пятого" /><W i="237" k="p" t="пятых" /><W i="238" k="p" t="пятому" /><W i="239" k="p" t="пятой" /><W i="240" k="p" t="пятому" /><W i="241" k="p" t="пятым" /><W i="243" k="p" t="пятого" /><W i="242" k="p" t="пятый" /><W i="243" k="p" t="пятую" /><W i="244" k="p" t="пятое" /><W i="245" k="p" t="пятые" /><W i="245" k="p" t="пятых" /><W i="246" k="p" t="пятым" /><W i="247" k="p" t="пятой" /><W i="248" k="p" t="пятым" /><W i="249" k="p" t="пятыми" /><W i="250" k="p" t="пятом" /><W i="251" k="p" t="пятой" /><W i="252" k="p" t="пятом" /><W i="253" k="p" t="пятых" /><W i="261" k="o" t="пять" /><W i="263" k="o" t="пяти" /><W i="266" k="o" t="пяти" /><W i="269" k="o" t="пять" /><W i="273" k="o" t="пятью" /><W i="283" k="o" t="пяти" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="243" t="го" /><E i="242" t="й" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="6"><W i="230" k="p" t="шестой" /><W i="231" k="p" t="шестая" /><W i="232" k="p" t="шестое" /><W i="233" k="p" t="шестые" /><W i="234" k="p" t="шестого" /><W i="235" k="p" t="шестой" /><W i="236" k="p" t="шестого" /><W i="237" k="p" t="шестых" /><W i="238" k="p" t="шестому" /><W i="239" k="p" t="шестой" /><W i="240" k="p" t="шестому" /><W i="241" k="p" t="шестым" /><W i="242" k="p" t="шестого" /><W i="242" k="p" t="шестой" /><W i="243" k="p" t="шестую" /><W i="244" k="p" t="шестое" /><W i="245" k="p" t="шестые" /><W i="245" k="p" t="шестых" /><W i="246" k="p" t="шестым" /><W i="247" k="p" t="шестой" /><W i="248" k="p" t="шестым" /><W i="249" k="p" t="шестыми" /><W i="250" k="p" t="шестом" /><W i="251" k="p" t="шестой" /><W i="252" k="p" t="шестом" /><W i="253" k="p" t="шестых" /><W i="261" k="o" t="шесть" /><W i="263" k="o" t="шести" /><W i="266" k="o" t="шести" /><W i="269" k="o" t="шесть" /><W i="273" k="o" t="шестью" /><W i="283" k="o" t="шести" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="7"><W i="230" k="p" t="седьмой" /><W i="231" k="p" t="седьмая" /><W i="232" k="p" t="седьмое" /><W i="233" k="p" t="седьмые" /><W i="234" k="p" t="седьмого" /><W i="235" k="p" t="седьмой" /><W i="236" k="p" t="седьмого" /><W i="237" k="p" t="седьмых" /><W i="238" k="p" t="седьмому" /><W i="239" k="p" t="седьмой" /><W i="240" k="p" t="седьмому" /><W i="241" k="p" t="седьмым" /><W i="242" k="p" t="седьмого" /><W i="242" k="p" t="седьмой" /><W i="243" k="p" t="седьмую" /><W i="244" k="p" t="седьмое" /><W i="245" k="p" t="седьмые" /><W i="245" k="p" t="седьмых" /><W i="246" k="p" t="седьмым" /><W i="247" k="p" t="седьмой" /><W i="248" k="p" t="седьмым" /><W i="249" k="p" t="седьмыми" /><W i="250" k="p" t="седьмом" /><W i="251" k="p" t="седьмой" /><W i="252" k="p" t="седьмом" /><W i="253" k="p" t="седьмых" /><W i="261" k="o" t="семь" /><W i="263" k="o" t="семи" /><W i="266" k="o" t="семи" /><W i="269" k="o" t="семь" /><W i="273" k="o" t="семью" /><W i="283" k="o" t="семи" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="8"><W i="230" k="p" t="восьмой" /><W i="231" k="p" t="восьмая" /><W i="232" k="p" t="восьмое" /><W i="233" k="p" t="восьмые" /><W i="234" k="p" t="восьмого" /><W i="235" k="p" t="восьмой" /><W i="236" k="p" t="восьмого" /><W i="237" k="p" t="восьмых" /><W i="238" k="p" t="восьмому" /><W i="239" k="p" t="восьмой" /><W i="240" k="p" t="восьмому" /><W i="241" k="p" t="восьмым" /><W i="242" k="p" t="восьмого" /><W i="242" k="p" t="восьмой" /><W i="243" k="p" t="восьмую" /><W i="244" k="p" t="восьмое" /><W i="245" k="p" t="восьмые" /><W i="245" k="p" t="восьмых" /><W i="246" k="p" t="восьмым" /><W i="247" k="p" t="восьмой" /><W i="248" k="p" t="восьмым" /><W i="249" k="p" t="восьмыми" /><W i="250" k="p" t="восьмом" /><W i="251" k="p" t="восьмой" /><W i="252" k="p" t="восьмом" /><W i="253" k="p" t="восьмых" /><W i="261" k="o" t="восемь" /><W i="263" k="o" t="восьми" /><W i="266" k="o" t="восьми" /><W i="269" k="o" t="восемь" /><W i="273" k="o" t="восемью" /><W i="273" k="o" t="восьмью" /><W i="283" k="o" t="восьми" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="9"><W i="230" k="p" t="девятый" /><W i="231" k="p" t="девятая" /><W i="232" k="p" t="девятое" /><W i="233" k="p" t="девятые" /><W i="234" k="p" t="девятого" /><W i="235" k="p" t="девятой" /><W i="236" k="p" t="девятого" /><W i="237" k="p" t="девятых" /><W i="238" k="p" t="девятому" /><W i="239" k="p" t="девятой" /><W i="240" k="p" t="девятому" /><W i="241" k="p" t="девятым" /><W i="242" k="p" t="девятого" /><W i="242" k="p" t="девятый" /><W i="243" k="p" t="девятую" /><W i="244" k="p" t="девятое" /><W i="245" k="p" t="девятые" /><W i="245" k="p" t="девятых" /><W i="246" k="p" t="девятым" /><W i="247" k="p" t="девятой" /><W i="248" k="p" t="девятым" /><W i="249" k="p" t="девятыми" /><W i="250" k="p" t="девятом" /><W i="251" k="p" t="девятой" /><W i="252" k="p" t="девятом" /><W i="253" k="p" t="девятых" /><W i="261" k="o" t="девять" /><W i="263" k="o" t="девяти" /><W i="266" k="o" t="девяти" /><W i="269" k="o" t="девять" /><W i="273" k="o" t="девятью" /><W i="283" k="o" t="девяти" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="10"><W i="230" k="p" t="десятый" /><W i="231" k="p" t="десятая" /><W i="232" k="p" t="десятое" /><W i="233" k="p" t="десятые" /><W i="234" k="p" t="десятого" /><W i="235" k="p" t="десятой" /><W i="236" k="p" t="десятого" /><W i="237" k="p" t="десятых" /><W i="238" k="p" t="десятому" /><W i="239" k="p" t="десятой" /><W i="240" k="p" t="десятому" /><W i="241" k="p" t="десятым" /><W i="242" k="p" t="десятого" /><W i="242" k="p" t="десятый" /><W i="243" k="p" t="десятую" /><W i="244" k="p" t="десятое" /><W i="245" k="p" t="десятые" /><W i="245" k="p" t="десятых" /><W i="246" k="p" t="десятым" /><W i="247" k="p" t="десятой" /><W i="248" k="p" t="десятым" /><W i="249" k="p" t="десятыми" /><W i="250" k="p" t="десятом" /><W i="251" k="p" t="десятой" /><W i="252" k="p" t="десятом" /><W i="253" k="p" t="десятых" /><W i="261" k="o" t="десять" /><W i="263" k="o" t="десяти" /><W i="266" k="o" t="десяти" /><W i="269" k="o" t="десять" /><W i="273" k="o" t="десятью" /><W i="283" k="o" t="десяти" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="11"><W i="230" k="p" t="одиннадцатый" /><W i="231" k="p" t="одиннадцатая" /><W i="232" k="p" t="одиннадцатое" /><W i="233" k="p" t="одиннадцатые" /><W i="234" k="p" t="одиннадцатого" /><W i="235" k="p" t="одиннадцатой" /><W i="236" k="p" t="одиннадцатого" /><W i="237" k="p" t="одиннадцатых" /><W i="238" k="p" t="одиннадцатому" /><W i="239" k="p" t="одиннадцатой" /><W i="240" k="p" t="одиннадцатому" /><W i="241" k="p" t="одиннадцат��м" /><W i="242" k="p" t="одиннадцатого" /><W i="242" k="p" t="одиннадцатый" /><W i="243" k="p" t="одиннадцатую" /><W i="244" k="p" t="одиннадцатое" /><W i="245" k="p" t="одиннадцатые" /><W i="245" k="p" t="одиннадцатых" /><W i="246" k="p" t="одиннадцатым" /><W i="247" k="p" t="одиннадцатой" /><W i="248" k="p" t="одиннадцатым" /><W i="249" k="p" t="одиннадцатыми" /><W i="250" k="p" t="одиннадцатом" /><W i="251" k="p" t="одиннадцатой" /><W i="252" k="p" t="одиннадцатом" /><W i="253" k="p" t="одиннадцатых" /><W i="261" k="o" t="одиннадцать" /><W i="263" k="o" t="одиннадцати" /><W i="266" k="o" t="одиннадцати" /><W i="269" k="o" t="одиннадцать" /><W i="273" k="o" t="одиннадцатью" /><W i="283" k="o" t="одиннадцати" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="12"><W i="231" k="ad" t="дюжина" /><W i="284" k="ad" t="дюжины" /><W i="235" k="ad" t="дюжины" /><W i="285" k="ad" t="дюжин" /><W i="239" k="ad" t="дюжине" /><W i="286" k="ad" t="дюжинам" /><W i="243" k="ad" t="дюжину" /><W i="287" k="ad" t="дюжины" /><W i="247" k="ad" t="дюжиной" /><W i="247" k="ad" t="дюжиною" /><W i="288" k="ad" t="дюжинами" /><W i="251" k="ad" t="дюжине" /><W i="289" k="ad" t="дюжинах" /><W i="230" k="p" t="двенадцатый" /><W i="231" k="p" t="двенадцатая" /><W i="232" k="p" t="двенадцатое" /><W i="233" k="p" t="двенадцатые" /><W i="234" k="p" t="двенадцатого" /><W i="235" k="p" t="двенадцатой" /><W i="236" k="p" t="двенадцатого" /><W i="237" k="p" t="двенадцатых" /><W i="238" k="p" t="двенадцатому" /><W i="239" k="p" t="двенадцатой" /><W i="240" k="p" t="двенадцатому" /><W i="241" k="p" t="двенадцатым" /><W i="242" k="p" t="двенадцатого" /><W i="242" k="p" t="двенадцатый" /><W i="243" k="p" t="двенадцатую" /><W i="244" k="p" t="двенадцатое" /><W i="245" k="p" t="двенадцатые" /><W i="245" k="p" t="двенадцатых" /><W i="246" k="p" t="двенадцатым" /><W i="247" k="p" t="двенадцатой" /><W i="248" k="p" t="двенадцатым" /><W i="249" k="p" t="двенадцатыми" /><W i="250" k="p" t="двенадцатом" /><W i="251" k="p" t="двенадцатой" /><W i="252" k="p" t="двенадцатом" /><W i="253" k="p" t="двенадцатых" /><W i="261" k="o" t="двенадцать" /><W i="263" k="o" t="двенадцати" /><W i="266" k="o" t="двенадцати" /><W i="269" k="o" t="двенадцать" /><W i="273" k="o" t="двенадцатью" /><W i="283" k="o" t="двенадцати" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="13"><W i="230" k="p" t="тринадцатый" /><W i="231" k="p" t="тринадцатая" /><W i="232" k="p" t="тринадцатое" /><W i="233" k="p" t="тринадцатые" /><W i="234" k="p" t="тринадцатого" /><W i="235" k="p" t="тринадцатой" /><W i="236" k="p" t="тринадцатого" /><W i="237" k="p" t="тринадцатых" /><W i="238" k="p" t="тринадцатому" /><W i="239" k="p" t="тринадцатой" /><W i="240" k="p" t="тринадцатому" /><W i="241" k="p" t="тринадцатым" /><W i="242" k="p" t="тринадцатого" /><W i="242" k="p" t="тринадцатый" /><W i="243" k="p" t="тринадцатую" /><W i="244" k="p" t="тринадцатое" /><W i="245" k="p" t="тринадцатые" /><W i="245" k="p" t="тринадцатых" /><W i="246" k="p" t="тринадцатым" /><W i="247" k="p" t="тринадцатой" /><W i="248" k="p" t="тринадцатым" /><W i="249" k="p" t="тринадцатыми" /><W i="250" k="p" t="тринадцатом" /><W i="251" k="p" t="тринадцатой" /><W i="252" k="p" t="тринадцатом" /><W i="253" k="p" t="тринадцатых" /><W i="261" k="o" t="тринадцать" /><W i="263" k="o" t="тринадцати" /><W i="266" k="o" t="тринадцати" /><W i="269" k="o" t="тринадцать" /><W i="273" k="o" t="тринадцатью" /><W i="283" k="o" t="тринадцати" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="14"><W i="230" k="p" t="четырнадцатый" /><W i="231" k="p" t="четырнадцатая" /><W i="232" k="p" t="четырнадцатое" /><W i="233" k="p" t="четырнадцатые" /><W i="234" k="p" t="четырнадцатого" /><W i="235" k="p" t="четырнадцатой" /><W i="236" k="p" t="четырнадцатого" /><W i="237" k="p" t="четырнадцатых" /><W i="238" k="p" t="четырнадцатому" /><W i="239" k="p" t="четырнадцатой" /><W i="240" k="p" t="четырнадцатому" /><W i="241" k="p" t="четырнадцатым" /><W i="242" k="p" t="четырнадцатого" /><W i="242" k="p" t="четырнадцатый" /><W i="243" k="p" t="четырнадцатую" /><W i="244" k="p" t="четырнадцатое" /><W i="245" k="p" t="четырнадцатые" /><W i="245" k="p" t="четырнадцатых" /><W i="246" k="p" t="четырнадцатым" /><W i="247" k="p" t="четырнадцатой" /><W i="248" k="p" t="четырнадцатым" /><W i="249" k="p" t="четырнадцатыми" /><W i="250" k="p" t="четырнадцатом" /><W i="251" k="p" t="четырнадцатой" /><W i="252" k="p" t="четырнадцатом" /><W i="253" k="p" t="четырнадцатых" /><W i="261" k="o" t="четырнадцать" /><W i="263" k="o" t="четырнадцати" /><W i="266" k="o" t="четырнадцати" /><W i="269" k="o" t="четырнадцать" /><W i="273" k="o" t="четырнадцатью" /><W i="283" k="o" t="четырнадцати" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="15"><W i="230" k="p" t="пятнадцатый" /><W i="231" k="p" t="пятнадцатая" /><W i="232" k="p" t="пятнадцатое" /><W i="233" k="p" t="пятнадцатые" /><W i="234" k="p" t="пятнадцатого" /><W i="235" k="p" t="пятнадцатой" /><W i="236" k="p" t="пятнадцатого" /><W i="237" k="p" t="пятнадцатых" /><W i="238" k="p" t="пятнадцатому" /><W i="239" k="p" t="пятнадцатой" /><W i="240" k="p" t="пятнадцатому" /><W i="241" k="p" t="пятнадцатым" /><W i="242" k="p" t="пятнадцатого" /><W i="242" k="p" t="пятнадцатый" /><W i="243" k="p" t="пятнадцатую" /><W i="244" k="p" t="пятнадцатое" /><W i="245" k="p" t="пятнадцатые" /><W i="245" k="p" t="пятнадцатых" /><W i="247" k="p" t="пятнадцатым" /><W i="247" k="p" t="пятнадцатой" /><W i="248" k="p" t="пятнадцатым" /><W i="249" k="p" t="пятнадцатыми" /><W i="251" k="p" t="пятнадцатом" /><W i="251" k="p" t="пятнадцатой" /><W i="252" k="p" t="пятнадцатом" /><W i="253" k="p" t="пятнадцатых" /><W i="261" k="o" t="пятнадцать" /><W i="263" k="o" t="пятнадцати" /><W i="266" k="o" t="пятнадцати" /><W i="269" k="o" t="пятнадцать" /><W i="273" k="o" t="пятнадцатью" /><W i="283" k="o" t="пятнадцати" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="247" t="м" /><E i="248" t="м" /><E i="249" t="ми" /><E i="251" t="м" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="16"><W i="230" k="p" t="шестнадцатый" /><W i="231" k="p" t="шестнадцатая" /><W i="232" k="p" t="шестнадцатое" /><W i="233" k="p" t="шестнадцатые" /><W i="234" k="p" t="шестнадцатого" /><W i="235" k="p" t="шестнадцатой" /><W i="236" k="p" t="шестнадцатого" /><W i="237" k="p" t="шестнадцатых" /><W i="238" k="p" t="шестнадцатому" /><W i="239" k="p" t="шестнадцатой" /><W i="240" k="p" t="шестнадцатому" /><W i="241" k="p" t="шестнадцатым" /><W i="242" k="p" t="шестнадцатого" /><W i="242" k="p" t="шестнадцатый" /><W i="243" k="p" t="шестнадцатую" /><W i="244" k="p" t="шестнадцатое" /><W i="245" k="p" t="шестнадцатые" /><W i="245" k="p" t="шестнадцатых" /><W i="246" k="p" t="шестнадцатым" /><W i="247" k="p" t="шестнадцатой" /><W i="248" k="p" t="шестнадцатым" /><W i="249" k="p" t="шестнадцатыми" /><W i="250" k="p" t="шестнадцатом" /><W i="251" k="p" t="шестнадцатой" /><W i="252" k="p" t="шестнадцатом" /><W i="253" k="p" t="шестнадцатых" /><W i="261" k="o" t="шестнадцать" /><W i="263" k="o" t="шестнадцати" /><W i="266" k="o" t="шестнадцати" /><W i="269" k="o" t="шестнадцать" /><W i="273" k="o" t="шестнадцатью" /><W i="283" k="o" t="шестнадцати" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="17"><W i="230" k="p" t="семнадцатый" /><W i="231" k="p" t="семнадцатая" /><W i="232" k="p" t="семнадцатое" /><W i="233" k="p" t="семнадцатые" /><W i="234" k="p" t="семнадцатого" /><W i="235" k="p" t="семнадцатой" /><W i="236" k="p" t="семнадцатого" /><W i="237" k="p" t="семнадцатых" /><W i="238" k="p" t="семнадцатому" /><W i="239" k="p" t="семнадцатой" /><W i="240" k="p" t="семнадцатому" /><W i="241" k="p" t="семнадцатым" /><W i="242" k="p" t="семнадцатого" /><W i="242" k="p" t="семнадцатый" /><W i="243" k="p" t="семнадцатую" /><W i="244" k="p" t="семнадцатое" /><W i="245" k="p" t="семнадцатые" /><W i="245" k="p" t="семнадцатых" /><W i="246" k="p" t="семнадцатым" /><W i="247" k="p" t="семнадцатой" /><W i="248" k="p" t="семнадцатым" /><W i="249" k="p" t="семнадцатыми" /><W i="250" k="p" t="семнадцатом" /><W i="251" k="p" t="семнадцатой" /><W i="252" k="p" t="семнадцатом" /><W i="253" k="p" t="семнадцатых" /><W i="261" k="o" t="семнадцать" /><W i="263" k="o" t="семнадцати" /><W i="266" k="o" t="семнадцати" /><W i="269" k="o" t="семнадцать" /><W i="273" k="o" t="семнадцатью" /><W i="283" k="o" t="семнадцати" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="18"><W i="230" k="p" t="восемнадцатый" /><W i="231" k="p" t="восемнадцатая" /><W i="232" k="p" t="восемнадцатое" /><W i="233" k="p" t="восемнадцатые" /><W i="234" k="p" t="восемнадцатого" /><W i="235" k="p" t="восемнадцатой" /><W i="236" k="p" t="восемнадцатого" /><W i="237" k="p" t="восемнадцатых" /><W i="238" k="p" t="восемнадцатому" /><W i="239" k="p" t="восемнадцатой" /><W i="240" k="p" t="восемнадцатому" /><W i="241" k="p" t="восемнадцатым" /><W i="242" k="p" t="восемнадцатого" /><W i="242" k="p" t="восемнадцатый" /><W i="243" k="p" t="восемнадцатую" /><W i="244" k="p" t="восемнадцатое" /><W i="245" k="p" t="восемнадцатые" /><W i="245" k="p" t="восемнадцатых" /><W i="246" k="p" t="восемнадцатым" /><W i="247" k="p" t="восемнадцатой" /><W i="248" k="p" t="восемнадцатым" /><W i="249" k="p" t="восемнадцатыми" /><W i="250" k="p" t="восемнадцатом" /><W i="251" k="p" t="восемнадцатой" /><W i="252" k="p" t="восемнадцатом" /><W i="253" k="p" t="восемнадцатых" /><W i="261" k="o" t="восемнадцать" /><W i="263" k="o" t="восемнадцати" /><W i="266" k="o" t="восемнадцати" /><W i="269" k="o" t="восемнадцать" /><W i="273" k="o" t="восемнадцатью" /><W i="283" k="o" t="восемнадцати" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="19"><W i="230" k="p" t="девятнадцатый" /><W i="231" k="p" t="девятнадцатая" /><W i="232" k="p" t="девятнадцатое" /><W i="233" k="p" t="девятнадцатые" /><W i="234" k="p" t="девятнадцатого" /><W i="235" k="p" t="девятнадцатой" /><W i="236" k="p" t="девятнадцатого" /><W i="237" k="p" t="девятнадцатых" /><W i="238" k="p" t="девятнадцатому" /><W i="239" k="p" t="девятнадцатой" /><W i="240" k="p" t="девятнадцатому" /><W i="241" k="p" t="девятнадцатым" /><W i="242" k="p" t="девятнадцатого" /><W i="242" k="p" t="девятнадцатый" /><W i="243" k="p" t="девятнадцатую" /><W i="244" k="p" t="девятнадцатое" /><W i="245" k="p" t="девятнадцатые" /><W i="245" k="p" t="девятнадцатых" /><W i="246" k="p" t="девятнадцатым" /><W i="247" k="p" t="девятнадцатой" /><W i="248" k="p" t="девятнадцатым" /><W i="249" k="p" t="девятнадцатыми" /><W i="250" k="p" t="девятнадцатом" /><W i="251" k="p" t="девятнадцатой" /><W i="252" k="p" t="девятнадцатом" /><W i="253" k="p" t="девятнадцатых" /><W i="261" k="o" t="девятнадцать" /><W i="263" k="o" t="девятнадцати" /><W i="266" k="o" t="девятнадцати" /><W i="269" k="o" t="девятнадцать" /><W i="273" k="o" t="девятнадцатью" /><W i="283" k="o" t="девятнадцати" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="20"><W i="230" k="p" t="двадцатый" /><W i="231" k="p" t="двадцатая" /><W i="232" k="p" t="двадцатое" /><W i="233" k="p" t="двадцатые" /><W i="234" k="p" t="двадцатого" /><W i="235" k="p" t="двадцатой" /><W i="236" k="p" t="двадцатого" /><W i="237" k="p" t="двадцатых" /><W i="238" k="p" t="двадцатому" /><W i="239" k="p" t="двадцатой" /><W i="240" k="p" t="двадцатому" /><W i="241" k="p" t="двадцатым" /><W i="242" k="p" t="двадцатого" /><W i="242" k="p" t="двадцатый" /><W i="243" k="p" t="двадцатую" /><W i="244" k="p" t="двадцатое" /><W i="245" k="p" t="двадцатые" /><W i="245" k="p" t="двадцатых" /><W i="246" k="p" t="двадцатым" /><W i="247" k="p" t="двадцатой" /><W i="248" k="p" t="двадцатым" /><W i="249" k="p" t="двадцатыми" /><W i="250" k="p" t="двадцатом" /><W i="251" k="p" t="двадцатой" /><W i="252" k="p" t="двадцатом" /><W i="253" k="p" t="двадцатых" /><W i="261" k="o" t="двадцать" /><W i="263" k="o" t="двадцати" /><W i="266" k="o" t="двадцати" /><W i="269" k="o" t="двадцать" /><W i="273" k="o" t="двадцатью" /><W i="283" k="o" t="двадцати" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="30"><W i="230" k="p" t="тридцатый" /><W i="231" k="p" t="тридцатая" /><W i="232" k="p" t="тридцатое" /><W i="233" k="p" t="тридцатые" /><W i="234" k="p" t="тридцатого" /><W i="235" k="p" t="тридцатой" /><W i="236" k="p" t="тридцатого" /><W i="237" k="p" t="тридцатых" /><W i="238" k="p" t="тридцатому" /><W i="239" k="p" t="тридцатой" /><W i="240" k="p" t="тридцатому" /><W i="241" k="p" t="тридцатым" /><W i="242" k="p" t="тридцатого" /><W i="242" k="p" t="тридцатый" /><W i="243" k="p" t="тридцатую" /><W i="244" k="p" t="тридцатое" /><W i="245" k="p" t="тридцатые" /><W i="245" k="p" t="тридцатых" /><W i="246" k="p" t="тридцатым" /><W i="247" k="p" t="тридцатой" /><W i="248" k="p" t="тридцатым" /><W i="249" k="p" t="тридцатыми" /><W i="250" k="p" t="тридцатом" /><W i="251" k="p" t="тридцатой" /><W i="252" k="p" t="тридцатом" /><W i="253" k="p" t="тридцатых" /><W i="261" k="o" t="тридцать" /><W i="263" k="o" t="тридцати" /><W i="266" k="o" t="тридцати" /><W i="269" k="o" t="тридцать" /><W i="273" k="o" t="тридцатью" /><W i="283" k="o" t="тридцати" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="40"><W i="230" k="p" t="сороковой" /><W i="231" k="p" t="сороковая" /><W i="232" k="p" t="сороковое" /><W i="233" k="p" t="сороковые" /><W i="234" k="p" t="сорокового" /><W i="235" k="p" t="сороковой" /><W i="236" k="p" t="сорокового" /><W i="237" k="p" t="сороковых" /><W i="238" k="p" t="сороковому" /><W i="239" k="p" t="сороковой" /><W i="240" k="p" t="сороковому" /><W i="241" k="p" t="сороковым" /><W i="242" k="p" t="сорокового" /><W i="242" k="p" t="сороковой" /><W i="243" k="p" t="сороковую" /><W i="244" k="p" t="сороковое" /><W i="245" k="p" t="сороковые" /><W i="245" k="p" t="сороковых" /><W i="246" k="p" t="сороковым" /><W i="247" k="p" t="сороковой" /><W i="248" k="p" t="сороковым" /><W i="249" k="p" t="сороковыми" /><W i="250" k="p" t="сороковом" /><W i="251" k="p" t="сороковой" /><W i="252" k="p" t="сороковом" /><W i="253" k="p" t="сороковых" /><W i="261" k="o" t="сорок" /><W i="263" k="o" t="сорока" /><W i="266" k="o" t="сорока" /><W i="269" k="o" t="сорок" /><W i="273" k="o" t="сорока" /><W i="283" k="o" t="сорока" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="50"><W i="232" k="ad" t="полсотни" /><W i="244" k="ad" t="полсотни" /><W i="230" k="p" t="пятидесятый" /><W i="231" k="p" t="пятидесятая" /><W i="232" k="p" t="пятидесятое" /><W i="233" k="p" t="пятидесятые" /><W i="234" k="p" t="пятидесятого" /><W i="235" k="p" t="пятидесятой" /><W i="236" k="p" t="пятидесятого" /><W i="237" k="p" t="пятидесятых" /><W i="238" k="p" t="пятидесятому" /><W i="239" k="p" t="пятидесятой" /><W i="240" k="p" t="пятидесятому" /><W i="241" k="p" t="пятидесятым" /><W i="242" k="p" t="пятидесятого" /><W i="242" k="p" t="пятидесятый" /><W i="243" k="p" t="пятидесятую" /><W i="244" k="p" t="пятидесятое" /><W i="245" k="p" t="пятидесятые" /><W i="245" k="p" t="пятидесятых" /><W i="246" k="p" t="пятидесятым" /><W i="247" k="p" t="пятидесятой" /><W i="248" k="p" t="пятидесятым" /><W i="249" k="p" t="пятидесятыми" /><W i="250" k="p" t="пятидесятом" /><W i="251" k="p" t="пятидесятой" /><W i="252" k="p" t="пятидесятом" /><W i="253" k="p" t="пятидесятых" /><W i="261" k="o" t="пятьдесят" /><W i="263" k="o" t="пятидесяти" /><W i="266" k="o" t="пятидесяти" /><W i="269" k="o" t="пятьдесят" /><W i="273" k="o" t="пятьюдесятью" /><W i="283" k="o" t="пятидесяти" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="60"><W i="230" k="p" t="шестидесятый" /><W i="231" k="p" t="шестидесятая" /><W i="232" k="p" t="шестидесятое" /><W i="233" k="p" t="шестидесятые" /><W i="234" k="p" t="шестидесятого" /><W i="235" k="p" t="шестидесятой" /><W i="236" k="p" t="шестидесятого" /><W i="237" k="p" t="шестидесятых" /><W i="238" k="p" t="шестидесятому" /><W i="239" k="p" t="шестидесятой" /><W i="240" k="p" t="шестидесятому" /><W i="241" k="p" t="шестидесятым" /><W i="242" k="p" t="шестидесятого" /><W i="242" k="p" t="шестидесятый" /><W i="243" k="p" t="шестидесятую" /><W i="244" k="p" t="шестидесятое" /><W i="245" k="p" t="шестидесятые" /><W i="245" k="p" t="шестидесятых" /><W i="246" k="p" t="шестидесятым" /><W i="247" k="p" t="шестидесятой" /><W i="248" k="p" t="шестидесятым" /><W i="249" k="p" t="шестидесятыми" /><W i="250" k="p" t="шестидесятом" /><W i="251" k="p" t="шестидесятой" /><W i="252" k="p" t="шестидесятом" /><W i="253" k="p" t="шестидесятых" /><W i="261" k="o" t="шестьдесят" /><W i="263" k="o" t="шестидесяти" /><W i="266" k="o" t="шестидесяти" /><W i="269" k="o" t="шестьдесят" /><W i="273" k="o" t="шестьюдесятью" /><W i="283" k="o" t="шестидесяти" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="70"><W i="230" k="p" t="семидесятый" /><W i="231" k="p" t="семидесятая" /><W i="232" k="p" t="семидесятое" /><W i="233" k="p" t="семидесятые" /><W i="234" k="p" t="семидесятого" /><W i="235" k="p" t="семидесятой" /><W i="236" k="p" t="семидесятого" /><W i="237" k="p" t="семидесятых" /><W i="238" k="p" t="семидесятому" /><W i="239" k="p" t="семидесятой" /><W i="240" k="p" t="семидесятому" /><W i="241" k="p" t="семидесятым" /><W i="242" k="p" t="семидесятого" /><W i="242" k="p" t="семидесятый" /><W i="243" k="p" t="семидесятую" /><W i="244" k="p" t="семидесятое" /><W i="245" k="p" t="семидесятые" /><W i="245" k="p" t="семидесятых" /><W i="246" k="p" t="семидесятым" /><W i="247" k="p" t="семидесятой" /><W i="248" k="p" t="семидесятым" /><W i="249" k="p" t="семидесятыми" /><W i="250" k="p" t="семидесятом" /><W i="251" k="p" t="семидесятой" /><W i="252" k="p" t="семидесятом" /><W i="253" k="p" t="семидесятых" /><W i="261" k="o" t="семьдесят" /><W i="263" k="o" t="семидесяти" /><W i="266" k="o" t="семидесяти" /><W i="269" k="o" t="семьдесят" /><W i="273" k="o" t="семьюдесятью" /><W i="283" k="o" t="семидесяти" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="80"><W i="230" k="p" t="восьмидесятый" /><W i="231" k="p" t="восьмидесятая" /><W i="232" k="p" t="восьмидесятое" /><W i="233" k="p" t="восьмидесятые" /><W i="234" k="p" t="восьмидесятого" /><W i="235" k="p" t="восьмидесятой" /><W i="236" k="p" t="восьмидесятого" /><W i="237" k="p" t="восьмидесятых" /><W i="238" k="p" t="восьмидесятому" /><W i="239" k="p" t="восьмидесятой" /><W i="240" k="p" t="восьмидесятому" /><W i="241" k="p" t="восьмидесятым" /><W i="242" k="p" t="восьмидесятого" /><W i="242" k="p" t="восьмидесятый" /><W i="243" k="p" t="восьмидесятую" /><W i="244" k="p" t="восьмидесятое" /><W i="245" k="p" t="восьмидесятые" /><W i="245" k="p" t="восьмидесятых" /><W i="246" k="p" t="восьмидесятым" /><W i="247" k="p" t="восьмидесятой" /><W i="248" k="p" t="восьмидесятым" /><W i="249" k="p" t="восьмидесятыми" /><W i="250" k="p" t="восьмидесятом" /><W i="251" k="p" t="восьмидесятой" /><W i="252" k="p" t="восьмидесятом" /><W i="253" k="p" t="восьмидесятых" /><W i="261" k="o" t="восемьдесят" /><W i="263" k="o" t="восьмидесяти" /><W i="266" k="o" t="восьмидесяти" /><W i="269" k="o" t="восемьдесят" /><W i="273" k="o" t="восемьюдесятью" /><W i="273" k="o" t="восьмьюдесятью" /><W i="283" k="o" t="восьмидесяти" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="90"><W i="230" k="p" t="девяностый" /><W i="231" k="p" t="девяностая" /><W i="232" k="p" t="девяностое" /><W i="233" k="p" t="девяностые" /><W i="234" k="p" t="девяностого" /><W i="235" k="p" t="девяностой" /><W i="236" k="p" t="девяностого" /><W i="237" k="p" t="девяностых" /><W i="238" k="p" t="девяностому" /><W i="239" k="p" t="девяностой" /><W i="240" k="p" t="девяностому" /><W i="241" k="p" t="девяностым" /><W i="242" k="p" t="девяностого" /><W i="242" k="p" t="девяностый" /><W i="243" k="p" t="девяностую" /><W i="244" k="p" t="девяностое" /><W i="245" k="p" t="девяностые" /><W i="245" k="p" t="девяностых" /><W i="246" k="p" t="девяностым" /><W i="247" k="p" t="девяностой" /><W i="248" k="p" t="девяностым" /><W i="249" k="p" t="девяностыми" /><W i="250" k="p" t="девяностом" /><W i="251" k="p" t="девяностой" /><W i="252" k="p" t="девяностом" /><W i="253" k="p" t="девяностых" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="100"><W i="231" k="ad" t="сотня" /><W i="284" k="ad" t="сотни" /><W i="235" k="ad" t="сотни" /><W i="285" k="ad" t="сот" /><W i="285" k="ad" t="сотен" /><W i="239" k="ad" t="сотне" /><W i="286" k="ad" t="сотням" /><W i="243" k="ad" t="сотню" /><W i="287" k="ad" t="сотни" /><W i="247" k="ad" t="сотней" /><W i="247" k="ad" t="сотнею" /><W i="288" k="ad" t="сотнями" /><W i="251" k="ad" t="сотне" /><W i="289" k="ad" t="сотнях" /><W i="230" k="p" t="сотый" /><W i="231" k="p" t="сотая" /><W i="232" k="p" t="сотое" /><W i="233" k="p" t="сотые" /><W i="234" k="p" t="сотого" /><W i="235" k="p" t="сотой" /><W i="236" k="p" t="сотого" /><W i="237" k="p" t="сотых" /><W i="238" k="p" t="сотому" /><W i="239" k="p" t="сотой" /><W i="240" k="p" t="сотому" /><W i="241" k="p" t="сотым" /><W i="242" k="p" t="сотого" /><W i="242" k="p" t="сотый" /><W i="243" k="p" t="сотую" /><W i="244" k="p" t="сотое" /><W i="245" k="p" t="сотые" /><W i="245" k="p" t="сотых" /><W i="246" k="p" t="сотым" /><W i="247" k="p" t="сотой" /><W i="248" k="p" t="сотым" /><W i="249" k="p" t="сотыми" /><W i="250" k="p" t="сотом" /><W i="251" k="p" t="сотой" /><W i="252" k="p" t="сотом" /><W i="253" k="p" t="сотых" /><W i="261" k="o" t="сто" /><W i="263" k="o" t="ста" /><W i="266" k="o" t="ста" /><W i="269" k="o" t="сто" /><W i="273" k="o" t="ста" /><W i="283" k="o" t="ста" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="200"><W i="230" k="p" t="двухсотый" /><W i="231" k="p" t="двухсотая" /><W i="232" k="p" t="двухсотое" /><W i="233" k="p" t="двухсотые" /><W i="234" k="p" t="двухсотого" /><W i="235" k="p" t="двухсотой" /><W i="236" k="p" t="двухсотого" /><W i="237" k="p" t="двухсотых" /><W i="238" k="p" t="двухсотому" /><W i="239" k="p" t="двухсотой" /><W i="240" k="p" t="двухсотому" /><W i="241" k="p" t="двухсотым" /><W i="242" k="p" t="двухсотого" /><W i="242" k="p" t="двухсотый" /><W i="243" k="p" t="двухсотую" /><W i="244" k="p" t="двухсотое" /><W i="245" k="p" t="двухсотые" /><W i="245" k="p" t="двухсотых" /><W i="246" k="p" t="двухсотым" /><W i="247" k="p" t="двухсотой" /><W i="248" k="p" t="двухсотым" /><W i="249" k="p" t="двухсотыми" /><W i="250" k="p" t="двухсотом" /><W i="251" k="p" t="двухсотой" /><W i="252" k="p" t="двухсотом" /><W i="253" k="p" t="двухсотых" /><W i="261" k="o" t="двести" /><W i="263" k="o" t="двухсот" /><W i="266" k="o" t="двумстам" /><W i="269" k="o" t="двести" /><W i="273" k="o" t="двумястами" /><W i="283" k="o" t="двухстах" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="300"><W i="230" k="p" t="трёхсотый" /><W i="231" k="p" t="трёхсотая" /><W i="232" k="p" t="трёхсотое" /><W i="233" k="p" t="трёхсотые" /><W i="234" k="p" t="трёхсотого" /><W i="235" k="p" t="трёхсотой" /><W i="236" k="p" t="трёхсотого" /><W i="237" k="p" t="трёхсотых" /><W i="238" k="p" t="трёхсотому" /><W i="239" k="p" t="трёхсотой" /><W i="240" k="p" t="трёхсотому" /><W i="241" k="p" t="трёхсотым" /><W i="242" k="p" t="трёхсотого" /><W i="242" k="p" t="трёхсотый" /><W i="243" k="p" t="трёхсотую" /><W i="244" k="p" t="трёхсотое" /><W i="245" k="p" t="трёхсотые" /><W i="245" k="p" t="трёхсотых" /><W i="246" k="p" t="трёхсотым" /><W i="247" k="p" t="трёхсотой" /><W i="248" k="p" t="трёхсотым" /><W i="249" k="p" t="трёхсотыми" /><W i="250" k="p" t="трёхсотом" /><W i="251" k="p" t="трёхсотой" /><W i="252" k="p" t="трёхсотом" /><W i="253" k="p" t="трёхсотых" /><W i="261" k="o" t="триста" /><W i="263" k="o" t="трехсот" /><W i="266" k="o" t="тремстам" /><W i="269" k="o" t="триста" /><W i="273" k="o" t="тремястами" /><W i="283" k="o" t="трехстах" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="400"><W i="230" k="p" t="четырёхсотый" /><W i="231" k="p" t="четырёхсотая" /><W i="232" k="p" t="четырёхсотое" /><W i="233" k="p" t="четырёхсотые" /><W i="234" k="p" t="четырёхсотого" /><W i="235" k="p" t="четырёхсотой" /><W i="236" k="p" t="четырёхсотого" /><W i="237" k="p" t="четырёхсотых" /><W i="238" k="p" t="четырёхсотому" /><W i="239" k="p" t="четырёхсотой" /><W i="240" k="p" t="четырёхсотому" /><W i="241" k="p" t="четырёхсотым" /><W i="242" k="p" t="четырёхсотого" /><W i="242" k="p" t="четырёхсотый" /><W i="243" k="p" t="четырёхсотую" /><W i="244" k="p" t="четырёхсотое" /><W i="245" k="p" t="четырёхсотые" /><W i="245" k="p" t="четырёхсотых" /><W i="246" k="p" t="четырёхсотым" /><W i="247" k="p" t="четырёхсотой" /><W i="248" k="p" t="четырёхсотым" /><W i="249" k="p" t="четырёхсотыми" /><W i="250" k="p" t="четырёхсотом" /><W i="251" k="p" t="четырёхсотой" /><W i="252" k="p" t="четырёхсотом" /><W i="253" k="p" t="четырёхсотых" /><W i="261" k="o" t="четыреста" /><W i="263" k="o" t="четырехсот" /><W i="266" k="o" t="четыремстам" /><W i="269" k="o" t="четыреста" /><W i="273" k="o" t="четырьмястами" /><W i="283" k="o" t="четырехстах" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="500"><W i="230" k="p" t="пятисотый" /><W i="231" k="p" t="пятисотая" /><W i="232" k="p" t="пятисотое" /><W i="233" k="p" t="пятисотые" /><W i="234" k="p" t="пятисотого" /><W i="235" k="p" t="пятисотой" /><W i="236" k="p" t="пятисотого" /><W i="237" k="p" t="пятисотых" /><W i="238" k="p" t="пятисотому" /><W i="239" k="p" t="пятисотой" /><W i="240" k="p" t="пятисотому" /><W i="241" k="p" t="пятисотым" /><W i="242" k="p" t="пятисотого" /><W i="242" k="p" t="пятисотый" /><W i="243" k="p" t="пятисотую" /><W i="244" k="p" t="пятисотое" /><W i="245" k="p" t="пятисотые" /><W i="245" k="p" t="пятисотых" /><W i="246" k="p" t="пятисотым" /><W i="247" k="p" t="пятисотой" /><W i="248" k="p" t="пятисотым" /><W i="249" k="p" t="пятисотыми" /><W i="250" k="p" t="пятисотом" /><W i="251" k="p" t="пятисотой" /><W i="252" k="p" t="пятисотом" /><W i="253" k="p" t="пятисотых" /><W i="261" k="o" t="пятьсот" /><W i="263" k="o" t="пятисот" /><W i="266" k="o" t="пятистам" /><W i="269" k="o" t="пятьсот" /><W i="273" k="o" t="пятьюстами" /><W i="283" k="o" t="пятистах" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="600"><W i="230" k="p" t="шестисотый" /><W i="231" k="p" t="шестисотая" /><W i="232" k="p" t="шестисотое" /><W i="233" k="p" t="шестисотые" /><W i="234" k="p" t="шестисотого" /><W i="235" k="p" t="шестисотой" /><W i="236" k="p" t="шестисотого" /><W i="237" k="p" t="шестисотых" /><W i="238" k="p" t="шестисотому" /><W i="239" k="p" t="шестисотой" /><W i="240" k="p" t="шестисотому" /><W i="241" k="p" t="шестисотым" /><W i="242" k="p" t="шестисотого" /><W i="242" k="p" t="шестисотый" /><W i="243" k="p" t="шестисотую" /><W i="244" k="p" t="шестисотое" /><W i="245" k="p" t="шестисотые" /><W i="245" k="p" t="шестисотых" /><W i="246" k="p" t="шестисотым" /><W i="247" k="p" t="шестисотой" /><W i="248" k="p" t="шестисотым" /><W i="249" k="p" t="шестисотыми" /><W i="250" k="p" t="шестисотом" /><W i="251" k="p" t="шестисотой" /><W i="252" k="p" t="шестисотом" /><W i="253" k="p" t="шестисотых" /><W i="261" k="o" t="шестьсот" /><W i="263" k="o" t="шестисот" /><W i="266" k="o" t="шестистам" /><W i="269" k="o" t="шестьсот" /><W i="273" k="o" t="шестьюстами" /><W i="283" k="o" t="шестистах" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="700"><W i="230" k="p" t="семисотый" /><W i="231" k="p" t="семисотая" /><W i="232" k="p" t="семисотое" /><W i="233" k="p" t="семисотые" /><W i="234" k="p" t="семисотого" /><W i="235" k="p" t="семисотой" /><W i="236" k="p" t="семисотого" /><W i="237" k="p" t="семисотых" /><W i="238" k="p" t="семисотому" /><W i="239" k="p" t="семисотой" /><W i="240" k="p" t="семисотому" /><W i="241" k="p" t="семисотым" /><W i="242" k="p" t="семисотого" /><W i="242" k="p" t="семисотый" /><W i="243" k="p" t="семисотую" /><W i="244" k="p" t="семисотое" /><W i="245" k="p" t="семисотые" /><W i="245" k="p" t="семисотых" /><W i="246" k="p" t="семисотым" /><W i="247" k="p" t="семисотой" /><W i="248" k="p" t="сем��сотым" /><W i="249" k="p" t="семисотыми" /><W i="250" k="p" t="семисотом" /><W i="251" k="p" t="семисотой" /><W i="252" k="p" t="семисотом" /><W i="253" k="p" t="семисотых" /><W i="261" k="o" t="семьсот" /><W i="263" k="o" t="семисот" /><W i="266" k="o" t="семистам" /><W i="269" k="o" t="семьсот" /><W i="273" k="o" t="семьюстами" /><W i="283" k="o" t="семистах" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="800"><W i="230" k="p" t="восьмисотый" /><W i="231" k="p" t="восьмисотая" /><W i="232" k="p" t="восьмисотое" /><W i="233" k="p" t="восьмисотые" /><W i="234" k="p" t="восьмисотого" /><W i="235" k="p" t="восьмисотой" /><W i="236" k="p" t="восьмисотого" /><W i="237" k="p" t="восьмисотых" /><W i="238" k="p" t="восьмисотому" /><W i="239" k="p" t="восьмисотой" /><W i="240" k="p" t="восьмисотому" /><W i="241" k="p" t="восьмисотым" /><W i="242" k="p" t="восьмисотого" /><W i="242" k="p" t="восьмисотый" /><W i="243" k="p" t="восьмисотую" /><W i="244" k="p" t="восьмисотое" /><W i="245" k="p" t="восьмисотые" /><W i="245" k="p" t="восьмисотых" /><W i="246" k="p" t="восьмисотым" /><W i="247" k="p" t="восьмисотой" /><W i="248" k="p" t="восьмисотым" /><W i="249" k="p" t="восьмисотыми" /><W i="250" k="p" t="восьмисотом" /><W i="251" k="p" t="восьмисотой" /><W i="252" k="p" t="восьмисотом" /><W i="253" k="p" t="восьмисотых" /><W i="261" k="o" t="восемьсот" /><W i="263" k="o" t="восьмисот" /><W i="266" k="o" t="восьмистам" /><W i="269" k="o" t="восемьсот" /><W i="273" k="o" t="восемьюстами" /><W i="283" k="o" t="восьмистах" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="900"><W i="230" k="p" t="девятисотый" /><W i="231" k="p" t="девятисотая" /><W i="232" k="p" t="девятисотое" /><W i="233" k="p" t="девятисотые" /><W i="234" k="p" t="девятисотого" /><W i="235" k="p" t="девятисотой" /><W i="236" k="p" t="девятисотого" /><W i="237" k="p" t="девятисотых" /><W i="238" k="p" t="девятисотому" /><W i="239" k="p" t="девятисотой" /><W i="240" k="p" t="девятисотому" /><W i="241" k="p" t="девятисотым" /><W i="242" k="p" t="девятисотого" /><W i="242" k="p" t="девятисотый" /><W i="243" k="p" t="девятисотую" /><W i="244" k="p" t="девятисотое" /><W i="245" k="p" t="девятисотые" /><W i="245" k="p" t="девятисотых" /><W i="246" k="p" t="девятисотым" /><W i="247" k="p" t="девятисотой" /><W i="248" k="p" t="девятисотым" /><W i="249" k="p" t="девятисотыми" /><W i="250" k="p" t="девятисотом" /><W i="251" k="p" t="девятисотой" /><W i="252" k="p" t="девятисотом" /><W i="253" k="p" t="девятисотых" /><W i="261" k="o" t="девятьсот" /><W i="263" k="o" t="девятисот" /><W i="266" k="o" t="девятистам" /><W i="269" k="o" t="девятьсот" /><W i="273" k="o" t="девятьюстами" /><W i="283" k="o" t="девятистах" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="1000"><W i="230" k="p" t="тысячный" /><W i="231" k="p" t="тысячная" /><W i="232" k="p" t="тысячное" /><W i="233" k="p" t="тысячные" /><W i="234" k="p" t="тысячн��го" /><W i="235" k="p" t="тысячной" /><W i="236" k="p" t="тысячного" /><W i="237" k="p" t="тысячных" /><W i="238" k="p" t="тысячному" /><W i="239" k="p" t="тысячной" /><W i="240" k="p" t="тысячному" /><W i="241" k="p" t="тысячным" /><W i="242" k="p" t="тысячного" /><W i="242" k="p" t="тысячный" /><W i="243" k="p" t="тысячную" /><W i="244" k="p" t="тысячное" /><W i="245" k="p" t="тысячные" /><W i="245" k="p" t="тысячных" /><W i="246" k="p" t="тысячным" /><W i="247" k="p" t="тысячной" /><W i="247" k="p" t="тысячною" /><W i="248" k="p" t="тысячным" /><W i="249" k="p" t="тысячными" /><W i="250" k="p" t="тысячном" /><W i="251" k="p" t="тысячной" /><W i="252" k="p" t="тысячном" /><W i="253" k="p" t="тысячных" /><W i="261" k="o" t="тысяча" /><W i="261" k="o" t="тысячи" /><W i="263" k="o" t="тысяч" /><W i="263" k="o" t="тысячи" /><W i="266" k="o" t="тысячам" /><W i="266" k="o" t="тысяче" /><W i="269" k="o" t="тысячи" /><W i="269" k="o" t="тысячу" /><W i="273" k="o" t="тысячами" /><W i="273" k="o" t="тысячей" /><W i="273" k="o" t="тысячью" /><W i="283" k="o" t="тысячах" /><W i="283" k="o" t="тысяче" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="1000000"><W i="230" k="p" t="миллионный" /><W i="231" k="p" t="миллионная" /><W i="232" k="p" t="миллионное" /><W i="233" k="p" t="миллионные" /><W i="234" k="p" t="миллионного" /><W i="235" k="p" t="миллионной" /><W i="236" k="p" t="миллионного" /><W i="237" k="p" t="миллионных" /><W i="238" k="p" t="миллионному" /><W i="239" k="p" t="миллионной" /><W i="240" k="p" t="миллионному" /><W i="241" k="p" t="миллионным" /><W i="242" k="p" t="миллионного" /><W i="242" k="p" t="миллионный" /><W i="243" k="p" t="миллионную" /><W i="244" k="p" t="миллионное" /><W i="245" k="p" t="миллионные" /><W i="245" k="p" t="миллионных" /><W i="246" k="p" t="миллионным" /><W i="247" k="p" t="миллионной" /><W i="247" k="p" t="миллионною" /><W i="248" k="p" t="миллионным" /><W i="249" k="p" t="миллионными" /><W i="250" k="p" t="миллионном" /><W i="251" k="p" t="миллионной" /><W i="252" k="p" t="миллионном" /><W i="253" k="p" t="миллионных" /><W i="261" k="o" t="миллион" /><W i="263" k="o" t="миллиона" /><W i="263" k="o" t="миллионов" /><W i="266" k="o" t="миллионам" /><W i="266" k="o" t="миллиону" /><W i="269" k="o" t="миллион" /><W i="273" k="o" t="миллионами" /><W i="273" k="o" t="миллионом" /><W i="283" k="o" t="миллионах" /><W i="283" k="o" t="миллионе" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="х" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="241" t="м" /><E i="242" t="го" /><E i="243" t="ю" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N><N v="1000000000"><W i="230" k="p" t="миллиардный" /><W i="231" k="p" t="миллиардная" /><W i="232" k="p" t="миллиардное" /><W i="233" k="p" t="миллиардные" /><W i="234" k="p" t="миллиардного" /><W i="235" k="p" t="миллиардной" /><W i="236" k="p" t="миллиардного" /><W i="237" k="p" t="миллиардным" /><W i="237" k="p" t="миллиардных" /><W i="238" k="p" t="миллиардному" /><W i="239" k="p" t="миллиардной" /><W i="240" k="p" t="миллиардному" /><W i="242" k="p" t="миллиардного" /><W i="243" k="p" t="миллиардный" /><W i="243" k="p" t="миллиардную" /><W i="244" k="p" t="миллиардное" /><W i="245" k="p" t="миллиардные" /><W i="245" k="p" t="миллиардных" /><W i="246" k="p" t="миллиардным" /><W i="247" k="p" t="миллиардной" /><W i="247" k="p" t="миллиардною" /><W i="248" k="p" t="миллиардным" /><W i="249" k="p" t="миллиардными" /><W i="250" k="p" t="миллиардном" /><W i="251" k="p" t="миллиардной" /><W i="252" k="p" t="миллиардном" /><W i="253" k="p" t="миллиардных" /><W i="261" k="o" t="миллиард" /><W i="261" k="o" t="миллиарды" /><W i="263" k="o" t="миллиарда" /><W i="263" k="o" t="миллиардов" /><W i="266" k="o" t="миллиардам" /><W i="266" k="o" t="миллиарду" /><W i="269" k="o" t="миллиард" /><W i="269" k="o" t="миллиарды" /><W i="273" k="o" t="миллиардами" /><W i="273" k="o" t="миллиардом" /><W i="283" k="o" t="миллиардах" /><W i="283" k="o" t="миллиарде" /><E i="230" t="й" /><E i="231" t="я" /><E i="232" t="е" /><E i="233" t="е" /><E i="234" t="го" /><E i="235" t="й" /><E i="236" t="го" /><E i="237" t="м" /><E i="238" t="му" /><E i="239" t="й" /><E i="240" t="му" /><E i="242" t="го" /><E i="243" t="й" /><E i="244" t="е" /><E i="245" t="е" /><E i="246" t="м" /><E i="247" t="й" /><E i="248" t="м" /><E i="249" t="ми" /><E i="250" t="м" /><E i="251" t="й" /><E i="252" t="м" /><E i="253" t="х" /></N></NumbData>

latest_release/release_small.xml

@@ -0,0 +1,2 @@
+<?xml version='1.0' encoding='utf-8'?>
+<Root batch_size="BatchSize" inflect_result="Inflect/Decoder_1/decoder/transpose_1" inflect_x_cls="Inflect/XClass" inflect_y_cls="Inflect/YClass" lem_cls_result="Reshape_1" lem_result="ExpandDims" lem_x_cls="Lemm_1/XClass" prob_case="Case/Softmax" prob_gndr="Gndr/Softmax" prob_mood="Mood/Softmax" prob_nmbr="Nmbr/Softmax" prob_pers="Pers/Softmax" prob_post="Post/Softmax" prob_tens="Tens/Softmax" prob_voic="Voic/Softmax" res_case="Case/Results" res_gndr="Gndr/Results" res_indexes="Main/Results" res_mood="Mood/Results" res_nmbr="Nmbr/Results" res_pers="Pers/Results" res_post="Post/Results" res_tens="Tens/Results" res_values="Main/Results" res_voic="Voic/Results" seq_len="SeqLen" x_ind="XIndexes" x_shape="XShape" x_val="XValues"><Chars end_char="35" start_char="36"><Char index="0" value="UNDEFINED" /><Char index="1" value="а" /><Char index="2" value="б" /><Char index="3" value="в" /><Char index="4" value="г" /><Char index="5" value="д" /><Char index="6" value="е" /><Char index="7" value="ё" /><Char index="8" value="ж" /><Char index="9" value="з" /><Char index="10" value="и" /><Char index="11" value="й" /><Char index="12" value="к" /><Char index="13" value="л" /><Char index="14" value="м" /><Char index="15" value="н" /><Char index="16" value="о" /><Char index="17" value="п" /><Char index="18" value="р" /><Char index="19" value="с" /><Char index="20" value="т" /><Char index="21" value="у" /><Char index="22" value="ф" /><Char index="23" value="х" /><Char index="24" value="ч" /><Char index="25" value="ц" /><Char index="26" value="ш" /><Char index="27" value="щ" /><Char index="28" value="ъ" /><Char index="29" value="ы" /><Char index="30" value="ь" /><Char index="31" value="э" /><Char index="32" value="ю" /><Char index="33" value="я" /><Char index="34" value="-" /></Chars><Grams><G key="post" op="Post/Softmax" /><G key="gndr" op="Gndr/Softmax" /><G key="nmbr" op="Nmbr/Softmax" /><G key="case" op="Case/Softmax" /><G key="pers" op="Pers/Softmax" /><G key="tens" op="Tens/Softmax" /><G key="mood" op="Mood/Softmax" /><G key="voic" op="Voic/Softmax" /></Grams><Inflect><Im i="48"><I i="173" /><I i="83" /><I i="75" /><I i="214" /><I i="203" /><I i="49" /><I i="213" /><I i="85" /><I i="29" /><I i="82" /><I i="212" /><I i="207" /><I i="57" /><I i="28" /><I i="38" /><I i="0" /></Im><Im i="58"><I i="8" /><I i="78" /><I i="6" /><I i="68" /><I i="10" /><I i="216" /><I i="144" /><I i="5" /><I i="217" /><I i="1" /><I i="7" /><I i="205" /><I i="4" /><I i="47" /></Im><Im i="135"><I i="177" /><I i="105" /><I i="123" /><I i="79" /><I i="220" /><I i="180" /><I i="215" /><I i="141" /><I i="34" /><I i="59" /><I i="101" /><I i="61" /><I i="131" /><I i="138" /><I i="197" /><I i="176" /><I i="86" /><I i="35" /><I i="161" /><I i="3" /><I i="183" /><I i="104" /><I i="115" /><I i="195" /><I i="143" /><I i="111" /><I i="120" /><I i="108" /><I i="127" /><I i="134" /><I i="133" /><I i="158" /><I i="51" /><I i="218" /><I i="56" /><I i="26" /><I i="185" /><I i="102" /><I i="149" /><I i="11" /><I i="192" /><I i="94" /><I i="174" /><I i="46" /><I i="53" /><I i="52" /><I i="107" /><I i="76" /><I i="124" /><I i="121" /><I i="97" /><I i="151" /><I i="159" /><I i="50" /><I i="87" /><I i="9" /><I i="19" /><I i="27" /><I i="194" /><I i="226" /><I i="55" /><I i="25" /><I i="125" /><I i="71" /><I i="132" /><I i="130" /><I i="100" /><I i="160" /><I i="179" /><I i="150" /><I i="129" /><I i="118" /><I i="198" /><I i="178" /><I i="162" /><I i="65" /><I i="21" /><I i="122" /><I i="80" /><I i="201" /><I i="202" /><I i="72" /><I i="190" /><I i="126" /><I i="20" /><I i="181" /><I i="99" /><I i="184" /><I i="66" /><I i="175" /><I i="44" /><I i="60" /><I i="199" /><I i="74" /><I i="204" /><I i="95" /><I i="110" /><I i="24" /><I i="92" /><I i="116" /><I i="88" /><I i="77" /><I i="191" /><I i="43" /><I i="45" /><I i="30" /><I i="117" /><I i="54" /><I i="148" /><I i="62" /><I i="137" /><I i="64" /><I i="93" /><I i="23" /><I i="196" /><I i="22" /><I i="73" /><I i="67" /><I i="2" /><I i="114" /><I i="140" /><I i="139" /><I i="128" /><I i="109" /><I i="186" /><I i="200" /><I i="187" /><I i="193" /></Im><Im i="18"><I i="12" /><I i="63" /><I i="32" /><I i="90" /><I i="152" /><I i="36" /><I i="39" /><I i="41" /><I i="96" /><I i="98" /><I i="31" /><I i="40" /><I i="189" /><I i="103" /><I i="14" /><I i="17" /><I i="13" /><I i="89" /><I i="42" /><I i="16" /><I i="119" /><I i="70" /><I i="221" /><I i="15" /><I i="33" /><I i="69" /><I i="182" /><I i="37" /><I i="91" /></Im><Im i="112"><I i="136" /><I i="228" /><I i="157" /><I i="142" /><I i="113" /><I i="146" /><I i="156" /><I i="147" /><I i="153" /><I i="224" /><I i="227" /><I i="145" /><I i="155" /><I i="222" /><I i="154" /></Im><Im i="163"><I i="165" /><I i="168" /><I i="172" /><I i="81" /><I i="171" /><I i="164" /><I i="106" /><I i="167" /><I i="219" /><I i="166" /><I i="169" /><I i="170" /></Im></Inflect></Root>

latest_release/tags.xml

@@ -0,0 +1,2 @@
+<?xml version='1.0' encoding='utf-8'?>
+<Tags><T i="0" o="219" p="nn" v="noun,masc,sing,datv,,,," /><T i="1" o="194" p="nn" v="noun,femn,plur,gent,,,," /><T i="2" o="21" p="nn" v="prtf,neut,sing,accs,,past,,actv" /><T i="3" o="33" p="nn" v="prtf,neut,sing,nomn,,past,,actv" /><T i="4" o="202" p="nn" v="noun,femn,sing,datv,,,," /><T i="5" o="195" p="nn" v="noun,femn,plur,nomn,,,," /><T i="6" o="199" p="nn" v="noun,femn,sing,loct,,,," /><T i="7" o="203" p="nn" v="noun,femn,sing,gent,,,," /><T i="8" o="192" p="nn" v="noun,femn,plur,accs,,,," /><T i="9" o="113" p="nn" v="verb,masc,sing,,,past,indc," /><T i="10" o="190" p="nn" v="noun,femn,plur,loct,,,," /><T i="11" o="131" p="nn" v="verb,,sing,,,,impr," /><T i="12" o="152" p="nn" v="adjf,masc,sing,datv,,,," /><T i="13" o="140" p="nn" v="adjf,neut,sing,datv,,,," /><T i="14" o="141" p="nn" v="adjf,neut,sing,gent,,,," /><T i="15" o="153" p="nn" v="adjf,masc,sing,gent,,,," /><T i="16" o="151" p="nn" v="adjf,masc,sing,accs,,,," /><T i="17" o="144" p="nn" v="adjf,femn,sing,ablt,,,," /><T i="18" l="1" o="154" p="nn" v="adjf,masc,sing,nomn,,,," /><T i="19" o="71" p="nn" v="prtf,masc,sing,accs,,pres,,actv" /><T i="20" o="83" p="nn" v="prtf,masc,sing,nomn,,pres,,actv" /><T i="21" o="46" p="nn" v="prtf,femn,sing,accs,,pres,,pssv" /><T i="22" o="75" p="nn" v="prtf,masc,sing,datv,,pres,,actv" /><T i="23" o="27" p="nn" v="prtf,neut,sing,datv,,pres,,actv" /><T i="24" o="51" p="nn" v="prtf,femn,sing,datv,,pres,,actv" /><T i="25" o="55" p="nn" v="prtf,femn,sing,gent,,pres,,actv" /><T i="26" o="43" p="nn" v="prtf,femn,sing,ablt,,pres,,actv" /><T i="27" o="39" p="nn" v="prtf,femn,sing,loct,,pres,,actv" /><T i="28" o="210" p="nn" v="noun,masc,plur,nomn,,,," /><T i="29" o="207" p="nn" v="noun,masc,plur,accs,,,," /><T i="30" o="121" p="nn" v="verb,,plur,,,past,indc," /><T i="31" o="158" p="nn" v="adjf,,plur,datv,,,," /><T i="32" o="138" p="nn" v="adjf,neut,sing,ablt,,,," /><T i="33" o="150" p="nn" v="adjf,masc,sing,ablt,,,," /><T i="34" o="8" p="nn" v="prts,masc,sing,,,past,,pssv" /><T i="35" o="115" p="nn" v="verb,,plur,,3per,futr,indc," /><T i="36" o="139" p="nn" v="adjf,neut,sing,accs,,,," /><T i="37" o="142" p="nn" v="adjf,neut,sing,nomn,,,," /><T i="38" o="209" p="nn" v="noun,masc,plur,gent,,,," /><T i="39" o="148" p="nn" v="adjf,femn,sing,nomn,,,," /><T i="40" o="143" p="nn" v="adjf,femn,sing,loct,,,," /><T i="41" o="146" p="nn" v="adjf,femn,sing,datv,,,," /><T i="42" o="147" p="nn" v="adjf,femn,sing,gent,,,," /><T i="43" o="103" p="nn" v="prtf,,plur,gent,,pres,,actv" /><T i="44" o="87" p="nn" v="prtf,,plur,loct,,pres,,actv" /><T i="45" o="95" p="nn" v="prtf,,plur,accs,,pres,,actv" /><T i="46" o="2" p="nn" v="grnd,,,,,past,," /><T i="47" o="200" p="nn" v="noun,femn,sing,ablt,,,," /><T i="48" l="1" o="221" p="nn" v="noun,masc,sing,nomn,,,," /><T i="49" o="218" p="nn" v="noun,masc,sing,accs,,,," /><T i="50" o="99" p="nn" v="prtf,,plur,datv,,pres,,actv" /><T i="51" o="67" p="nn" v="prtf,masc,sing,ablt,,pres,,actv" /><T i="52" o="19" p="nn" v="prtf,neut,sing,ablt,,pres,,actv" /><T i="53" o="53" p="nn" v="prtf,femn,sing,gent,,past,,actv" /><T i="54" o="37" p="nn" v="prtf,femn,sing,loct,,past,,actv" /><T i="55" o="49" p="nn" v="prtf,femn,sing,datv,,past,,actv" /><T i="56" o="41" p="nn" v="prtf,femn,sing,ablt,,past,,actv" /><T i="57" o="220" p="nn" v="noun,masc,sing,gent,,,," /><T i="58" l="1" o="204" p="nn" v="noun,femn,sing,nomn,,,," /><T i="59" o="111" p="nn" v="verb,femn,sing,,,past,indc," /><T i="60" o="65" p="nn" v="prtf,masc,sing,ablt,,past,,actv" /><T i="61" o="17" p="nn" v="prtf,neut,sing,ablt,,past,,actv" /><T i="62" o="97" p="nn" v="prtf,,plur,datv,,past,,actv" /><T i="63" o="145" p="nn" v="adjf,femn,sing,accs,,,," /><T i="64" o="72" p="nn" v="prtf,masc,sing,datv,,past,,pssv" /><T i="65" o="24" p="nn" v="prtf,neut,sing,datv,,past,,pssv" /><T i="66" o="25" p="nn" v="prtf,neut,sing,datv,,past,,actv" /><T i="67" o="73" p="nn" v="prtf,masc,sing,datv,,past,,actv" /><T i="68" o="193" p="nn" v="noun,femn,plur,datv,,,," /><T i="69" o="137" p="nn" v="adjf,neut,sing,loct,,,," /><T i="70" o="149" p="nn" v="adjf,masc,sing,loct,,,," /><T i="71" o="119" p="nn" v="verb,,plur,,1per,futr,indc," /><T i="72" o="107" p="nn" v="prtf,,plur,nomn,,pres,,actv" /><T i="73" o="10" p="nn" v="prts,,plur,,,past,,pssv" /><T i="74" o="56" p="nn" v="prtf,femn,sing,nomn,,past,,pssv" /><T i="75" o="216" p="nn" v="noun,masc,sing,loct,,,," /><T i="76" o="92" p="nn" v="prtf,,plur,accs,,past,,pssv" /><T i="77" o="104" p="nn" v="prtf,,plur,nomn,,past,,pssv" /><T i="78" o="201" p="nn" v="noun,femn,sing,accs,,,," /><T i="79" o="117" p="nn" v="verb,,plur,,2per,futr,indc," /><T i="80" o="122" p="nn" v="verb,,plur,,,,impr," /><T i="81" o="164" p="nn" v="noun,msf,plur,datv,,,," /><T i="82" o="217" p="nn" v="noun,masc,sing,ablt,,,," /><T i="83" o="208" p="nn" v="noun,masc,plur,datv,,,," /><T i="84" o="223" p="nn" v="noun,,plur,loct,,,," /><T i="85" o="205" p="nn" v="noun,masc,plur,loct,,,," /><T i="86" o="60" p="nn" v="prtf,masc,sing,loct,,past,,pssv" /><T i="87" o="12" p="nn" v="prtf,neut,sing,loct,,past,,pssv" /><T i="88" o="4" p="nn" v="prts,neut,sing,,,past,,pssv" /><T i="89" o="159" p="nn" v="adjf,,plur,gent,,,," /><T i="90" o="155" p="nn" v="adjf,,plur,loct,,,," /><T i="91" o="157" p="nn" v="adjf,,plur,accs,,,," /><T i="92" o="81" p="nn" v="prtf,masc,sing,nomn,,past,,actv" /><T i="93" o="69" p="nn" v="prtf,masc,sing,accs,,past,,actv" /><T i="94" o="63" p="nn" v="prtf,masc,sing,loct,,pres,,actv" /><T i="95" o="15" p="nn" v="prtf,neut,sing,loct,,pres,,actv" /><T i="96" o="160" p="nn" v="adjf,,plur,nomn,,,," /><T i="97" o="7" p="nn" v="prts,femn,sing,,,pres,,pssv" /><T i="98" o="108" p="nn" v="comp,,,,,,," /><T i="99" o="61" p="nn" v="prtf,masc,sing,loct,,past,,actv" /><T i="100" o="13" p="nn" v="prtf,neut,sing,loct,,past,,actv" /><T i="101" o="40" p="nn" v="prtf,femn,sing,ablt,,past,,pssv" /><T i="102" o="47" p="nn" v="prtf,femn,sing,accs,,pres,,actv" /><T i="103" o="156" p="nn" v="adjf,,plur,ablt,,,," /><T i="104" o="35" p="nn" v="prtf,neut,sing,nomn,,pres,,actv" /><T i="105" o="23" p="nn" v="prtf,neut,sing,accs,,pres,,actv" /><T i="106" o="162" p="nn" v="noun,msf,plur,ablt,,,," /><T i="107" o="70" p="nn" v="prtf,masc,sing,accs,,pres,,pssv" /><T i="108" o="78" p="nn" v="prtf,masc,sing,gent,,pres,,pssv" /><T i="109" o="30" p="nn" v="prtf,neut,sing,gent,,pres,,pssv" /><T i="110" o="77" p="nn" v="prtf,masc,sing,gent,,past,,actv" /><T i="111" o="29" p="nn" v="prtf,neut,sing,gent,,past,,actv" /><T i="112" l="1" o="189" p="nn" v="noun,neut,sing,nomn,,,," /><T i="113" o="186" p="nn" v="noun,neut,sing,accs,,,," /><T i="114" o="31" p="nn" v="prtf,neut,sing,gent,,pres,,actv" /><T i="115" o="79" p="nn" v="prtf,masc,sing,gent,,pres,,actv" /><T i="116" o="124" p="nn" v="verb,,sing,,3per,pres,indc," /><T i="117" o="118" p="nn" v="verb,,plur,,2per,pres,indc," /><T i="118" o="59" p="nn" v="prtf,femn,sing,nomn,,pres,,actv" /><T i="119" o="135" p="nn" v="adjs,masc,sing,,,,," /><T i="120" o="101" p="nn" v="prtf,,plur,gent,,past,,actv" /><T i="121" o="85" p="nn" v="prtf,,plur,loct,,past,,actv" /><T i="122" o="93" p="nn" v="prtf,,plur,accs,,past,,actv" /><T i="123" o="128" p="nn" v="verb,,sing,,1per,pres,indc," /><T i="124" o="98" p="nn" v="prtf,,plur,datv,,pres,,pssv" /><T i="125" o="18" p="nn" v="prtf,neut,sing,ablt,,pres,,pssv" /><T i="126" o="66" p="nn" v="prtf,masc,sing,ablt,,pres,,pssv" /><T i="127" o="89" p="nn" v="prtf,,plur,ablt,,past,,actv" /><T i="128" o="94" p="nn" v="prtf,,plur,accs,,pres,,pssv" /><T i="129" o="86" p="nn" v="prtf,,plur,loct,,pres,,pssv" /><T i="130" o="102" p="nn" v="prtf,,plur,gent,,pres,,pssv" /><T i="131" o="14" p="nn" v="prtf,neut,sing,loct,,pres,,pssv" /><T i="132" o="62" p="nn" v="prtf,masc,sing,loct,,pres,,pssv" /><T i="133" o="116" p="nn" v="verb,,plur,,3per,pres,indc," /><T i="134" o="88" p="nn" v="prtf,,plur,ablt,,past,,pssv" /><T i="135" l="1" o="109" p="nn" v="infn,,,,,,," /><T i="136" o="184" p="nn" v="noun,neut,sing,loct,,,," /><T i="137" o="68" p="nn" v="prtf,masc,sing,accs,,past,,pssv" /><T i="138" o="28" p="nn" v="prtf,neut,sing,gent,,past,,pssv" /><T i="139" o="76" p="nn" v="prtf,masc,sing,gent,,past,,pssv" /><T i="140" o="105" p="nn" v="prtf,,plur,nomn,,past,,actv" /><T i="141" o="42" p="nn" v="prtf,femn,sing,ablt,,pres,,pssv" /><T i="142" o="187" p="nn" v="noun,neut,sing,datv,,,," /><T i="143" o="9" p="nn" v="prts,masc,sing,,,pres,,pssv" /><T i="144" o="191" p="nn" v="noun,femn,plur,ablt,,,," /><T i="145" o="179" p="nn" v="noun,neut,plur,nomn,,,," /><T i="146" o="188" p="nn" v="noun,neut,sing,gent,,,," /><T i="147" o="176" p="nn" v="noun,neut,plur,accs,,,," /><T i="148" o="96" p="nn" v="prtf,,plur,datv,,past,,pssv" /><T i="149" o="16" p="nn" v="prtf,neut,sing,ablt,,past,,pssv" /><T i="150" o="64" p="nn" v="prtf,masc,sing,ablt,,past,,pssv" /><T i="151" o="58" p="nn" v="prtf,femn,sing,nomn,,pres,,pssv" /><T i="152" o="136" p="nn" v="adjs,,plur,,,,," /><T i="153" o="185" p="nn" v="noun,neut,sing,ablt,,,," /><T i="154" o="177" p="nn" v="noun,neut,plur,datv,,,," /><T i="155" o="175" p="nn" v="noun,neut,plur,ablt,,,," /><T i="156" o="174" p="nn" v="noun,neut,plur,loct,,,," /><T i="157" o="178" p="nn" v="noun,neut,plur,gent,,,," /><T i="158" o="100" p="nn" v="prtf,,plur,gent,,past,,pssv" /><T i="159" o="84" p="nn" v="prtf,,plur,loct,,past,,pssv" /><T i="160" o="20" p="nn" v="prtf,neut,sing,accs,,past,,pssv" /><T i="161" o="32" p="nn" v="prtf,neut,sing,nomn,,past,,pssv" /><T i="162" o="110" p="nn" v="verb,neut,sing,,,past,indc," /><T i="163" l="1" o="173" p="nn" v="noun,msf,sing,nomn,,,," /><T i="164" o="169" p="nn" v="noun,msf,sing,ablt,,,," /><T i="165" o="168" p="nn" v="noun,msf,sing,loct,,,," /><T i="166" o="170" p="nn" v="noun,msf,sing,accs,,,," /><T i="167" o="165" p="nn" v="noun,msf,plur,gent,,,," /><T i="168" o="166" p="nn" v="noun,msf,plur,nomn,,,," /><T i="169" o="171" p="nn" v="noun,msf,sing,datv,,,," /><T i="170" o="161" p="nn" v="noun,msf,plur,loct,,,," /><T i="171" o="163" p="nn" v="noun,msf,plur,accs,,,," /><T i="172" o="172" p="nn" v="noun,msf,sing,gent,,,," /><T i="173" o="206" p="nn" v="noun,masc,plur,ablt,,,," /><T i="174" o="120" p="nn" v="verb,,plur,,1per,pres,indc," /><T i="175" o="45" p="nn" v="prtf,femn,sing,accs,,past,,actv" /><T i="176" o="123" p="nn" v="verb,,sing,,3per,futr,indc," /><T i="177" o="52" p="nn" v="prtf,femn,sing,gent,,past,,pssv" /><T i="178" o="36" p="nn" v="prtf,femn,sing,loct,,past,,pssv" /><T i="179" o="48" p="nn" v="prtf,femn,sing,datv,,past,,pssv" /><T i="180" o="11" p="nn" v="prts,,plur,,,pres,,pssv" /><T i="181" o="44" p="nn" v="prtf,femn,sing,accs,,past,,pssv" /><T i="182" o="132" p="nn" v="adjs,neut,sing,,,,," /><T i="183" o="80" p="nn" v="prtf,masc,sing,nomn,,past,,pssv" /><T i="184" o="3" p="nn" v="grnd,,,,,pres,," /><T i="185" o="91" p="nn" v="prtf,,plur,ablt,,pres,,actv" /><T i="186" o="26" p="nn" v="prtf,neut,sing,datv,,pres,,pssv" /><T i="187" o="74" p="nn" v="prtf,masc,sing,datv,,pres,,pssv" /><T i="188" l="1" o="1" p="nn" v="advb,,,,,,," /><T i="189" o="133" p="nn" v="adjs,femn,sing,,,,," /><T i="190" o="125" p="nn" v="verb,,sing,,2per,futr,indc," /><T i="191" o="126" p="nn" v="verb,,sing,,2per,pres,indc," /><T i="192" o="54" p="nn" v="prtf,femn,sing,gent,,pres,,pssv" /><T i="193" o="50" p="nn" v="prtf,femn,sing,datv,,pres,,pssv" /><T i="194" o="38" p="nn" v="prtf,femn,sing,loct,,pres,,pssv" /><T i="195" o="127" p="nn" v="verb,,sing,,1per,futr,indc," /><T i="196" o="22" p="nn" v="prtf,neut,sing,accs,,pres,,pssv" /><T i="197" o="34" p="nn" v="prtf,neut,sing,nomn,,pres,,pssv" /><T i="198" o="57" p="nn" v="prtf,femn,sing,nomn,,past,,actv" /><T i="199" o="106" p="nn" v="prtf,,plur,nomn,,pres,,pssv" /><T i="200" o="90" p="nn" v="prtf,,plur,ablt,,pres,,pssv" /><T i="201" o="5" p="nn" v="prts,neut,sing,,,pres,,pssv" /><T i="202" o="82" p="nn" v="prtf,masc,sing,nomn,,pres,,pssv" /><T i="203" o="211" p="nn" v="noun,masc,sing,loct2,,,," /><T i="204" o="6" p="nn" v="prts,femn,sing,,,past,,pssv" /><T i="205" o="198" p="nn" v="noun,femn,sing,voct,,,," /><T i="206" o="227" p="nn" v="noun,,plur,gent,,,," /><T i="207" o="214" p="nn" v="noun,masc,sing,gent2,,,," /><T i="208" o="226" p="nn" v="noun,,plur,datv,,,," /><T i="209" o="225" p="nn" v="noun,,plur,accs,,,," /><T i="210" o="228" p="nn" v="noun,,plur,nomn,,,," /><T i="211" o="224" p="nn" v="noun,,plur,ablt,,,," /><T i="212" o="215" p="nn" v="noun,masc,sing,gent1,,,," /><T i="213" o="213" p="nn" v="noun,masc,sing,voct,,,," /><T i="214" o="212" p="nn" v="noun,masc,sing,loct1,,,," /><T i="215" o="130" p="nn" v="verb,,sing,,,pres,indc," /><T i="216" o="197" p="nn" v="noun,femn,sing,loct1,,,," /><T i="217" o="196" p="nn" v="noun,femn,sing,loct2,,,," /><T i="218" o="114" p="nn" v="verb,masc,sing,,,,impr," /><T i="219" o="167" p="nn" v="noun,msf,sing,voct,,,," /><T i="220" o="129" p="nn" v="verb,,sing,,,futr,indc," /><T i="221" o="134" p="nn" v="adjs,masc,sing,gent,,,," /><T i="222" o="180" p="nn" v="noun,neut,sing,loct2,,,," /><T i="223" o="222" p="nn" v="noun,,plur,voct,,,," /><T i="224" o="182" p="nn" v="noun,neut,sing,gent2,,,," /><T i="225" o="0" p="nn" v="advb,neut,sing,,,,," /><T i="226" o="112" p="nn" v="verb,femn,sing,,,,impr," /><T i="227" o="183" p="nn" v="noun,neut,sing,gent1,,,," /><T i="228" o="181" p="nn" v="noun,neut,sing,loct1,,,," /><T i="230" l="1" o="275" p="numb" v="numb,masc,sing,nomn,,,," /><T i="231" l="1" o="257" p="numb" v="numb,femn,sing,nomn,,,," /><T i="232" l="1" o="239" p="numb" v="numb,neut,sing,nomn,,,," /><T i="233" o="287" p="numb" v="numb,,plur,nomn,,,," /><T i="234" o="274" p="numb" v="numb,masc,sing,gent,,,," /><T i="235" o="256" p="numb" v="numb,femn,sing,gent,,,," /><T i="236" o="238" p="numb" v="numb,neut,sing,gent,,,," /><T i="237" o="286" p="numb" v="numb,,plur,gent,,,," /><T i="238" o="273" p="numb" v="numb,masc,sing,datv,,,," /><T i="239" o="255" p="numb" v="numb,femn,sing,datv,,,," /><T i="240" o="237" p="numb" v="numb,neut,sing,datv,,,," /><T i="241" o="285" p="numb" v="numb,,plur,datv,,,," /><T i="242" o="272" p="numb" v="numb,masc,sing,accs,,,," /><T i="243" o="254" p="numb" v="numb,femn,sing,accs,,,," /><T i="244" o="236" p="numb" v="numb,neut,sing,accs,,,," /><T i="245" o="284" p="numb" v="numb,,plur,accs,,,," /><T i="246" o="271" p="numb" v="numb,masc,sing,ablt,,,," /><T i="247" o="253" p="numb" v="numb,femn,sing,ablt,,,," /><T i="248" o="235" p="numb" v="numb,neut,sing,ablt,,,," /><T i="249" o="283" p="numb" v="numb,,plur,ablt,,,," /><T i="250" o="270" p="numb" v="numb,masc,sing,loct,,,," /><T i="251" o="252" p="numb" v="numb,femn,sing,loct,,,," /><T i="252" o="234" p="numb" v="numb,neut,sing,loct,,,," /><T i="253" o="282" p="numb" v="numb,,plur,loct,,,," /><T i="254" o="269" p="numb" v="numb,masc,plur,nomn,,,," /><T i="255" o="268" p="numb" v="numb,masc,plur,gent,,,," /><T i="256" o="267" p="numb" v="numb,masc,plur,datv,,,," /><T i="257" o="266" p="numb" v="numb,masc,plur,accs,,,," /><T i="258" o="265" p="numb" v="numb,masc,plur,ablt,,,," /><T i="259" o="264" p="numb" v="numb,masc,plur,loct,,,," /><T i="260" o="263" p="numb" v="numb,femn,,nomn,,,," /><T i="261" l="1" o="293" p="numb" v="numb,,,nomn,,,," /><T i="262" o="245" p="numb" v="numb,neut,,nomn,,,," /><T i="263" o="292" p="numb" v="numb,,,gent,,,," /><T i="264" o="262" p="numb" v="numb,femn,,gent,,,," /><T i="265" o="244" p="numb" v="numb,neut,,gent,,,," /><T i="266" o="291" p="numb" v="numb,,,datv,,,," /><T i="267" o="261" p="numb" v="numb,femn,,datv,,,," /><T i="268" o="243" p="numb" v="numb,neut,,datv,,,," /><T i="269" o="290" p="numb" v="numb,,,accs,,,," /><T i="270" o="278" p="numb" v="numb,masc,,accs,,,," /><T i="271" o="260" p="numb" v="numb,femn,,accs,,,," /><T i="272" o="242" p="numb" v="numb,neut,,accs,,,," /><T i="273" o="289" p="numb" v="numb,,,ablt,,,," /><T i="274" o="277" p="numb" v="numb,masc,,ablt,,,," /><T i="275" o="259" p="numb" v="numb,femn,,ablt,,,," /><T i="276" o="241" p="numb" v="numb,neut,,ablt,,,," /><T i="277" o="258" p="numb" v="numb,femn,,loct,,,," /><T i="278" o="240" p="numb" v="numb,neut,,loct,,,," /><T i="279" l="1" o="281" p="numb" v="numb,masc,,nomn,,,," /><T i="280" o="280" p="numb" v="numb,masc,,gent,,,," /><T i="281" o="279" p="numb" v="numb,masc,,datv,,,," /><T i="282" o="276" p="numb" v="numb,masc,,loct,,,," /><T i="283" o="288" p="numb" v="numb,,,loct,,,," /><T i="284" o="251" p="numb" v="numb,femn,plur,nomn,,,," /><T i="285" o="250" p="numb" v="numb,femn,plur,gent,,,," /><T i="286" o="249" p="numb" v="numb,femn,plur,datv,,,," /><T i="287" o="248" p="numb" v="numb,femn,plur,accs,,,," /><T i="288" o="247" p="numb" v="numb,femn,plur,ablt,,,," /><T i="289" o="246" p="numb" v="numb,femn,plur,loct,,,," /><T i="290" l="1" o="232" p="reg" v="punct,,,,,,," /><T i="291" l="1" o="231" p="reg" v="int,,,,,,," /><T i="292" l="1" o="230" p="reg" v="romn,,,,,,," /><T i="294" l="1" o="229" p="reg" v="unkn,,,,,,," /><T i="295" l="1" o="374" p="dict" v="npro,,sing,nomn,1per,,," /><T i="296" l="1" o="294" p="dict" v="intj,,,,,,," /><T i="297" l="1" o="296" p="dict" v="conj,,,,,,," /><T i="298" o="325" p="dict" v="npro,masc,sing,ablt,,,," /><T i="299" l="1" o="309" p="dict" v="npro,neut,sing,nomn,,,," /><T i="300" l="1" o="321" p="dict" v="npro,femn,sing,nomn,,,," /><T i="301" l="1" o="297" p="dict" v="prep,,,,,,," /><T i="302" l="1" o="295" p="dict" v="prcl,,,,,,," /><T i="303" o="299" p="dict" v="npro,neut,sing,loct,,,," /><T i="304" o="329" p="dict" v="npro,masc,sing,datv,,,," /><T i="305" o="327" p="dict" v="npro,masc,sing,accs,,,," /><T i="306" l="1" o="233" p="dict" v="pred,,,,,pres,," /><T i="307" o="303" p="dict" v="npro,neut,sing,accs,,,," /><T i="308" l="1" o="333" p="dict" v="npro,masc,sing,nomn,,,," /><T i="309" o="368" p="dict" v="npro,,sing,datv,1per,,," /><T i="310" o="372" p="dict" v="npro,,sing,gent,,,," /><T i="311" o="328" p="dict" v="npro,masc,sing,datv,3per,,," /><T i="312" o="312" p="dict" v="npro,femn,sing,ablt,3per,,," /><T i="313" o="371" p="dict" v="npro,,sing,gent,1per,,," /><T i="314" o="361" p="dict" v="npro,,sing,ablt,2per,,," /><T i="315" o="301" p="dict" v="npro,neut,sing,ablt,,,," /><T i="316" o="311" p="dict" v="npro,femn,sing,loct,,,," /><T i="317" o="323" p="dict" v="npro,masc,sing,loct,,,," /><T i="318" o="315" p="dict" v="npro,femn,sing,accs,,,," /><T i="319" o="363" p="dict" v="npro,,sing,ablt,,,," /><T i="320" o="313" p="dict" v="npro,femn,sing,ablt,,,," /><T i="321" o="342" p="dict" v="npro,,plur,accs,3per,,," /><T i="322" o="339" p="dict" v="npro,,plur,ablt,2per,,," /><T i="323" l="1" o="356" p="dict" v="npro,,plur,nomn,1per,,," /><T i="324" o="305" p="dict" v="npro,neut,sing,datv,,,," /><T i="325" l="1" o="320" p="dict" v="npro,femn,sing,nomn,3per,,," /><T i="326" o="347" p="dict" v="npro,,plur,datv,2per,,," /><T i="327" l="1" o="355" p="dict" v="npro,,plur,nomn,2per,,," /><T i="328" o="318" p="dict" v="npro,femn,sing,gent,3per,,," /><T i="329" o="351" p="dict" v="npro,,plur,gent,2per,,," /><T i="330" o="369" p="dict" v="npro,,sing,datv,,,," /><T i="331" o="344" p="dict" v="npro,,plur,accs,1per,,," /><T i="332" o="307" p="dict" v="npro,neut,sing,gent,,,," /><T i="333" o="348" p="dict" v="npro,,plur,datv,1per,,," /><T i="334" o="304" p="dict" v="npro,neut,sing,datv,3per,,," /><T i="335" o="349" p="dict" v="npro,,plur,datv,,,," /><T i="336" o="300" p="dict" v="npro,neut,sing,ablt,3per,,," /><T i="337" o="326" p="dict" v="npro,masc,sing,accs,3per,,," /><T i="338" o="314" p="dict" v="npro,femn,sing,accs,3per,,," /><T i="339" o="324" p="dict" v="npro,masc,sing,ablt,3per,,," /><T i="340" o="340" p="dict" v="npro,,plur,ablt,1per,,," /><T i="341" o="317" p="dict" v="npro,femn,sing,datv,,,," /><T i="342" o="310" p="dict" v="npro,femn,sing,loct,3per,,," /><T i="343" l="1" o="357" p="dict" v="npro,,plur,nomn,,,," /><T i="344" o="367" p="dict" v="npro,,sing,datv,2per,,," /><T i="345" o="337" p="dict" v="npro,,plur,loct,,,," /><T i="346" l="1" o="373" p="dict" v="npro,,sing,nomn,2per,,," /><T i="347" o="370" p="dict" v="npro,,sing,gent,2per,,," /><T i="348" o="341" p="dict" v="npro,,plur,ablt,,,," /><T i="349" o="352" p="dict" v="npro,,plur,gent,1per,,," /><T i="350" o="362" p="dict" v="npro,,sing,ablt,1per,,," /><T i="351" o="306" p="dict" v="npro,neut,sing,gent,3per,,," /><T i="352" o="331" p="dict" v="npro,masc,sing,gent,,,," /><T i="353" o="353" p="dict" v="npro,,plur,gent,,,," /><T i="354" o="338" p="dict" v="npro,,plur,ablt,3per,,," /><T i="355" o="360" p="dict" v="npro,,sing,loct,,,," /><T i="356" l="1" o="332" p="dict" v="npro,masc,sing,nomn,3per,,," /><T i="357" o="345" p="dict" v="npro,,plur,accs,,,," /><T i="358" o="366" p="dict" v="npro,,sing,accs,,,," /><T i="359" o="302" p="dict" v="npro,neut,sing,accs,3per,,," /><T i="360" o="365" p="dict" v="npro,,sing,accs,1per,,," /><T i="361" l="1" o="354" p="dict" v="npro,,plur,nomn,3per,,," /><T i="362" l="1" o="308" p="dict" v="npro,neut,sing,nomn,3per,,," /><T i="363" o="335" p="dict" v="npro,,plur,loct,2per,,," /><T i="364" o="334" p="dict" v="npro,,plur,loct,3per,,," /><T i="365" o="319" p="dict" v="npro,femn,sing,gent,,,," /><T i="366" o="330" p="dict" v="npro,masc,sing,gent,3per,,," /><T i="367" o="336" p="dict" v="npro,,plur,loct,1per,,," /><T i="368" o="346" p="dict" v="npro,,plur,datv,3per,,," /><T i="369" o="358" p="dict" v="npro,,sing,loct,2per,,," /><T i="370" o="350" p="dict" v="npro,,plur,gent,3per,,," /><T i="371" o="298" p="dict" v="npro,neut,sing,loct,3per,,," /><T i="372" o="322" p="dict" v="npro,masc,sing,loct,3per,,," /><T i="373" o="359" p="dict" v="npro,,sing,loct,1per,,," /><T i="374" l="1" o="375" p="dict" v="npro,,sing,nomn,,,," /><T i="375" o="316" p="dict" v="npro,femn,sing,datv,3per,,," /><T i="376" o="364" p="dict" v="npro,,sing,accs,2per,,," /><T i="377" o="343" p="dict" v="npro,,plur,accs,2per,,," /></Tags>

latest_release/test_info.txt

@@ -0,0 +1,11 @@
+post. full_cls_acc: 0.987678530574893; part_cls_acc: 0.9878190630048466
+gndr. full_cls_acc: 0.9604920851889414; part_cls_acc: 0.9672658190851966
+nmbr. full_cls_acc: 0.9931398234969664; part_cls_acc: 0.9935646606086085
+case. full_cls_acc: 0.9806930486635252; part_cls_acc: 0.9876671968294527
+pers. full_cls_acc: 1.0; part_cls_acc: 1.0
+tens. full_cls_acc: 1.0; part_cls_acc: 1.0
+mood. full_cls_acc: 0.9986893840104849; part_cls_acc: 0.9987577639751553
+voic. full_cls_acc: 1.0; part_cls_acc: 1.0
+main. full_cls_acc: 0.9551451187335093; part_cls_acc: 0.9647049460825275
+lemma_acc: 0.9955360766171577
+inflect_acc: 0.9919720286154625

model.py

@@ -0,0 +1,264 @@
+import os
+import shutil
+import tensorflow as tf
+from graph.gram_cls import GramCls
+from graph.main_cls import MainCls
+from graph.lemm import Lemm
+from graph.inflect import Inflect
+from graph.base import TfContext
+from utils import MyDefaultDict, CONFIG
+from tensorflow.python.tools import freeze_graph
+
+
+class RNN:
+    def __init__(self, for_usage):
+        self.config = CONFIG
+        self.filler = self.config['filler']
+        self.checkpoints_keep = 200000
+        self.for_usage = for_usage
+        self.default_config = self.config['graph_part_configs']['default']
+        self.key_configs = MyDefaultDict(
+            lambda key: self.default_config,
+            {
+                key: MyDefaultDict(lambda prop_key: self.default_config[prop_key], self.config['graph_part_configs'][key])
+                for key in self.config['graph_part_configs']
+                if key != 'default'
+            }
+        )
+        self.export_path = self.config['export_path']
+        self.save_path = self.config['save_path']
+        self.model_key = self.config['model_key']
+        self.miss_steps = self.config['miss_steps'] if 'miss_steps' in self.config else []
+        self.start_char = self.config['start_token']
+        self.end_char = self.config['end_token']
+        self.gram_keys = [
+            key
+            for key in sorted(self.config['grammemes_types'], key=lambda x: self.config['grammemes_types'][x]['index'])
+        ]
+        self.main_class_k = self.config['main_class_k']
+        self.train_steps = self.config['train_steps']
+
+        if for_usage:
+            self.devices = ['/cpu:0']
+        else:
+            self.devices = self.config['train_devices']
+
+        if not os.path.exists(self.save_path):
+            os.mkdir(self.save_path)
+
+        self._build_graph()
+
+    def _build_graph(self):
+        self.graph = tf.Graph()
+        self.checks = []
+        self.xs = []
+        self.x_seq_lens = []
+
+        self.x_inds = []
+        self.x_vals = []
+        self.x_shape = []
+
+        self.prints = []
+
+        with self.graph.as_default(), tf.device('/cpu:0'):
+            self.is_training = tf.placeholder(tf.bool, name="IsTraining")
+            self.learn_rate = tf.placeholder(tf.float32, name="LearningRate")
+            self.batch_size = tf.placeholder(tf.int32, [], name="BatchSize") if self.for_usage else None
+            self.optimiser = tf.train.AdamOptimizer(self.learn_rate)
+            self.reset_optimizer = tf.variables_initializer(self.optimiser.variables())
+            self.gram_graph_parts = {
+                gram: GramCls(gram, self.for_usage, self.config, self.key_configs[gram], self.optimiser, self.reset_optimizer)
+                for gram in self.gram_keys
+            }
+            self.lem_graph_part = Lemm(self.for_usage, self.config, self.key_configs["lemm"], self.optimiser, self.reset_optimizer)
+            self.main_graph_part = MainCls(self.for_usage, self.config, self.key_configs["main"], self.optimiser, self.reset_optimizer)
+            self.inflect_graph_part = Inflect(self.for_usage, self.config, self.key_configs['inflect'], self.optimiser, self.reset_optimizer)
+
+            for device_index, device_name in enumerate(self.devices):
+                with tf.device(device_name):
+                    if self.for_usage:
+                        x_ind_pl = tf.placeholder(dtype=tf.int32, shape=(None, None), name='XIndexes')
+                        x_val_pl = tf.placeholder(dtype=tf.int32, shape=(None,), name='XValues')
+                        x_shape_pl = tf.placeholder(dtype=tf.int32, shape=(2,), name='XShape')
+                        x_ind = tf.dtypes.cast(x_ind_pl, dtype=tf.int64)
+                        x_val = tf.dtypes.cast(x_val_pl, dtype=tf.int64)
+                        x_shape = tf.dtypes.cast(x_shape_pl, dtype=tf.int64)
+
+                        x_sparse = tf.sparse.SparseTensor(x_ind, x_val, x_shape)
+                        x = tf.sparse.to_dense(x_sparse, default_value=self.end_char)
+                        self.x_inds.append(x_ind_pl)
+                        self.x_vals.append(x_val_pl)
+                        self.x_shape.append(x_shape_pl)
+                    else:
+                        x = tf.placeholder(dtype=tf.int32, shape=(None, None), name='X')
+                        self.xs.append(x)
+
+                    x_seq_len = tf.placeholder(dtype=tf.int32, shape=(None,), name='SeqLen')
+                    self.x_seq_lens.append(x_seq_len)
+
+                    for gram in self.gram_keys:
+                        self.gram_graph_parts[gram].build_graph_for_device(x, x_seq_len)
+
+                    gram_probs = [self.gram_graph_parts[gram].probs[-1] for gram in self.gram_keys]
+                    gram_keep_drops = [self.gram_graph_parts[gram].keep_drops[-1] for gram in self.gram_keys]
+                    self.main_graph_part.build_graph_for_device(x, x_seq_len, gram_probs, gram_keep_drops)
+                    self.prints.append(tf.print("main_result", self.main_graph_part.results[0].indices))
+                    if self.for_usage:
+                        x_tiled = tf.contrib.seq2seq.tile_batch(x, multiplier=self.main_class_k)
+                        seq_len_tiled = tf.contrib.seq2seq.tile_batch(x_seq_len, multiplier=self.main_class_k)
+                        cls = tf.reshape(self.main_graph_part.results[0].indices, (-1,))
+                        batch_size_tiled = self.batch_size * self.main_class_k
+                        self.lem_graph_part.build_graph_for_device(x_tiled,
+                                                                   seq_len_tiled,
+                                                                   batch_size_tiled,
+                                                                   cls)
+                        self.lem_cls_result = tf.reshape(self.lem_graph_part.results[0],
+                                                         (self.batch_size, self.main_class_k, -1))
+                        self.lem_graph_part.build_graph_for_device(x,
+                                                                   x_seq_len,
+                                                                   self.batch_size)
+                        self.lem_result = tf.expand_dims(self.lem_graph_part.results[1], 1)
+                        self.lem_class_pl = self.lem_graph_part.cls[0]
+                    else:
+                        self.lem_graph_part.build_graph_for_device(x,
+                                                                   x_seq_len,
+                                                                   self.batch_size)
+
+                    self.inflect_graph_part.build_graph_for_device(x, x_seq_len, self.batch_size)
+                    if self.for_usage:
+                        self.inflect_result = self.inflect_graph_part.results[0]
+                        self.inflect_x_class_pl = self.inflect_graph_part.x_cls[0]
+                        self.inflect_y_class_pl = self.inflect_graph_part.y_cls[0]
+
+            for gram in self.gram_keys:
+                self.gram_graph_parts[gram].build_graph_end()
+            self.main_graph_part.build_graph_end()
+            self.lem_graph_part.build_graph_end()
+            self.inflect_graph_part.build_graph_end()
+            self.saver = tf.train.Saver(tf.global_variables(), max_to_keep=self.checkpoints_keep)
+
+    def restore(self, sess):
+        latest_checkpoint = tf.train.latest_checkpoint(self.save_path)
+        for gram in self.gram_keys:
+            if gram not in self.config['ignore_restore']:
+                self.gram_graph_parts[gram].restore(sess, latest_checkpoint)
+        if self.main_graph_part.key not in self.config['ignore_restore']:
+            self.main_graph_part.restore(sess, latest_checkpoint)
+        if self.lem_graph_part.key not in self.config['ignore_restore']:
+            self.lem_graph_part.restore(sess, latest_checkpoint)
+        if self.inflect_graph_part not in self.config['ignore_restore']:
+            self.inflect_graph_part.restore(sess, latest_checkpoint)
+
+        if self.inflect_graph_part.settings['transfer_init']:
+            self.inflect_graph_part.transfer_learning_init(sess)
+
+    def train(self):
+        config = tf.ConfigProto(allow_soft_placement=True)
+        if not os.path.isdir(self.save_path):
+            os.mkdir(self.save_path)
+
+        with tf.Session(config = config, graph=self.graph) as sess:
+            sess.run(tf.global_variables_initializer())
+            sess.run(tf.local_variables_initializer())
+
+            tc = TfContext(sess, self.saver, self.learn_rate)
+            self.restore(sess)
+            sess.run(self.reset_optimizer)
+
+            for gram in self.gram_keys:
+                if gram in self.train_steps:
+                    self.gram_graph_parts[gram].train(tc)
+
+            if self.main_graph_part.key in self.train_steps:
+                self.main_graph_part.train(tc)
+
+            if self.lem_graph_part.key in self.train_steps:
+                self.lem_graph_part.train(tc)
+
+            if self.inflect_graph_part.key in self.train_steps:
+                self.inflect_graph_part.train(tc)
+
+    def release(self):
+        with tf.Session(graph=self.graph) as sess:
+            sess.run(tf.global_variables_initializer())
+            sess.run(tf.local_variables_initializer())
+
+            # Loading checkpoint
+            latest_checkpoint = tf.train.latest_checkpoint(self.save_path)
+            if latest_checkpoint:
+                self.restore(sess)
+
+            if os.path.isdir(self.export_path):
+                shutil.rmtree(self.export_path)
+
+            output_dic = {}
+            gram_op_dic = {}
+            for gram in self.gram_keys:
+                res = self.gram_graph_parts[gram].results[0]
+                prob = self.gram_graph_parts[gram].probs[0]
+                output_dic[f'res_{gram}'] = res
+                output_dic[f'prob_{gram}'] = prob
+                gram_op_dic[gram] = {
+                    'res': res.op.name,
+                    'prob': prob.op.name
+                }
+
+            output_dic['res_values'] = self.main_graph_part.results[0].values
+            output_dic['res_indexes'] = self.main_graph_part.results[0].indices
+
+            output_dic['lem_cls_result'] = self.lem_cls_result
+            output_dic['lem_result'] = self.lem_result
+
+            output_dic['inflect_result'] = self.inflect_graph_part.results[0]
+
+            # Saving model
+            tf.saved_model.simple_save(sess,
+                                       self.export_path,
+                                       inputs={
+                                           'x_ind': self.x_inds[0],
+                                           'x_val': self.x_vals[0],
+                                           'x_shape': self.x_shape[0],
+                                           'seq_len': self.x_seq_lens[0],
+                                           'batch_size': self.batch_size,
+                                           'lem_x_cls': self.lem_class_pl,
+                                           'inflect_x_cls': self.inflect_x_class_pl,
+                                           'inflect_y_cls': self.inflect_y_class_pl
+                                       },
+                                       outputs=output_dic)
+
+            # Freezing graph
+            input_graph = 'graph.pbtxt'
+            tf.train.write_graph(sess.graph.as_graph_def(), self.export_path, input_graph, as_text=True)
+            input_graph = os.path.join(self.export_path, input_graph)
+            frozen_path = os.path.join(self.export_path, 'frozen_model.pb')
+            output_ops = [output_dic[key].op.name for key in output_dic]
+            output_ops = ",".join(output_ops)
+            freeze_graph.freeze_graph(input_graph,
+                                      "",
+                                      False,
+                                      latest_checkpoint,
+                                      output_ops,
+                                      "",
+                                      "",
+                                      frozen_path,
+                                      True,
+                                      "",
+                                      input_saved_model_dir=self.export_path)
+
+            op_dic = {
+                key: output_dic[key].op.name
+                for key in output_dic
+            }
+
+            op_dic['x_ind'] = self.x_inds[0].op.name
+            op_dic['x_val'] = self.x_vals[0].op.name
+            op_dic['x_shape'] = self.x_shape[0].op.name
+            op_dic['seq_len'] = self.main_graph_part.x_seq_lens[0].op.name
+            op_dic['batch_size'] = self.batch_size.op.name
+            op_dic['lem_x_cls'] = self.lem_class_pl.op.name
+            op_dic['inflect_x_cls'] = self.inflect_x_class_pl.op.name
+            op_dic['inflect_y_cls'] = self.inflect_y_class_pl.op.name
+
+            return frozen_path, \
+                   gram_op_dic , \
+                   op_dic

numb.yml

@@ -0,0 +1,1532 @@
+0:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: нулевой}
+  - {case: nomn, gndr: femn, nmbr: sing, text: нулевая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: нулевое}
+  - {case: nomn, nmbr: plur, text: нулевые}
+  - {case: gent, gndr: masc, nmbr: sing, text: нулевого}
+  - {case: gent, gndr: femn, nmbr: sing, text: нулевой}
+  - {case: gent, gndr: neut, nmbr: sing, text: нулевого}
+  - {case: gent, nmbr: plur, text: нулевых}
+  - {case: datv, gndr: masc, nmbr: sing, text: нулевому}
+  - {case: datv, gndr: femn, nmbr: sing, text: нулевой}
+  - {case: datv, gndr: neut, nmbr: sing, text: нулевому}
+  - {case: datv, nmbr: plur, text: нулевым}
+  - {case: accs, gndr: masc, nmbr: sing, text: нулевого}
+  - {case: accs, gndr: masc, nmbr: sing, text: нулевой}
+  - {case: accs, gndr: femn, nmbr: sing, text: нулевую}
+  - {case: accs, gndr: neut, nmbr: sing, text: нулевое}
+  - {case: accs, nmbr: plur, text: нулевых}
+  - {case: accs, nmbr: plur, text: нулевые}
+  - {case: ablt, gndr: masc, nmbr: sing, text: нулевым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: нулевой}
+  - {case: ablt, gndr: femn, nmbr: sing, text: нулевою}
+  - {case: ablt, gndr: neut, nmbr: sing, text: нулевым}
+  - {case: ablt, nmbr: plur, text: нулевыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: нулевом}
+  - {case: loct, gndr: femn, nmbr: sing, text: нулевой}
+  - {case: loct, gndr: neut, nmbr: sing, text: нулевом}
+  - {case: loct, nmbr: plur, text: нулевых}
+  o:
+  - {case: nomn, gndr: masc, nmbr: sing, text: ноль}
+  - {case: nomn, gndr: masc, nmbr: plur, text: ноли}
+  - {case: gent, gndr: masc, nmbr: sing, text: ноля}
+  - {case: gent, gndr: masc, nmbr: plur, text: нолей}
+  - {case: datv, gndr: masc, nmbr: sing, text: нолю}
+  - {case: datv, gndr: masc, nmbr: plur, text: нолям}
+  - {case: accs, gndr: masc, nmbr: sing, text: ноль}
+  - {case: accs, gndr: masc, nmbr: plur, text: ноли}
+  - {case: ablt, gndr: masc, nmbr: sing, text: нолём}
+  - {case: ablt, gndr: masc, nmbr: plur, text: нолями}
+  - {case: loct, gndr: masc, nmbr: sing, text: ноле}
+  - {case: loct, gndr: masc, nmbr: plur, text: нолях}
+  o1:
+  - {case: nomn, gndr: masc, nmbr: sing, text: нуль}
+  - {case: nomn, gndr: masc, nmbr: plur, text: нули}
+  - {case: gent, gndr: masc, nmbr: sing, text: нуля}
+  - {case: gent, gndr: masc, nmbr: plur, text: нулей}
+  - {case: datv, gndr: masc, nmbr: sing, text: нулю}
+  - {case: datv, gndr: masc, nmbr: plur, text: нулям}
+  - {case: accs, gndr: masc, nmbr: sing, text: нуль}
+  - {case: accs, gndr: masc, nmbr: plur, text: нули}
+  - {case: ablt, gndr: masc, nmbr: sing, text: нулём}
+  - {case: ablt, gndr: masc, nmbr: plur, text: нулями}
+  - {case: loct, gndr: masc, nmbr: sing, text: нуле}
+  - {case: loct, gndr: masc, nmbr: plur, text: нулях}
+1:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: первый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: первая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: первое}
+  - {case: nomn, nmbr: plur, text: первые}
+  - {case: gent, gndr: masc, nmbr: sing, text: первого}
+  - {case: gent, gndr: femn, nmbr: sing, text: первой}
+  - {case: gent, gndr: neut, nmbr: sing, text: первого}
+  - {case: gent, nmbr: plur, text: первых}
+  - {case: datv, gndr: masc, nmbr: sing, text: первому}
+  - {case: datv, gndr: femn, nmbr: sing, text: первой}
+  - {case: datv, gndr: neut, nmbr: sing, text: первому}
+  - {case: datv, nmbr: plur, text: первым}
+  - {case: accs, gndr: masc, nmbr: sing, text: первого}
+  - {case: accs, gndr: masc, nmbr: sing, text: первый}
+  - {case: accs, gndr: femn, nmbr: sing, text: первую}
+  - {case: accs, gndr: neut, nmbr: sing, text: первое}
+  - {case: accs, nmbr: plur, text: первые}
+  - {case: accs, nmbr: plur, text: первых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: первым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: первой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: первым}
+  - {case: ablt, nmbr: plur, text: первыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: первом}
+  - {case: loct, gndr: femn, nmbr: sing, text: первой}
+  - {case: loct, gndr: neut, nmbr: sing, text: первом}
+  - {case: loct, nmbr: plur, text: первых}
+  o:
+  - {case: nomn, gndr: masc, nmbr: sing, text: один}
+  - {case: nomn, gndr: femn, nmbr: sing, text: одна}
+  - {case: nomn, gndr: neut, nmbr: sing, text: одно}
+  - {case: nomn, nmbr: plur, text: одни}
+  - {case: gent, gndr: masc, nmbr: sing, text: одного}
+  - {case: gent, gndr: femn, nmbr: sing, text: одной}
+  - {case: gent, gndr: neut, nmbr: sing, text: одного}
+  - {case: gent, nmbr: plur, text: одних}
+  - {case: datv, gndr: masc, nmbr: sing, text: одному}
+  - {case: datv, gndr: femn, nmbr: sing, text: одной}
+  - {case: datv, gndr: neut, nmbr: sing, text: одному}
+  - {case: datv, nmbr: plur, text: одним}
+  - {case: accs, gndr: femn, nmbr: sing, text: одного}
+  - {case: accs, gndr: masc, nmbr: sing, text: один}
+  - {case: accs, gndr: femn, nmbr: sing, text: одну}
+  - {case: accs, gndr: neut, nmbr: sing, text: одно}
+  - {case: accs, nmbr: plur, text: одни}
+  - {case: accs, nmbr: plur, text: одних}
+  - {case: ablt, gndr: masc, nmbr: sing, text: одним}
+  - {case: ablt, gndr: femn, nmbr: sing, text: одной}
+  - {case: ablt, gndr: neut, nmbr: sing, text: одним}
+  - {case: ablt, nmbr: plur, text: одними}
+  - {case: loct, gndr: masc, nmbr: sing, text: одном}
+  - {case: loct, gndr: femn, nmbr: sing, text: одной}
+  - {case: loct, gndr: neut, nmbr: sing, text: одном}
+  - {case: loct, nmbr: plur, text: одних}
+  - {case: nomn, gndr: femn, text: один}
+  - {case: nomn, text: одни}
+  - {case: nomn, gndr: femn, text: одна}
+  - {case: nomn, gndr: neut, text: одно}
+  - {case: gent, text: одних}
+  - {case: gent, gndr: femn, text: одного}
+  - {case: gent, gndr: femn, text: одной}
+  - {case: gent, gndr: neut, text: одного}
+  - {case: datv, text: одним}
+  - {case: datv, gndr: femn, text: одному}
+  - {case: datv, gndr: femn, text: одной}
+  - {case: datv, gndr: neut, text: одному}
+  - {case: accs, text: одни}
+  - {case: accs, text: одних}
+  - {case: accs, gndr: masc, text: один}
+  - {case: accs, gndr: masc, text: одного}
+  - {case: accs, gndr: femn, text: одну}
+  - {case: accs, gndr: neut, text: одно}
+  - {case: ablt, text: одними}
+  - {case: ablt, gndr: masc, text: одним}
+  - {case: ablt, gndr: femn, text: одной}
+  - {case: ablt, gndr: femn, text: одною}
+  - {case: ablt, gndr: neut, text: одним}
+  - {case: loct, gndr: femn, text: одном}
+  - {case: loct, gndr: femn, text: одной}
+  - {case: loct, gndr: neut, text: одном}
+2:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: второй}
+  - {case: nomn, gndr: femn, nmbr: sing, text: вторая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: второе}
+  - {case: nomn, nmbr: plur, text: вторые}
+  - {case: gent, gndr: masc, nmbr: sing, text: второго}
+  - {case: gent, gndr: femn, nmbr: sing, text: второй}
+  - {case: gent, gndr: neut, nmbr: sing, text: второго}
+  - {case: gent, nmbr: plur, text: вторых}
+  - {case: datv, gndr: masc, nmbr: sing, text: второму}
+  - {case: datv, gndr: femn, nmbr: sing, text: второй}
+  - {case: datv, gndr: neut, nmbr: sing, text: второму}
+  - {case: datv, nmbr: plur, text: вторым}
+  - {case: accs, gndr: femn, nmbr: sing, text: второго}
+  - {case: accs, gndr: masc, nmbr: sing, text: второй}
+  - {case: accs, gndr: femn, nmbr: sing, text: вторую}
+  - {case: accs, gndr: neut, nmbr: sing, text: второе}
+  - {case: accs, nmbr: plur, text: вторые}
+  - {case: accs, nmbr: plur, text: вторых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: вторым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: второй}
+  - {case: ablt, gndr: neut, nmbr: sing, text: вторым}
+  - {case: ablt, nmbr: plur, text: вторыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: втором}
+  - {case: loct, gndr: femn, nmbr: sing, text: второй}
+  - {case: loct, gndr: neut, nmbr: sing, text: втором}
+  - {case: loct, nmbr: plur, text: вторых}
+  o:
+  - {case: nomn, gndr: masc, text: два}
+  - {case: nomn, gndr: femn, text: две}
+  - {case: nomn, gndr: neut, text: два}
+  - {case: gent, gndr: masc, text: двух}
+  - {case: gent, gndr: femn, text: двух}
+  - {case: gent, gndr: neut, text: двух}
+  - {case: datv, gndr: masc, text: двум}
+  - {case: datv, gndr: femn, text: двум}
+  - {case: datv, gndr: neut, text: двум}
+  - {case: accs, gndr: masc, text: два}
+  - {case: accs, gndr: masc, text: двух}
+  - {case: accs, gndr: femn, text: две}
+  - {case: accs, gndr: femn, text: двух}
+  - {case: accs, gndr: neut, text: два}
+  - {case: ablt, gndr: masc, text: двумя}
+  - {case: ablt, gndr: femn, text: двумя}
+  - {case: ablt, gndr: neut, text: двумя}
+  - {case: loct, gndr: masc, text: двух}
+  - {case: loct, gndr: femn, text: двух}
+  - {case: loct, gndr: neut, text: двух}
+3:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: третий}
+  - {case: nomn, gndr: femn, nmbr: sing, text: третья}
+  - {case: nomn, gndr: neut, nmbr: sing, text: третье}
+  - {case: nomn, nmbr: plur, text: третьи}
+  - {case: gent, gndr: masc, nmbr: sing, text: третьего}
+  - {case: gent, gndr: femn, nmbr: sing, text: третьей}
+  - {case: gent, gndr: neut, nmbr: sing, text: третьего}
+  - {case: gent, nmbr: plur, text: третьих}
+  - {case: datv, gndr: masc, nmbr: sing, text: третьему}
+  - {case: datv, gndr: femn, nmbr: sing, text: третьей}
+  - {case: datv, gndr: neut, nmbr: sing, text: третьему}
+  - {case: datv, nmbr: plur, text: третьим}
+  - {case: accs, gndr: masc, nmbr: sing, text: третьего}
+  - {case: accs, gndr: masc, nmbr: sing, text: третий}
+  - {case: accs, gndr: femn, nmbr: sing, text: третью}
+  - {case: accs, gndr: neut, nmbr: sing, text: третье}
+  - {case: accs, nmbr: plur, text: третьи}
+  - {case: accs, nmbr: plur, text: третьих}
+  - {case: ablt, gndr: masc, nmbr: sing, text: третьим}
+  - {case: ablt, gndr: femn, nmbr: sing, text: третьей}
+  - {case: ablt, gndr: femn, nmbr: sing, text: третьею}
+  - {case: ablt, gndr: neut, nmbr: sing, text: третьим}
+  - {case: ablt, nmbr: plur, text: третьими}
+  - {case: loct, gndr: masc, nmbr: sing, text: третьем}
+  - {case: loct, gndr: femn, nmbr: sing, text: третьей}
+  - {case: loct, gndr: neut, nmbr: sing, text: третьем}
+  - {case: loct, nmbr: plur, text: третьих}
+  o:
+  - {case: nomn, text: три}
+  - {case: gent, text: трех}
+  - {case: datv, text: трем}
+  - {case: accs, text: трех}
+  - {case: accs, text: три}
+  - {case: ablt, text: тремя}
+  - {case: loct, text: трех}
+4:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: четвёртый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: четвёртая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: четвёртое}
+  - {case: nomn, nmbr: plur, text: четвёртые}
+  - {case: gent, gndr: masc, nmbr: sing, text: четвёртого}
+  - {case: gent, gndr: femn, nmbr: sing, text: четвёртой}
+  - {case: gent, gndr: neut, nmbr: sing, text: четвёртого}
+  - {case: gent, nmbr: plur, text: четвёртых}
+  - {case: datv, gndr: masc, nmbr: sing, text: четвёртому}
+  - {case: datv, gndr: femn, nmbr: sing, text: четвёртой}
+  - {case: datv, gndr: neut, nmbr: sing, text: четвёртому}
+  - {case: datv, nmbr: plur, text: четвёртым}
+  - {case: accs, gndr: masc, nmbr: sing, text: четвёртого}
+  - {case: accs, gndr: masc, nmbr: sing, text: четвёртый}
+  - {case: accs, gndr: femn, nmbr: sing, text: четвёртую}
+  - {case: accs, gndr: neut, nmbr: sing, text: четвёртое}
+  - {case: accs, nmbr: plur, text: четвёртые}
+  - {case: accs, nmbr: plur, text: четвёртых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: четвёртым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: четвёртой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: четвёртым}
+  - {case: ablt, nmbr: plur, text: четвёртыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: четвёртом}
+  - {case: loct, gndr: femn, nmbr: sing, text: четвёртой}
+  - {case: loct, gndr: neut, nmbr: sing, text: четвёртом}
+  - {case: loct, nmbr: plur, text: четвёртых}
+  o:
+  - {case: nomn, text: четыре}
+  - {case: gent, text: четырех}
+  - {case: datv, text: четырем}
+  - {case: accs, text: четыре}
+  - {case: accs, text: четырех}
+  - {case: ablt, text: четырьмя}
+  - {case: loct, text: четырех}
+5:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: пятый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: пятая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: пятое}
+  - {case: nomn, nmbr: plur, text: пятые}
+  - {case: gent, gndr: masc, nmbr: sing, text: пятого}
+  - {case: gent, gndr: femn, nmbr: sing, text: пятой}
+  - {case: gent, gndr: neut, nmbr: sing, text: пятого}
+  - {case: gent, nmbr: plur, text: пятых}
+  - {case: datv, gndr: masc, nmbr: sing, text: пятому}
+  - {case: datv, gndr: femn, nmbr: sing, text: пятой}
+  - {case: datv, gndr: neut, nmbr: sing, text: пятому}
+  - {case: datv, nmbr: plur, text: пятым}
+  - {case: accs, gndr: femn, nmbr: sing, text: пятого}
+  - {case: accs, gndr: masc, nmbr: sing, text: пятый}
+  - {case: accs, gndr: femn, nmbr: sing, text: пятую}
+  - {case: accs, gndr: neut, nmbr: sing, text: пятое}
+  - {case: accs, nmbr: plur, text: пятые}
+  - {case: accs, nmbr: plur, text: пятых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: пятым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: пятой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: пятым}
+  - {case: ablt, nmbr: plur, text: пятыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: пятом}
+  - {case: loct, gndr: femn, nmbr: sing, text: пятой}
+  - {case: loct, gndr: neut, nmbr: sing, text: пятом}
+  - {case: loct, nmbr: plur, text: пятых}
+  o:
+  - {case: nomn, text: пять}
+  - {case: gent, text: пяти}
+  - {case: datv, text: пяти}
+  - {case: accs, text: пять}
+  - {case: ablt, text: пятью}
+  - {case: loct, text: пяти}
+6:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: шестой}
+  - {case: nomn, gndr: femn, nmbr: sing, text: шестая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: шестое}
+  - {case: nomn, nmbr: plur, text: шестые}
+  - {case: gent, gndr: masc, nmbr: sing, text: шестого}
+  - {case: gent, gndr: femn, nmbr: sing, text: шестой}
+  - {case: gent, gndr: neut, nmbr: sing, text: шестого}
+  - {case: gent, nmbr: plur, text: шестых}
+  - {case: datv, gndr: masc, nmbr: sing, text: шестому}
+  - {case: datv, gndr: femn, nmbr: sing, text: шестой}
+  - {case: datv, gndr: neut, nmbr: sing, text: шестому}
+  - {case: datv, nmbr: plur, text: шестым}
+  - {case: accs, gndr: masc, nmbr: sing, text: шестого}
+  - {case: accs, gndr: masc, nmbr: sing, text: шестой}
+  - {case: accs, gndr: femn, nmbr: sing, text: шестую}
+  - {case: accs, gndr: neut, nmbr: sing, text: шестое}
+  - {case: accs, nmbr: plur, text: шестые}
+  - {case: accs, nmbr: plur, text: шестых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: шестым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: шестой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: шестым}
+  - {case: ablt, nmbr: plur, text: шестыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: шестом}
+  - {case: loct, gndr: femn, nmbr: sing, text: шестой}
+  - {case: loct, gndr: neut, nmbr: sing, text: шестом}
+  - {case: loct, nmbr: plur, text: шестых}
+  o:
+  - {case: nomn, text: шесть}
+  - {case: gent, text: шести}
+  - {case: datv, text: шести}
+  - {case: accs, text: шесть}
+  - {case: ablt, text: шестью}
+  - {case: loct, text: шести}
+7:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: седьмой}
+  - {case: nomn, gndr: femn, nmbr: sing, text: седьмая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: седьмое}
+  - {case: nomn, nmbr: plur, text: седьмые}
+  - {case: gent, gndr: masc, nmbr: sing, text: седьмого}
+  - {case: gent, gndr: femn, nmbr: sing, text: седьмой}
+  - {case: gent, gndr: neut, nmbr: sing, text: седьмого}
+  - {case: gent, nmbr: plur, text: седьмых}
+  - {case: datv, gndr: masc, nmbr: sing, text: седьмому}
+  - {case: datv, gndr: femn, nmbr: sing, text: седьмой}
+  - {case: datv, gndr: neut, nmbr: sing, text: седьмому}
+  - {case: datv, nmbr: plur, text: седьмым}
+  - {case: accs, gndr: masc, nmbr: sing, text: седьмого}
+  - {case: accs, gndr: masc, nmbr: sing, text: седьмой}
+  - {case: accs, gndr: femn, nmbr: sing, text: седьмую}
+  - {case: accs, gndr: neut, nmbr: sing, text: седьмое}
+  - {case: accs, nmbr: plur, text: седьмые}
+  - {case: accs, nmbr: plur, text: седьмых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: седьмым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: седьмой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: седьмым}
+  - {case: ablt, nmbr: plur, text: седьмыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: седьмом}
+  - {case: loct, gndr: femn, nmbr: sing, text: седьмой}
+  - {case: loct, gndr: neut, nmbr: sing, text: седьмом}
+  - {case: loct, nmbr: plur, text: седьмых}
+  o:
+  - {case: nomn, text: семь}
+  - {case: gent, text: семи}
+  - {case: datv, text: семи}
+  - {case: accs, text: семь}
+  - {case: ablt, text: семью}
+  - {case: loct, text: семи}
+8:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: восьмой}
+  - {case: nomn, gndr: femn, nmbr: sing, text: восьмая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: восьмое}
+  - {case: nomn, nmbr: plur, text: восьмые}
+  - {case: gent, gndr: masc, nmbr: sing, text: восьмого}
+  - {case: gent, gndr: femn, nmbr: sing, text: восьмой}
+  - {case: gent, gndr: neut, nmbr: sing, text: восьмого}
+  - {case: gent, nmbr: plur, text: восьмых}
+  - {case: datv, gndr: masc, nmbr: sing, text: восьмому}
+  - {case: datv, gndr: femn, nmbr: sing, text: восьмой}
+  - {case: datv, gndr: neut, nmbr: sing, text: восьмому}
+  - {case: datv, nmbr: plur, text: восьмым}
+  - {case: accs, gndr: masc, nmbr: sing, text: восьмого}
+  - {case: accs, gndr: masc, nmbr: sing, text: восьмой}
+  - {case: accs, gndr: femn, nmbr: sing, text: восьмую}
+  - {case: accs, gndr: neut, nmbr: sing, text: восьмое}
+  - {case: accs, nmbr: plur, text: восьмые}
+  - {case: accs, nmbr: plur, text: восьмых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: восьмым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: восьмой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: восьмым}
+  - {case: ablt, nmbr: plur, text: восьмыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: восьмом}
+  - {case: loct, gndr: femn, nmbr: sing, text: восьмой}
+  - {case: loct, gndr: neut, nmbr: sing, text: восьмом}
+  - {case: loct, nmbr: plur, text: восьмых}
+  o:
+  - {case: nomn, text: восемь}
+  - {case: gent, text: восьми}
+  - {case: datv, text: восьми}
+  - {case: accs, text: восемь}
+  - {case: ablt, text: восемью}
+  - {case: ablt, text: восьмью}
+  - {case: loct, text: восьми}
+9:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: девятый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: девятая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: девятое}
+  - {case: nomn, nmbr: plur, text: девятые}
+  - {case: gent, gndr: masc, nmbr: sing, text: девятого}
+  - {case: gent, gndr: femn, nmbr: sing, text: девятой}
+  - {case: gent, gndr: neut, nmbr: sing, text: девятого}
+  - {case: gent, nmbr: plur, text: девятых}
+  - {case: datv, gndr: masc, nmbr: sing, text: девятому}
+  - {case: datv, gndr: femn, nmbr: sing, text: девятой}
+  - {case: datv, gndr: neut, nmbr: sing, text: девятому}
+  - {case: datv, nmbr: plur, text: девятым}
+  - {case: accs, gndr: masc, nmbr: sing, text: девятого}
+  - {case: accs, gndr: masc, nmbr: sing, text: девятый}
+  - {case: accs, gndr: femn, nmbr: sing, text: девятую}
+  - {case: accs, gndr: neut, nmbr: sing, text: девятое}
+  - {case: accs, nmbr: plur, text: девятые}
+  - {case: accs, nmbr: plur, text: девятых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: девятым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: девятой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: девятым}
+  - {case: ablt, nmbr: plur, text: девятыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: девятом}
+  - {case: loct, gndr: femn, nmbr: sing, text: девятой}
+  - {case: loct, gndr: neut, nmbr: sing, text: девятом}
+  - {case: loct, nmbr: plur, text: девятых}
+  o:
+  - {case: nomn, text: девять}
+  - {case: gent, text: девяти}
+  - {case: datv, text: девяти}
+  - {case: accs, text: девять}
+  - {case: ablt, text: девятью}
+  - {case: loct, text: девяти}
+10:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: десятый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: десятая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: десятое}
+  - {case: nomn, nmbr: plur, text: десятые}
+  - {case: gent, gndr: masc, nmbr: sing, text: десятого}
+  - {case: gent, gndr: femn, nmbr: sing, text: десятой}
+  - {case: gent, gndr: neut, nmbr: sing, text: десятого}
+  - {case: gent, nmbr: plur, text: десятых}
+  - {case: datv, gndr: masc, nmbr: sing, text: десятому}
+  - {case: datv, gndr: femn, nmbr: sing, text: десятой}
+  - {case: datv, gndr: neut, nmbr: sing, text: десятому}
+  - {case: datv, nmbr: plur, text: десятым}
+  - {case: accs, gndr: masc, nmbr: sing, text: десятого}
+  - {case: accs, gndr: masc, nmbr: sing, text: десятый}
+  - {case: accs, gndr: femn, nmbr: sing, text: десятую}
+  - {case: accs, gndr: neut, nmbr: sing, text: десятое}
+  - {case: accs, nmbr: plur, text: десятые}
+  - {case: accs, nmbr: plur, text: десятых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: десятым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: десятой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: десятым}
+  - {case: ablt, nmbr: plur, text: десятыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: десятом}
+  - {case: loct, gndr: femn, nmbr: sing, text: десятой}
+  - {case: loct, gndr: neut, nmbr: sing, text: десятом}
+  - {case: loct, nmbr: plur, text: десятых}
+  o:
+  - {case: nomn, text: десять}
+  - {case: gent, text: десяти}
+  - {case: datv, text: десяти}
+  - {case: accs, text: десять}
+  - {case: ablt, text: десятью}
+  - {case: loct, text: десяти}
+11:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: одиннадцатый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: одиннадцатая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: одиннадцатое}
+  - {case: nomn, nmbr: plur, text: одиннадцатые}
+  - {case: gent, gndr: masc, nmbr: sing, text: одиннадцатого}
+  - {case: gent, gndr: femn, nmbr: sing, text: одиннадцатой}
+  - {case: gent, gndr: neut, nmbr: sing, text: одиннадцатого}
+  - {case: gent, nmbr: plur, text: одиннадцатых}
+  - {case: datv, gndr: masc, nmbr: sing, text: одиннадцатому}
+  - {case: datv, gndr: femn, nmbr: sing, text: одиннадцатой}
+  - {case: datv, gndr: neut, nmbr: sing, text: одиннадцатому}
+  - {case: datv, nmbr: plur, text: одиннадцатым}
+  - {case: accs, gndr: masc, nmbr: sing, text: одиннадцатого}
+  - {case: accs, gndr: masc, nmbr: sing, text: одиннадцатый}
+  - {case: accs, gndr: femn, nmbr: sing, text: одиннадцатую}
+  - {case: accs, gndr: neut, nmbr: sing, text: одиннадцатое}
+  - {case: accs, nmbr: plur, text: одиннадцатые}
+  - {case: accs, nmbr: plur, text: одиннадцатых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: одиннадцатым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: одиннадцатой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: одиннадцатым}
+  - {case: ablt, nmbr: plur, text: одиннадцатыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: одиннадцатом}
+  - {case: loct, gndr: femn, nmbr: sing, text: одиннадцатой}
+  - {case: loct, gndr: neut, nmbr: sing, text: одиннадцатом}
+  - {case: loct, nmbr: plur, text: одиннадцатых}
+  o:
+  - {case: nomn, text: одиннадцать}
+  - {case: gent, text: одиннадцати}
+  - {case: datv, text: одиннадцати}
+  - {case: accs, text: одиннадцать}
+  - {case: ablt, text: одиннадцатью}
+  - {case: loct, text: одиннадцати}
+12:
+  ad:
+  - {case: nomn, gndr: femn, nmbr: sing, text: дюжина}
+  - {case: nomn, gndr: femn, nmbr: plur, text: дюжины}
+  - {case: gent, gndr: femn, nmbr: sing, text: дюжины}
+  - {case: gent, gndr: femn, nmbr: plur, text: дюжин}
+  - {case: datv, gndr: femn, nmbr: sing, text: дюжине}
+  - {case: datv, gndr: femn, nmbr: plur, text: дюжинам}
+  - {case: accs, gndr: femn, nmbr: sing, text: дюжину}
+  - {case: accs, gndr: femn, nmbr: plur, text: дюжины}
+  - {case: ablt, gndr: femn, nmbr: sing, text: дюжиной}
+  - {case: ablt, gndr: femn, nmbr: sing, text: дюжиною}
+  - {case: ablt, gndr: femn, nmbr: plur, text: дюжинами}
+  - {case: loct, gndr: femn, nmbr: sing, text: дюжине}
+  - {case: loct, gndr: femn, nmbr: plur, text: дюжинах}
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: двенадцатый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: двенадцатая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: двенадцатое}
+  - {case: nomn, nmbr: plur, text: двенадцатые}
+  - {case: gent, gndr: masc, nmbr: sing, text: двенадцатого}
+  - {case: gent, gndr: femn, nmbr: sing, text: двенадцатой}
+  - {case: gent, gndr: neut, nmbr: sing, text: двенадцатого}
+  - {case: gent, nmbr: plur, text: двенадцатых}
+  - {case: datv, gndr: masc, nmbr: sing, text: двенадцатому}
+  - {case: datv, gndr: femn, nmbr: sing, text: двенадцатой}
+  - {case: datv, gndr: neut, nmbr: sing, text: двенадцатому}
+  - {case: datv, nmbr: plur, text: двенадцатым}
+  - {case: accs, gndr: masc, nmbr: sing, text: двенадцатого}
+  - {case: accs, gndr: masc, nmbr: sing, text: двенадцатый}
+  - {case: accs, gndr: femn, nmbr: sing, text: двенадцатую}
+  - {case: accs, gndr: neut, nmbr: sing, text: двенадцатое}
+  - {case: accs, nmbr: plur, text: двенадцатые}
+  - {case: accs, nmbr: plur, text: двенадцатых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: двенадцатым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: двенадцатой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: двенадцатым}
+  - {case: ablt, nmbr: plur, text: двенадцатыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: двенадцатом}
+  - {case: loct, gndr: femn, nmbr: sing, text: двенадцатой}
+  - {case: loct, gndr: neut, nmbr: sing, text: двенадцатом}
+  - {case: loct, nmbr: plur, text: двенадцатых}
+  o:
+  - {case: nomn, text: двенадцать}
+  - {case: gent, text: двенадцати}
+  - {case: datv, text: двенадцати}
+  - {case: accs, text: двенадцать}
+  - {case: ablt, text: двенадцатью}
+  - {case: loct, text: двенадцати}
+13:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: тринадцатый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: тринадцатая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: тринадцатое}
+  - {case: nomn, nmbr: plur, text: тринадцатые}
+  - {case: gent, gndr: masc, nmbr: sing, text: тринадцатого}
+  - {case: gent, gndr: femn, nmbr: sing, text: тринадцатой}
+  - {case: gent, gndr: neut, nmbr: sing, text: тринадцатого}
+  - {case: gent, nmbr: plur, text: тринадцатых}
+  - {case: datv, gndr: masc, nmbr: sing, text: тринадцатому}
+  - {case: datv, gndr: femn, nmbr: sing, text: тринадцатой}
+  - {case: datv, gndr: neut, nmbr: sing, text: тринадцатому}
+  - {case: datv, nmbr: plur, text: тринадцатым}
+  - {case: accs, gndr: masc, nmbr: sing, text: тринадцатого}
+  - {case: accs, gndr: masc, nmbr: sing, text: тринадцатый}
+  - {case: accs, gndr: femn, nmbr: sing, text: тринадцатую}
+  - {case: accs, gndr: neut, nmbr: sing, text: трина��цатое}
+  - {case: accs, nmbr: plur, text: тринадцатые}
+  - {case: accs, nmbr: plur, text: тринадцатых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: тринадцатым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: тринадцатой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: тринадцатым}
+  - {case: ablt, nmbr: plur, text: тринадцатыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: тринадцатом}
+  - {case: loct, gndr: femn, nmbr: sing, text: тринадцатой}
+  - {case: loct, gndr: neut, nmbr: sing, text: тринадцатом}
+  - {case: loct, nmbr: plur, text: тринадцатых}
+  o:
+  - {case: nomn, text: тринадцать}
+  - {case: gent, text: тринадцати}
+  - {case: datv, text: тринадцати}
+  - {case: accs, text: тринадцать}
+  - {case: ablt, text: тринадцатью}
+  - {case: loct, text: тринадцати}
+14:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: четырнадцатый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: четырнадцатая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: четырнадцатое}
+  - {case: nomn, nmbr: plur, text: четырнадцатые}
+  - {case: gent, gndr: masc, nmbr: sing, text: четырнадцатого}
+  - {case: gent, gndr: femn, nmbr: sing, text: четырнадцатой}
+  - {case: gent, gndr: neut, nmbr: sing, text: четырнадцатого}
+  - {case: gent, nmbr: plur, text: четырнадцатых}
+  - {case: datv, gndr: masc, nmbr: sing, text: четырнадцатому}
+  - {case: datv, gndr: femn, nmbr: sing, text: четырнадцатой}
+  - {case: datv, gndr: neut, nmbr: sing, text: четырнадцатому}
+  - {case: datv, nmbr: plur, text: четырнадцатым}
+  - {case: accs, gndr: masc, nmbr: sing, text: четырнадцатого}
+  - {case: accs, gndr: masc, nmbr: sing, text: четырнадцатый}
+  - {case: accs, gndr: femn, nmbr: sing, text: четырнадцатую}
+  - {case: accs, gndr: neut, nmbr: sing, text: четырнадцатое}
+  - {case: accs, nmbr: plur, text: четырнадцатые}
+  - {case: accs, nmbr: plur, text: четырнадцатых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: четырнадцатым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: четырнадцатой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: четырнадцатым}
+  - {case: ablt, nmbr: plur, text: четырнадцатыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: четырнадцатом}
+  - {case: loct, gndr: femn, nmbr: sing, text: четырнадцатой}
+  - {case: loct, gndr: neut, nmbr: sing, text: четырнадцатом}
+  - {case: loct, nmbr: plur, text: четырнадцатых}
+  o:
+  - {case: nomn, text: четырнадцать}
+  - {case: gent, text: четырнадцати}
+  - {case: datv, text: четырнадцати}
+  - {case: accs, text: четырнадцать}
+  - {case: ablt, text: четырнадцатью}
+  - {case: loct, text: четырнадцати}
+15:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: пятнадцатый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: пятнадцатая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: пятнадцатое}
+  - {case: nomn, nmbr: plur, text: пятнадцатые}
+  - {case: gent, gndr: masc, nmbr: sing, text: пятнадцатого}
+  - {case: gent, gndr: femn, nmbr: sing, text: пятнадцатой}
+  - {case: gent, gndr: neut, nmbr: sing, text: пятнадцатого}
+  - {case: gent, nmbr: plur, text: пятнадцатых}
+  - {case: datv, gndr: masc, nmbr: sing, text: пятнадцатому}
+  - {case: datv, gndr: femn, nmbr: sing, text: пятнадцатой}
+  - {case: datv, gndr: neut, nmbr: sing, text: пятнадцатому}
+  - {case: datv, nmbr: plur, text: пятнадцатым}
+  - {case: accs, gndr: masc, nmbr: sing, text: пятнадцатого}
+  - {case: accs, gndr: masc, nmbr: sing, text: пятнадцатый}
+  - {case: accs, gndr: femn, nmbr: sing, text: пятнадцатую}
+  - {case: accs, gndr: neut, nmbr: sing, text: пятнадцатое}
+  - {case: accs, nmbr: plur, text: пятнадцатые}
+  - {case: accs, nmbr: plur, text: пятнадцатых}
+  - {case: ablt, gndr: femn, nmbr: sing, text: пятнадцатым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: пятнадцатой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: пятнадцатым}
+  - {case: ablt, nmbr: plur, text: пятнадцатыми}
+  - {case: loct, gndr: femn, nmbr: sing, text: пятнадцатом}
+  - {case: loct, gndr: femn, nmbr: sing, text: пятнадцатой}
+  - {case: loct, gndr: neut, nmbr: sing, text: пятнадцатом}
+  - {case: loct, nmbr: plur, text: пятнадцатых}
+  o:
+  - {case: nomn, text: пятнадцать}
+  - {case: gent, text: пятнадцати}
+  - {case: datv, text: пятнадцати}
+  - {case: accs, text: пятнадцать}
+  - {case: ablt, text: пятнадцатью}
+  - {case: loct, text: пятнадцати}
+16:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: шестнадцатый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: шестнадцатая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: шестнадцатое}
+  - {case: nomn, nmbr: plur, text: шестнадцатые}
+  - {case: gent, gndr: masc, nmbr: sing, text: шестнадцатого}
+  - {case: gent, gndr: femn, nmbr: sing, text: шестнадцатой}
+  - {case: gent, gndr: neut, nmbr: sing, text: шестнадцатого}
+  - {case: gent, nmbr: plur, text: шестнадцатых}
+  - {case: datv, gndr: masc, nmbr: sing, text: шестнадцатому}
+  - {case: datv, gndr: femn, nmbr: sing, text: шестнадцатой}
+  - {case: datv, gndr: neut, nmbr: sing, text: шестнадцатому}
+  - {case: datv, nmbr: plur, text: шестнадцатым}
+  - {case: accs, gndr: masc, nmbr: sing, text: шестнадцатого}
+  - {case: accs, gndr: masc, nmbr: sing, text: шестнадцатый}
+  - {case: accs, gndr: femn, nmbr: sing, text: шестнадцатую}
+  - {case: accs, gndr: neut, nmbr: sing, text: шестнадцатое}
+  - {case: accs, nmbr: plur, text: шестнадцатые}
+  - {case: accs, nmbr: plur, text: шестнадцатых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: шестнадцатым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: шестнадцатой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: шестнадцатым}
+  - {case: ablt, nmbr: plur, text: шестнадцатыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: шестнадцатом}
+  - {case: loct, gndr: femn, nmbr: sing, text: шестнадцатой}
+  - {case: loct, gndr: neut, nmbr: sing, text: шестнадцатом}
+  - {case: loct, nmbr: plur, text: шестнадцатых}
+  o:
+  - {case: nomn, text: шестнадцать}
+  - {case: gent, text: шестнадцати}
+  - {case: datv, text: шестнадцати}
+  - {case: accs, text: шестнадцать}
+  - {case: ablt, text: шестнадцатью}
+  - {case: loct, text: шестнадцати}
+17:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: семнадцатый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: семнадцатая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: семнадцатое}
+  - {case: nomn, nmbr: plur, text: семнадцатые}
+  - {case: gent, gndr: masc, nmbr: sing, text: семнадцатого}
+  - {case: gent, gndr: femn, nmbr: sing, text: семнадцатой}
+  - {case: gent, gndr: neut, nmbr: sing, text: семнадцатого}
+  - {case: gent, nmbr: plur, text: семнадцатых}
+  - {case: datv, gndr: masc, nmbr: sing, text: семнадцатому}
+  - {case: datv, gndr: femn, nmbr: sing, text: семнадцатой}
+  - {case: datv, gndr: neut, nmbr: sing, text: семнадцатому}
+  - {case: datv, nmbr: plur, text: семнадцатым}
+  - {case: accs, gndr: masc, nmbr: sing, text: семнадцатого}
+  - {case: accs, gndr: masc, nmbr: sing, text: семнадцатый}
+  - {case: accs, gndr: femn, nmbr: sing, text: семнадцатую}
+  - {case: accs, gndr: neut, nmbr: sing, text: семнадцатое}
+  - {case: accs, nmbr: plur, text: семнадцатые}
+  - {case: accs, nmbr: plur, text: семнадцатых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: семнадцатым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: семнадцатой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: семнадцатым}
+  - {case: ablt, nmbr: plur, text: семнадцатыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: семнадцатом}
+  - {case: loct, gndr: femn, nmbr: sing, text: семнадцатой}
+  - {case: loct, gndr: neut, nmbr: sing, text: семнадцатом}
+  - {case: loct, nmbr: plur, text: семнадцатых}
+  o:
+  - {case: nomn, text: семнадцать}
+  - {case: gent, text: семнадцати}
+  - {case: datv, text: семнадцати}
+  - {case: accs, text: семнадцать}
+  - {case: ablt, text: семнадцатью}
+  - {case: loct, text: семнадцати}
+18:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: восемнадцатый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: восемнадцатая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: восемнадцатое}
+  - {case: nomn, nmbr: plur, text: восемнадцатые}
+  - {case: gent, gndr: masc, nmbr: sing, text: восемнадцатого}
+  - {case: gent, gndr: femn, nmbr: sing, text: восемнадцатой}
+  - {case: gent, gndr: neut, nmbr: sing, text: восемнадцатого}
+  - {case: gent, nmbr: plur, text: восемнадцатых}
+  - {case: datv, gndr: masc, nmbr: sing, text: восемнадцатому}
+  - {case: datv, gndr: femn, nmbr: sing, text: восемнадцатой}
+  - {case: datv, gndr: neut, nmbr: sing, text: восемнадцатому}
+  - {case: datv, nmbr: plur, text: восемнадцатым}
+  - {case: accs, gndr: masc, nmbr: sing, text: восемнадцатого}
+  - {case: accs, gndr: masc, nmbr: sing, text: восемнадцатый}
+  - {case: accs, gndr: femn, nmbr: sing, text: восемнадцатую}
+  - {case: accs, gndr: neut, nmbr: sing, text: восемнадцатое}
+  - {case: accs, nmbr: plur, text: восемнадцатые}
+  - {case: accs, nmbr: plur, text: восемнадцатых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: восемнадцатым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: восемнадцатой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: восемнадцатым}
+  - {case: ablt, nmbr: plur, text: восемнадцатыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: восемнадцатом}
+  - {case: loct, gndr: femn, nmbr: sing, text: восемнадцатой}
+  - {case: loct, gndr: neut, nmbr: sing, text: восемнадцатом}
+  - {case: loct, nmbr: plur, text: восемнадцатых}
+  o:
+  - {case: nomn, text: восемнадцать}
+  - {case: gent, text: восемнадцати}
+  - {case: datv, text: восемнадцати}
+  - {case: accs, text: восемнадцать}
+  - {case: ablt, text: восемнадцатью}
+  - {case: loct, text: восемнадцати}
+19:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: девятнадцатый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: девятнадцатая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: девятнадцатое}
+  - {case: nomn, nmbr: plur, text: девятнадцатые}
+  - {case: gent, gndr: masc, nmbr: sing, text: девятнадцатого}
+  - {case: gent, gndr: femn, nmbr: sing, text: девятнадцатой}
+  - {case: gent, gndr: neut, nmbr: sing, text: девятнадцатого}
+  - {case: gent, nmbr: plur, text: девятнадцатых}
+  - {case: datv, gndr: masc, nmbr: sing, text: девятнадцатому}
+  - {case: datv, gndr: femn, nmbr: sing, text: девятнадцатой}
+  - {case: datv, gndr: neut, nmbr: sing, text: девятнадцатому}
+  - {case: datv, nmbr: plur, text: девятнадцатым}
+  - {case: accs, gndr: masc, nmbr: sing, text: девятнадцатого}
+  - {case: accs, gndr: masc, nmbr: sing, text: девятнадцатый}
+  - {case: accs, gndr: femn, nmbr: sing, text: девятнадцатую}
+  - {case: accs, gndr: neut, nmbr: sing, text: девятнадцатое}
+  - {case: accs, nmbr: plur, text: девятнадцатые}
+  - {case: accs, nmbr: plur, text: девятнадцатых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: девятнадцатым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: девятнадцатой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: девятнадцатым}
+  - {case: ablt, nmbr: plur, text: девятнадцатыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: девятнадцатом}
+  - {case: loct, gndr: femn, nmbr: sing, text: девятнадцатой}
+  - {case: loct, gndr: neut, nmbr: sing, text: девятнадцатом}
+  - {case: loct, nmbr: plur, text: девятнадцатых}
+  o:
+  - {case: nomn, text: девятнадцать}
+  - {case: gent, text: девятнадцати}
+  - {case: datv, text: девятнадцати}
+  - {case: accs, text: девятнадцать}
+  - {case: ablt, text: девятнадцатью}
+  - {case: loct, text: девятнадцати}
+20:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: двадцатый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: двадцатая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: двадцатое}
+  - {case: nomn, nmbr: plur, text: двадцатые}
+  - {case: gent, gndr: masc, nmbr: sing, text: двадцатого}
+  - {case: gent, gndr: femn, nmbr: sing, text: двадцатой}
+  - {case: gent, gndr: neut, nmbr: sing, text: двадцатого}
+  - {case: gent, nmbr: plur, text: двадцатых}
+  - {case: datv, gndr: masc, nmbr: sing, text: двадцатому}
+  - {case: datv, gndr: femn, nmbr: sing, text: двадцатой}
+  - {case: datv, gndr: neut, nmbr: sing, text: двадцатому}
+  - {case: datv, nmbr: plur, text: двадцатым}
+  - {case: accs, gndr: masc, nmbr: sing, text: двадцатого}
+  - {case: accs, gndr: masc, nmbr: sing, text: двадцатый}
+  - {case: accs, gndr: femn, nmbr: sing, text: двадцатую}
+  - {case: accs, gndr: neut, nmbr: sing, text: двадцатое}
+  - {case: accs, nmbr: plur, text: двадцатые}
+  - {case: accs, nmbr: plur, text: двадцатых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: двадцатым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: двадцатой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: двадцатым}
+  - {case: ablt, nmbr: plur, text: двадцатыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: двадцатом}
+  - {case: loct, gndr: femn, nmbr: sing, text: двадцатой}
+  - {case: loct, gndr: neut, nmbr: sing, text: двадцатом}
+  - {case: loct, nmbr: plur, text: двадцатых}
+  o:
+  - {case: nomn, text: двадцать}
+  - {case: gent, text: двадцати}
+  - {case: datv, text: двадцати}
+  - {case: accs, text: двадцать}
+  - {case: ablt, text: двадцатью}
+  - {case: loct, text: двадцати}
+30:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: тридцатый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: тридцатая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: тридцатое}
+  - {case: nomn, nmbr: plur, text: тридцатые}
+  - {case: gent, gndr: masc, nmbr: sing, text: тридцатого}
+  - {case: gent, gndr: femn, nmbr: sing, text: тридцатой}
+  - {case: gent, gndr: neut, nmbr: sing, text: тридцатого}
+  - {case: gent, nmbr: plur, text: тридцатых}
+  - {case: datv, gndr: masc, nmbr: sing, text: тридцатому}
+  - {case: datv, gndr: femn, nmbr: sing, text: тридцатой}
+  - {case: datv, gndr: neut, nmbr: sing, text: тридцатому}
+  - {case: datv, nmbr: plur, text: тридцатым}
+  - {case: accs, gndr: masc, nmbr: sing, text: тридцатого}
+  - {case: accs, gndr: masc, nmbr: sing, text: тридцатый}
+  - {case: accs, gndr: femn, nmbr: sing, text: тридцатую}
+  - {case: accs, gndr: neut, nmbr: sing, text: тридцатое}
+  - {case: accs, nmbr: plur, text: тридцатые}
+  - {case: accs, nmbr: plur, text: тридцатых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: тридцатым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: тридцатой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: тридцатым}
+  - {case: ablt, nmbr: plur, text: тридцатыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: тридцатом}
+  - {case: loct, gndr: femn, nmbr: sing, text: тридцатой}
+  - {case: loct, gndr: neut, nmbr: sing, text: тридцатом}
+  - {case: loct, nmbr: plur, text: тридцатых}
+  o:
+  - {case: nomn, text: тридцать}
+  - {case: gent, text: тридцати}
+  - {case: datv, text: тридцати}
+  - {case: accs, text: тридцать}
+  - {case: ablt, text: тридцатью}
+  - {case: loct, text: тридцати}
+40:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: сороковой}
+  - {case: nomn, gndr: femn, nmbr: sing, text: сороковая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: сороковое}
+  - {case: nomn, nmbr: plur, text: сороковые}
+  - {case: gent, gndr: masc, nmbr: sing, text: сорокового}
+  - {case: gent, gndr: femn, nmbr: sing, text: сороковой}
+  - {case: gent, gndr: neut, nmbr: sing, text: сорокового}
+  - {case: gent, nmbr: plur, text: сороковых}
+  - {case: datv, gndr: masc, nmbr: sing, text: сороковому}
+  - {case: datv, gndr: femn, nmbr: sing, text: сороковой}
+  - {case: datv, gndr: neut, nmbr: sing, text: сороковому}
+  - {case: datv, nmbr: plur, text: сороковым}
+  - {case: accs, gndr: masc, nmbr: sing, text: сорокового}
+  - {case: accs, gndr: masc, nmbr: sing, text: сороковой}
+  - {case: accs, gndr: femn, nmbr: sing, text: сороковую}
+  - {case: accs, gndr: neut, nmbr: sing, text: сороковое}
+  - {case: accs, nmbr: plur, text: сороковые}
+  - {case: accs, nmbr: plur, text: сороковых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: сороковым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: сороковой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: сороковым}
+  - {case: ablt, nmbr: plur, text: сороковыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: сороковом}
+  - {case: loct, gndr: femn, nmbr: sing, text: сороковой}
+  - {case: loct, gndr: neut, nmbr: sing, text: сороковом}
+  - {case: loct, nmbr: plur, text: сороковых}
+  o:
+  - {case: nomn, text: сорок}
+  - {case: gent, text: сорока}
+  - {case: datv, text: сорока}
+  - {case: accs, text: сорок}
+  - {case: ablt, text: сорока}
+  - {case: loct, text: сорока}
+50:
+  ad:
+  - {case: nomn, gndr: neut, nmbr: sing, text: полсотни}
+  - {case: accs, gndr: neut, nmbr: sing, text: полсотни}
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: пятидесятый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: пятидесятая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: пятидесятое}
+  - {case: nomn, nmbr: plur, text: пятидесятые}
+  - {case: gent, gndr: masc, nmbr: sing, text: пятидесятого}
+  - {case: gent, gndr: femn, nmbr: sing, text: пятидесятой}
+  - {case: gent, gndr: neut, nmbr: sing, text: пятидесятого}
+  - {case: gent, nmbr: plur, text: пятидесятых}
+  - {case: datv, gndr: masc, nmbr: sing, text: пятидесятому}
+  - {case: datv, gndr: femn, nmbr: sing, text: пятидесятой}
+  - {case: datv, gndr: neut, nmbr: sing, text: пятидесятому}
+  - {case: datv, nmbr: plur, text: пятидесятым}
+  - {case: accs, gndr: masc, nmbr: sing, text: пятидесятого}
+  - {case: accs, gndr: masc, nmbr: sing, text: пятидесятый}
+  - {case: accs, gndr: femn, nmbr: sing, text: пятидесятую}
+  - {case: accs, gndr: neut, nmbr: sing, text: пятидесятое}
+  - {case: accs, nmbr: plur, text: пятидесятые}
+  - {case: accs, nmbr: plur, text: пятидесятых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: пятидесятым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: пятидесятой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: пятидесятым}
+  - {case: ablt, nmbr: plur, text: пятидесятыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: пятидесятом}
+  - {case: loct, gndr: femn, nmbr: sing, text: пятидесятой}
+  - {case: loct, gndr: neut, nmbr: sing, text: пятидесятом}
+  - {case: loct, nmbr: plur, text: пятидесятых}
+  o:
+  - {case: nomn, text: пятьдесят}
+  - {case: gent, text: пятидесяти}
+  - {case: datv, text: пятидесяти}
+  - {case: accs, text: пятьдесят}
+  - {case: ablt, text: пятьюдесятью}
+  - {case: loct, text: пятидесяти}
+60:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: шестидесятый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: шестидесятая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: шестидесятое}
+  - {case: nomn, nmbr: plur, text: шестидесятые}
+  - {case: gent, gndr: masc, nmbr: sing, text: шестидесятого}
+  - {case: gent, gndr: femn, nmbr: sing, text: шестидесятой}
+  - {case: gent, gndr: neut, nmbr: sing, text: шестидесятого}
+  - {case: gent, nmbr: plur, text: шестидесятых}
+  - {case: datv, gndr: masc, nmbr: sing, text: шестидесятому}
+  - {case: datv, gndr: femn, nmbr: sing, text: шестидесятой}
+  - {case: datv, gndr: neut, nmbr: sing, text: шестидесятому}
+  - {case: datv, nmbr: plur, text: шестидесятым}
+  - {case: accs, gndr: masc, nmbr: sing, text: шестидесятого}
+  - {case: accs, gndr: masc, nmbr: sing, text: шестидесятый}
+  - {case: accs, gndr: femn, nmbr: sing, text: шестидесятую}
+  - {case: accs, gndr: neut, nmbr: sing, text: шестидесятое}
+  - {case: accs, nmbr: plur, text: шестидесятые}
+  - {case: accs, nmbr: plur, text: шестидесятых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: шестидесятым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: шестидесятой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: шестидесятым}
+  - {case: ablt, nmbr: plur, text: шестидесятыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: шестидесятом}
+  - {case: loct, gndr: femn, nmbr: sing, text: шестидесятой}
+  - {case: loct, gndr: neut, nmbr: sing, text: шестидесятом}
+  - {case: loct, nmbr: plur, text: шестидесятых}
+  o:
+  - {case: nomn, text: шестьдесят}
+  - {case: gent, text: шестидесяти}
+  - {case: datv, text: шестидесяти}
+  - {case: accs, text: шестьдесят}
+  - {case: ablt, text: шестьюдесятью}
+  - {case: loct, text: шестидесяти}
+70:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: семидесятый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: семидесятая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: семидесятое}
+  - {case: nomn, nmbr: plur, text: семидесятые}
+  - {case: gent, gndr: masc, nmbr: sing, text: семидесятого}
+  - {case: gent, gndr: femn, nmbr: sing, text: семидесятой}
+  - {case: gent, gndr: neut, nmbr: sing, text: семидесятого}
+  - {case: gent, nmbr: plur, text: семидесятых}
+  - {case: datv, gndr: masc, nmbr: sing, text: семидесятому}
+  - {case: datv, gndr: femn, nmbr: sing, text: семидесятой}
+  - {case: datv, gndr: neut, nmbr: sing, text: семидесятому}
+  - {case: datv, nmbr: plur, text: семидесятым}
+  - {case: accs, gndr: masc, nmbr: sing, text: семидесятого}
+  - {case: accs, gndr: masc, nmbr: sing, text: семидесятый}
+  - {case: accs, gndr: femn, nmbr: sing, text: семидесятую}
+  - {case: accs, gndr: neut, nmbr: sing, text: семидесятое}
+  - {case: accs, nmbr: plur, text: семидесятые}
+  - {case: accs, nmbr: plur, text: семидесятых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: семидесятым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: семидесятой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: семидесятым}
+  - {case: ablt, nmbr: plur, text: семидесятыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: семидесятом}
+  - {case: loct, gndr: femn, nmbr: sing, text: семидесятой}
+  - {case: loct, gndr: neut, nmbr: sing, text: семидесятом}
+  - {case: loct, nmbr: plur, text: семидесятых}
+  o:
+  - {case: nomn, text: семьдесят}
+  - {case: gent, text: семидесяти}
+  - {case: datv, text: семидесяти}
+  - {case: accs, text: семьдесят}
+  - {case: ablt, text: семьюдесятью}
+  - {case: loct, text: семидесяти}
+80:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: восьмидесятый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: восьмидесятая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: восьмидесятое}
+  - {case: nomn, nmbr: plur, text: восьмидесятые}
+  - {case: gent, gndr: masc, nmbr: sing, text: восьмидесятого}
+  - {case: gent, gndr: femn, nmbr: sing, text: восьмидесятой}
+  - {case: gent, gndr: neut, nmbr: sing, text: восьмидесятого}
+  - {case: gent, nmbr: plur, text: восьмидесятых}
+  - {case: datv, gndr: masc, nmbr: sing, text: восьмидесятому}
+  - {case: datv, gndr: femn, nmbr: sing, text: восьмидесятой}
+  - {case: datv, gndr: neut, nmbr: sing, text: восьмидесятому}
+  - {case: datv, nmbr: plur, text: восьмидесятым}
+  - {case: accs, gndr: masc, nmbr: sing, text: восьмидесятого}
+  - {case: accs, gndr: masc, nmbr: sing, text: восьмидесятый}
+  - {case: accs, gndr: femn, nmbr: sing, text: восьмидесятую}
+  - {case: accs, gndr: neut, nmbr: sing, text: восьмидесятое}
+  - {case: accs, nmbr: plur, text: восьмидесятые}
+  - {case: accs, nmbr: plur, text: восьмидесятых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: восьмидесятым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: восьмидесятой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: восьмидесятым}
+  - {case: ablt, nmbr: plur, text: восьмидесятыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: восьмидесятом}
+  - {case: loct, gndr: femn, nmbr: sing, text: восьмидесятой}
+  - {case: loct, gndr: neut, nmbr: sing, text: восьмидесятом}
+  - {case: loct, nmbr: plur, text: восьмидесятых}
+  o:
+  - {case: nomn, text: восемьдесят}
+  - {case: gent, text: восьмидесяти}
+  - {case: datv, text: восьмидесяти}
+  - {case: accs, text: восемьдесят}
+  - {case: ablt, text: восемьюдесятью}
+  - {case: ablt, text: восьмьюдесятью}
+  - {case: loct, text: восьмидесяти}
+90:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: девяностый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: девяностая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: девяностое}
+  - {case: nomn, nmbr: plur, text: девяностые}
+  - {case: gent, gndr: masc, nmbr: sing, text: девяностого}
+  - {case: gent, gndr: femn, nmbr: sing, text: девяностой}
+  - {case: gent, gndr: neut, nmbr: sing, text: девяностого}
+  - {case: gent, nmbr: plur, text: девяностых}
+  - {case: datv, gndr: masc, nmbr: sing, text: девяностому}
+  - {case: datv, gndr: femn, nmbr: sing, text: девяностой}
+  - {case: datv, gndr: neut, nmbr: sing, text: девяностому}
+  - {case: datv, nmbr: plur, text: девяностым}
+  - {case: accs, gndr: masc, nmbr: sing, text: девяностого}
+  - {case: accs, gndr: masc, nmbr: sing, text: девяностый}
+  - {case: accs, gndr: femn, nmbr: sing, text: девяностую}
+  - {case: accs, gndr: neut, nmbr: sing, text: девяностое}
+  - {case: accs, nmbr: plur, text: девяностые}
+  - {case: accs, nmbr: plur, text: девяностых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: девяностым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: девяностой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: девяностым}
+  - {case: ablt, nmbr: plur, text: девяностыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: девяностом}
+  - {case: loct, gndr: femn, nmbr: sing, text: девяностой}
+  - {case: loct, gndr: neut, nmbr: sing, text: девяностом}
+  - {case: loct, nmbr: plur, text: девяностых}
+  o: []
+100:
+  ad:
+  - {case: nomn, gndr: femn, nmbr: sing, text: сотня}
+  - {case: nomn, gndr: femn, nmbr: plur, text: сотни}
+  - {case: gent, gndr: femn, nmbr: sing, text: сотни}
+  - {case: gent, gndr: femn, nmbr: plur, text: сот}
+  - {case: gent, gndr: femn, nmbr: plur, text: сотен}
+  - {case: datv, gndr: femn, nmbr: sing, text: сотне}
+  - {case: datv, gndr: femn, nmbr: plur, text: сотням}
+  - {case: accs, gndr: femn, nmbr: sing, text: сотню}
+  - {case: accs, gndr: femn, nmbr: plur, text: сотни}
+  - {case: ablt, gndr: femn, nmbr: sing, text: сотней}
+  - {case: ablt, gndr: femn, nmbr: sing, text: сотнею}
+  - {case: ablt, gndr: femn, nmbr: plur, text: сотнями}
+  - {case: loct, gndr: femn, nmbr: sing, text: сотне}
+  - {case: loct, gndr: femn, nmbr: plur, text: сотнях}
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: сотый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: сотая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: сотое}
+  - {case: nomn, nmbr: plur, text: сотые}
+  - {case: gent, gndr: masc, nmbr: sing, text: сотого}
+  - {case: gent, gndr: femn, nmbr: sing, text: сотой}
+  - {case: gent, gndr: neut, nmbr: sing, text: сотого}
+  - {case: gent, nmbr: plur, text: сотых}
+  - {case: datv, gndr: masc, nmbr: sing, text: сотому}
+  - {case: datv, gndr: femn, nmbr: sing, text: сотой}
+  - {case: datv, gndr: neut, nmbr: sing, text: сотому}
+  - {case: datv, nmbr: plur, text: сотым}
+  - {case: accs, gndr: masc, nmbr: sing, text: сотого}
+  - {case: accs, gndr: masc, nmbr: sing, text: сотый}
+  - {case: accs, gndr: femn, nmbr: sing, text: сотую}
+  - {case: accs, gndr: neut, nmbr: sing, text: сотое}
+  - {case: accs, nmbr: plur, text: сотые}
+  - {case: accs, nmbr: plur, text: сотых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: сотым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: сотой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: сотым}
+  - {case: ablt, nmbr: plur, text: сотыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: сотом}
+  - {case: loct, gndr: femn, nmbr: sing, text: сотой}
+  - {case: loct, gndr: neut, nmbr: sing, text: сотом}
+  - {case: loct, nmbr: plur, text: сотых}
+  o:
+  - {case: nomn, text: сто}
+  - {case: gent, text: ста}
+  - {case: datv, text: ста}
+  - {case: accs, text: сто}
+  - {case: ablt, text: ста}
+  - {case: loct, text: ста}
+200:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: двухсотый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: двухсотая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: двухсотое}
+  - {case: nomn, nmbr: plur, text: двухсотые}
+  - {case: gent, gndr: masc, nmbr: sing, text: двухсотого}
+  - {case: gent, gndr: femn, nmbr: sing, text: двухсотой}
+  - {case: gent, gndr: neut, nmbr: sing, text: двухсотого}
+  - {case: gent, nmbr: plur, text: двухсотых}
+  - {case: datv, gndr: masc, nmbr: sing, text: двухсотому}
+  - {case: datv, gndr: femn, nmbr: sing, text: двухсотой}
+  - {case: datv, gndr: neut, nmbr: sing, text: двухсотому}
+  - {case: datv, nmbr: plur, text: двухсотым}
+  - {case: accs, gndr: masc, nmbr: sing, text: двухсотого}
+  - {case: accs, gndr: masc, nmbr: sing, text: двухсотый}
+  - {case: accs, gndr: femn, nmbr: sing, text: двухсотую}
+  - {case: accs, gndr: neut, nmbr: sing, text: двухсотое}
+  - {case: accs, nmbr: plur, text: двухсотые}
+  - {case: accs, nmbr: plur, text: двухсотых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: двухсотым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: двухсотой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: двухсотым}
+  - {case: ablt, nmbr: plur, text: двухсотыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: двухсотом}
+  - {case: loct, gndr: femn, nmbr: sing, text: двухсотой}
+  - {case: loct, gndr: neut, nmbr: sing, text: двухсотом}
+  - {case: loct, nmbr: plur, text: двухсотых}
+  o:
+  - {case: nomn, text: двести}
+  - {case: gent, text: двухсот}
+  - {case: datv, text: двумстам}
+  - {case: accs, text: двести}
+  - {case: ablt, text: двумястами}
+  - {case: loct, text: двухстах}
+300:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: трёхсотый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: трёхсотая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: трёхсотое}
+  - {case: nomn, nmbr: plur, text: трёхсотые}
+  - {case: gent, gndr: masc, nmbr: sing, text: трёхсотого}
+  - {case: gent, gndr: femn, nmbr: sing, text: трёхсотой}
+  - {case: gent, gndr: neut, nmbr: sing, text: трёхсотого}
+  - {case: gent, nmbr: plur, text: трёхсотых}
+  - {case: datv, gndr: masc, nmbr: sing, text: трёхсотому}
+  - {case: datv, gndr: femn, nmbr: sing, text: трёхсотой}
+  - {case: datv, gndr: neut, nmbr: sing, text: трёхсотому}
+  - {case: datv, nmbr: plur, text: трёхсотым}
+  - {case: accs, gndr: masc, nmbr: sing, text: трёхсотого}
+  - {case: accs, gndr: masc, nmbr: sing, text: трёхсотый}
+  - {case: accs, gndr: femn, nmbr: sing, text: трёхсотую}
+  - {case: accs, gndr: neut, nmbr: sing, text: трёхсотое}
+  - {case: accs, nmbr: plur, text: трёхсотые}
+  - {case: accs, nmbr: plur, text: трёхсотых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: трёхсотым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: трёхсотой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: трёхсотым}
+  - {case: ablt, nmbr: plur, text: трёхсотыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: трёхсотом}
+  - {case: loct, gndr: femn, nmbr: sing, text: трёхсотой}
+  - {case: loct, gndr: neut, nmbr: sing, text: трёхсотом}
+  - {case: loct, nmbr: plur, text: трёхсотых}
+  o:
+  - {case: nomn, text: триста}
+  - {case: gent, text: трехсот}
+  - {case: datv, text: тремстам}
+  - {case: accs, text: триста}
+  - {case: ablt, text: тремястами}
+  - {case: loct, text: трехстах}
+400:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: четырёхсотый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: четырёхсотая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: четырёхсотое}
+  - {case: nomn, nmbr: plur, text: четырёхсотые}
+  - {case: gent, gndr: masc, nmbr: sing, text: четырёхсотого}
+  - {case: gent, gndr: femn, nmbr: sing, text: четырёхсотой}
+  - {case: gent, gndr: neut, nmbr: sing, text: четырёхсотого}
+  - {case: gent, nmbr: plur, text: четырёхсотых}
+  - {case: datv, gndr: masc, nmbr: sing, text: четырёхсотому}
+  - {case: datv, gndr: femn, nmbr: sing, text: четырёхсотой}
+  - {case: datv, gndr: neut, nmbr: sing, text: четырёхсотому}
+  - {case: datv, nmbr: plur, text: четырёхсотым}
+  - {case: accs, gndr: masc, nmbr: sing, text: четырёхсотого}
+  - {case: accs, gndr: masc, nmbr: sing, text: четырёхсотый}
+  - {case: accs, gndr: femn, nmbr: sing, text: четырёхсотую}
+  - {case: accs, gndr: neut, nmbr: sing, text: четырёхсотое}
+  - {case: accs, nmbr: plur, text: четырёхсотые}
+  - {case: accs, nmbr: plur, text: четырёхсотых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: четырёхсотым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: четырёхсотой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: четырёхсотым}
+  - {case: ablt, nmbr: plur, text: четырёхсотыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: четырёхсотом}
+  - {case: loct, gndr: femn, nmbr: sing, text: четырёхсотой}
+  - {case: loct, gndr: neut, nmbr: sing, text: четырёхсотом}
+  - {case: loct, nmbr: plur, text: четырёхсотых}
+  o:
+  - {case: nomn, text: четыреста}
+  - {case: gent, text: четырехсот}
+  - {case: datv, text: четыремстам}
+  - {case: accs, text: четыреста}
+  - {case: ablt, text: четырьмястами}
+  - {case: loct, text: четырехстах}
+500:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: пятисотый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: пятисотая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: пятисотое}
+  - {case: nomn, nmbr: plur, text: пятисотые}
+  - {case: gent, gndr: masc, nmbr: sing, text: пятисотого}
+  - {case: gent, gndr: femn, nmbr: sing, text: пятисотой}
+  - {case: gent, gndr: neut, nmbr: sing, text: пятисотого}
+  - {case: gent, nmbr: plur, text: пятисотых}
+  - {case: datv, gndr: masc, nmbr: sing, text: пятисотому}
+  - {case: datv, gndr: femn, nmbr: sing, text: пятисотой}
+  - {case: datv, gndr: neut, nmbr: sing, text: пятисотому}
+  - {case: datv, nmbr: plur, text: пятисотым}
+  - {case: accs, gndr: masc, nmbr: sing, text: пятисотого}
+  - {case: accs, gndr: masc, nmbr: sing, text: пятисотый}
+  - {case: accs, gndr: femn, nmbr: sing, text: пятисотую}
+  - {case: accs, gndr: neut, nmbr: sing, text: пятисотое}
+  - {case: accs, nmbr: plur, text: пятисотые}
+  - {case: accs, nmbr: plur, text: пятисотых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: пятисотым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: пятисотой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: пятисотым}
+  - {case: ablt, nmbr: plur, text: пятисотыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: пятисотом}
+  - {case: loct, gndr: femn, nmbr: sing, text: пятисотой}
+  - {case: loct, gndr: neut, nmbr: sing, text: пятисотом}
+  - {case: loct, nmbr: plur, text: пятисотых}
+  o:
+  - {case: nomn, text: пятьсот}
+  - {case: gent, text: пятисот}
+  - {case: datv, text: пятистам}
+  - {case: accs, text: пятьсот}
+  - {case: ablt, text: пятьюстами}
+  - {case: loct, text: пятистах}
+600:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: шестисотый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: шестисотая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: шестисотое}
+  - {case: nomn, nmbr: plur, text: шестисотые}
+  - {case: gent, gndr: masc, nmbr: sing, text: шестисотого}
+  - {case: gent, gndr: femn, nmbr: sing, text: шестисотой}
+  - {case: gent, gndr: neut, nmbr: sing, text: шестисотого}
+  - {case: gent, nmbr: plur, text: шестисотых}
+  - {case: datv, gndr: masc, nmbr: sing, text: шестисотому}
+  - {case: datv, gndr: femn, nmbr: sing, text: шестисотой}
+  - {case: datv, gndr: neut, nmbr: sing, text: шестисотому}
+  - {case: datv, nmbr: plur, text: шестисотым}
+  - {case: accs, gndr: masc, nmbr: sing, text: шестисотого}
+  - {case: accs, gndr: masc, nmbr: sing, text: шестисотый}
+  - {case: accs, gndr: femn, nmbr: sing, text: шестисотую}
+  - {case: accs, gndr: neut, nmbr: sing, text: шестисотое}
+  - {case: accs, nmbr: plur, text: шестисотые}
+  - {case: accs, nmbr: plur, text: шестисотых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: шестисотым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: шестисотой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: шестисотым}
+  - {case: ablt, nmbr: plur, text: шестисотыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: шестисотом}
+  - {case: loct, gndr: femn, nmbr: sing, text: шестисотой}
+  - {case: loct, gndr: neut, nmbr: sing, text: шестисотом}
+  - {case: loct, nmbr: plur, text: шестисотых}
+  o:
+  - {case: nomn, text: шестьсот}
+  - {case: gent, text: шестисот}
+  - {case: datv, text: шестистам}
+  - {case: accs, text: шестьсот}
+  - {case: ablt, text: шестьюстами}
+  - {case: loct, text: шестистах}
+700:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: семисотый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: семисотая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: семисотое}
+  - {case: nomn, nmbr: plur, text: семисотые}
+  - {case: gent, gndr: masc, nmbr: sing, text: семисотого}
+  - {case: gent, gndr: femn, nmbr: sing, text: семисотой}
+  - {case: gent, gndr: neut, nmbr: sing, text: семисотого}
+  - {case: gent, nmbr: plur, text: семисотых}
+  - {case: datv, gndr: masc, nmbr: sing, text: семисотому}
+  - {case: datv, gndr: femn, nmbr: sing, text: семисотой}
+  - {case: datv, gndr: neut, nmbr: sing, text: семисотому}
+  - {case: datv, nmbr: plur, text: семисотым}
+  - {case: accs, gndr: masc, nmbr: sing, text: семисотого}
+  - {case: accs, gndr: masc, nmbr: sing, text: семисотый}
+  - {case: accs, gndr: femn, nmbr: sing, text: семисотую}
+  - {case: accs, gndr: neut, nmbr: sing, text: семисотое}
+  - {case: accs, nmbr: plur, text: семисотые}
+  - {case: accs, nmbr: plur, text: семисотых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: семисотым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: семисотой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: семисотым}
+  - {case: ablt, nmbr: plur, text: семисотыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: семисотом}
+  - {case: loct, gndr: femn, nmbr: sing, text: семисотой}
+  - {case: loct, gndr: neut, nmbr: sing, text: семисотом}
+  - {case: loct, nmbr: plur, text: семисотых}
+  o:
+  - {case: nomn, text: семьсот}
+  - {case: gent, text: семисот}
+  - {case: datv, text: семистам}
+  - {case: accs, text: семьсот}
+  - {case: ablt, text: семьюстами}
+  - {case: loct, text: семистах}
+800:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: восьмисотый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: восьмисотая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: восьмисотое}
+  - {case: nomn, nmbr: plur, text: восьмисотые}
+  - {case: gent, gndr: masc, nmbr: sing, text: восьмисотого}
+  - {case: gent, gndr: femn, nmbr: sing, text: восьмисотой}
+  - {case: gent, gndr: neut, nmbr: sing, text: восьмисотого}
+  - {case: gent, nmbr: plur, text: восьмисотых}
+  - {case: datv, gndr: masc, nmbr: sing, text: восьмисотому}
+  - {case: datv, gndr: femn, nmbr: sing, text: восьмисотой}
+  - {case: datv, gndr: neut, nmbr: sing, text: восьмисотому}
+  - {case: datv, nmbr: plur, text: восьмисотым}
+  - {case: accs, gndr: masc, nmbr: sing, text: восьмисотого}
+  - {case: accs, gndr: masc, nmbr: sing, text: восьмисотый}
+  - {case: accs, gndr: femn, nmbr: sing, text: восьмисотую}
+  - {case: accs, gndr: neut, nmbr: sing, text: восьмисотое}
+  - {case: accs, nmbr: plur, text: восьмисотые}
+  - {case: accs, nmbr: plur, text: восьмисотых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: восьмисотым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: восьмисотой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: восьмисотым}
+  - {case: ablt, nmbr: plur, text: восьмисотыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: восьмисотом}
+  - {case: loct, gndr: femn, nmbr: sing, text: восьмисотой}
+  - {case: loct, gndr: neut, nmbr: sing, text: восьмисотом}
+  - {case: loct, nmbr: plur, text: восьмисотых}
+  o:
+  - {case: nomn, text: восемьсот}
+  - {case: gent, text: восьмисот}
+  - {case: datv, text: восьмистам}
+  - {case: accs, text: восемьсот}
+  - {case: ablt, text: восемьюстами}
+  - {case: loct, text: восьмистах}
+900:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: девятисотый}
+  - {case: nomn, gndr: femn, nmbr: sing, text: девятисотая}
+  - {case: nomn, gndr: neut, nmbr: sing, text: девятисотое}
+  - {case: nomn, nmbr: plur, text: девятисотые}
+  - {case: gent, gndr: masc, nmbr: sing, text: девятисотого}
+  - {case: gent, gndr: femn, nmbr: sing, text: девятисотой}
+  - {case: gent, gndr: neut, nmbr: sing, text: девятисотого}
+  - {case: gent, nmbr: plur, text: девятисотых}
+  - {case: datv, gndr: masc, nmbr: sing, text: девятисотому}
+  - {case: datv, gndr: femn, nmbr: sing, text: девятисотой}
+  - {case: datv, gndr: neut, nmbr: sing, text: девятисотому}
+  - {case: datv, nmbr: plur, text: девятисотым}
+  - {case: accs, gndr: masc, nmbr: sing, text: девятисотого}
+  - {case: accs, gndr: masc, nmbr: sing, text: девятисотый}
+  - {case: accs, gndr: femn, nmbr: sing, text: девятисотую}
+  - {case: accs, gndr: neut, nmbr: sing, text: девятисотое}
+  - {case: accs, nmbr: plur, text: девятисотые}
+  - {case: accs, nmbr: plur, text: девятисотых}
+  - {case: ablt, gndr: masc, nmbr: sing, text: девятисотым}
+  - {case: ablt, gndr: femn, nmbr: sing, text: девятисотой}
+  - {case: ablt, gndr: neut, nmbr: sing, text: девятисотым}
+  - {case: ablt, nmbr: plur, text: девятисотыми}
+  - {case: loct, gndr: masc, nmbr: sing, text: девятисотом}
+  - {case: loct, gndr: femn, nmbr: sing, text: девятисотой}
+  - {case: loct, gndr: neut, nmbr: sing, text: девятисотом}
+  - {case: loct, nmbr: plur, text: девятисотых}
+  o:
+  - {case: nomn, text: девятьсот}
+  - {case: gent, text: девятисот}
+  - {case: datv, text: девятистам}
+  - {case: accs, text: девятьсот}
+  - {case: ablt, text: девятьюстами}
+  - {case: loct, text: девятистах}
+1000:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: тысячный}
+  - {case: nomn, gndr: femn, nmbr: sing, text: тысячная}
+  - {case: nomn, gndr: neut, nmbr: sing, text: тысячное}
+  - {case: nomn, nmbr: plur, text: тысячные}
+  - {case: gent, gndr: masc, nmbr: sing, text: тысячного}
+  - {case: gent, gndr: femn, nmbr: sing, text: тысячной}
+  - {case: gent, gndr: neut, nmbr: sing, text: тысячного}
+  - {case: gent, nmbr: plur, text: тысячных}
+  - {case: datv, gndr: masc, nmbr: sing, text: тысячному}
+  - {case: datv, gndr: femn, nmbr: sing, text: тысячной}
+  - {case: datv, gndr: neut, nmbr: sing, text: тысячному}
+  - {case: datv, nmbr: plur, text: тысячным}
+  - {case: accs, gndr: masc, nmbr: sing, text: тысячного}
+  - {case: accs, gndr: masc, nmbr: sing, text: тысячный}
+  - {case: accs, gndr: femn, nmbr: sing, text: тысячную}
+  - {case: accs, gndr: neut, nmbr: sing, text: тысячное}
+  - {case: accs, nmbr: plur, text: тысячные}
+  - {case: accs, nmbr: plur, text: тысячных}
+  - {case: ablt, gndr: masc, nmbr: sing, text: тысячным}
+  - {case: ablt, gndr: femn, nmbr: sing, text: тысячной}
+  - {case: ablt, gndr: femn, nmbr: sing, text: тысячною}
+  - {case: ablt, gndr: neut, nmbr: sing, text: тысячным}
+  - {case: ablt, nmbr: plur, text: тысячными}
+  - {case: loct, gndr: masc, nmbr: sing, text: тысячном}
+  - {case: loct, gndr: femn, nmbr: sing, text: тысячной}
+  - {case: loct, gndr: neut, nmbr: sing, text: тысячном}
+  - {case: loct, nmbr: plur, text: тысячных}
+  o:
+  - {case: nomn, text: тысяча}
+  - {case: nomn, text: тысячи}
+  - {case: gent, text: тысяч}
+  - {case: gent, text: тысячи}
+  - {case: datv, text: тысячам}
+  - {case: datv, text: тысяче}
+  - {case: accs, text: тысячи}
+  - {case: accs, text: тысячу}
+  - {case: ablt, text: тысячами}
+  - {case: ablt, text: тысячей}
+  - {case: ablt, text: тысячью}
+  - {case: loct, text: тысячах}
+  - {case: loct, text: тысяче}
+1000000:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: миллионный}
+  - {case: nomn, gndr: femn, nmbr: sing, text: миллионная}
+  - {case: nomn, gndr: neut, nmbr: sing, text: миллионное}
+  - {case: nomn, nmbr: plur, text: миллионные}
+  - {case: gent, gndr: masc, nmbr: sing, text: миллионного}
+  - {case: gent, gndr: femn, nmbr: sing, text: миллионной}
+  - {case: gent, gndr: neut, nmbr: sing, text: миллионного}
+  - {case: gent, nmbr: plur, text: миллионных}
+  - {case: datv, gndr: masc, nmbr: sing, text: миллионному}
+  - {case: datv, gndr: femn, nmbr: sing, text: миллионной}
+  - {case: datv, gndr: neut, nmbr: sing, text: миллионному}
+  - {case: datv, nmbr: plur, text: миллионным}
+  - {case: accs, gndr: masc, nmbr: sing, text: миллионного}
+  - {case: accs, gndr: masc, nmbr: sing, text: миллионный}
+  - {case: accs, gndr: femn, nmbr: sing, text: миллионную}
+  - {case: accs, gndr: neut, nmbr: sing, text: миллионное}
+  - {case: accs, nmbr: plur, text: миллионные}
+  - {case: accs, nmbr: plur, text: миллионных}
+  - {case: ablt, gndr: masc, nmbr: sing, text: миллионным}
+  - {case: ablt, gndr: femn, nmbr: sing, text: миллионной}
+  - {case: ablt, gndr: femn, nmbr: sing, text: миллионною}
+  - {case: ablt, gndr: neut, nmbr: sing, text: миллионным}
+  - {case: ablt, nmbr: plur, text: миллионными}
+  - {case: loct, gndr: masc, nmbr: sing, text: миллионном}
+  - {case: loct, gndr: femn, nmbr: sing, text: миллионной}
+  - {case: loct, gndr: neut, nmbr: sing, text: миллионном}
+  - {case: loct, nmbr: plur, text: миллионных}
+  o:
+  - {case: nomn, text: миллион}
+  - {case: gent, text: миллиона}
+  - {case: gent, text: миллионов}
+  - {case: datv, text: миллионам}
+  - {case: datv, text: миллиону}
+  - {case: accs, text: миллион}
+  - {case: ablt, text: миллионами}
+  - {case: ablt, text: миллионом}
+  - {case: loct, text: миллионах}
+  - {case: loct, text: миллионе}
+
+1000000000:
+  p:
+  - {case: nomn, gndr: masc, nmbr: sing, text: миллиардный}
+  - {case: nomn, gndr: femn, nmbr: sing, text: миллиардная}
+  - {case: nomn, gndr: neut, nmbr: sing, text: миллиардное}
+  - {case: nomn, nmbr: plur, text: миллиардные}
+  - {case: gent, gndr: masc, nmbr: sing, text: миллиардного}
+  - {case: gent, gndr: femn, nmbr: sing, text: миллиардной}
+  - {case: gent, gndr: neut, nmbr: sing, text: миллиардного}
+  - {case: gent, nmbr: plur, text: миллиардным}
+  - {case: gent, nmbr: plur, text: миллиардных}
+  - {case: datv, gndr: masc, nmbr: sing, text: миллиардному}
+  - {case: datv, gndr: femn, nmbr: sing, text: миллиардной}
+  - {case: datv, gndr: neut, nmbr: sing, text: миллиардному}
+  - {case: accs, gndr: masc, nmbr: sing, text: миллиардного}
+  - {case: accs, gndr: femn, nmbr: sing, text: миллиардный}
+  - {case: accs, gndr: femn, nmbr: sing, text: миллиардную}
+  - {case: accs, gndr: neut, nmbr: sing, text: миллиардное}
+  - {case: accs, nmbr: plur, text: миллиардные}
+  - {case: accs, nmbr: plur, text: миллиардных}
+  - {case: ablt, gndr: masc, nmbr: sing, text: миллиардным}
+  - {case: ablt, gndr: femn, nmbr: sing, text: миллиардной}
+  - {case: ablt, gndr: femn, nmbr: sing, text: миллиардною}
+  - {case: ablt, gndr: neut, nmbr: sing, text: миллиардным}
+  - {case: ablt, nmbr: plur, text: миллиардными}
+  - {case: loct, gndr: masc, nmbr: sing, text: миллиардном}
+  - {case: loct, gndr: femn, nmbr: sing, text: миллиардной}
+  - {case: loct, gndr: neut, nmbr: sing, text: миллиардном}
+  - {case: loct, nmbr: plur, text: миллиардных}
+  o:
+  - {case: nomn, text: миллиард}
+  - {case: nomn, text: миллиарды}
+  - {case: gent, text: миллиарда}
+  - {case: gent, text: миллиардов}
+  - {case: datv, text: миллиардам}
+  - {case: datv, text: миллиарду}
+  - {case: accs, text: миллиард}
+  - {case: accs, text: миллиарды}
+  - {case: ablt, text: миллиардами}
+  - {case: ablt, text: миллиардом}
+  - {case: loct, text: миллиардах}
+  - {case: loct, text: миллиарде}

requirements.txt

@@ -0,0 +1,6 @@
+scikit-learn=0.20.3
+tqdm=4.31.1
+numpy==1.16.2
+PyYAML==5.1
+tensorflow-gpu==1.13.1
+lxml=4.3.3

tester.py

@@ -0,0 +1,256 @@
+import os, pickle
+import numpy as np
+import tensorflow as tf
+from tqdm import tqdm
+from model import RNN
+from utils import CONFIG, decode_word, load_datasets
+
+
+class Tester:
+    def __init__(self):
+        self.config = CONFIG
+        self.config['graph_part_configs']['lemm']['use_cls_placeholder'] = True
+        self.rnn = RNN(True)
+        self.chars = {c: index for index, c in enumerate(self.config['chars'])}
+        self.batch_size = 65536
+        self.show_bad_items = False
+
+    def test(self):
+        config = tf.ConfigProto(allow_soft_placement=True)
+        results = []
+        with tf.Session(config=config, graph=self.rnn.graph) as sess:
+            sess.run(tf.global_variables_initializer())
+            sess.run(tf.local_variables_initializer())
+            self.rnn.restore(sess)
+            for gram in self.rnn.gram_keys:
+                full_cls_acc, part_cls_acc, _ = self.__test_classification__(sess, gram, self.rnn.gram_graph_parts[gram], 'test')
+                result = f"{gram}. full_cls_acc: {full_cls_acc}; part_cls_acc: {part_cls_acc}"
+                results.append(result)
+                tqdm.write(result)
+
+            full_cls_acc, part_cls_acc, _ = self.__test_classification__(sess, 'main', self.rnn.main_graph_part, 'test')
+            result = f"main. full_cls_acc: {full_cls_acc}; part_cls_acc: {part_cls_acc}"
+            results.append(result)
+            tqdm.write(result)
+            lemm_acc, _ = self.__test_lemmas__(sess, 'test')
+            result = f"lemma_acc: {lemm_acc}"
+            tqdm.write(result)
+            results.append(result)
+            inflect_acc, _ = self.__test_inflect__(sess, 'test')
+            result = f"inflect_acc: {inflect_acc}"
+            tqdm.write(result)
+            results.append(result)
+            tqdm.write(result)
+
+        return "\n".join(results)
+
+    def __get_classification_items__(self, sess, items, graph_part):
+        wi = 0
+        pbar = tqdm(total=len(items), desc='Getting classification info')
+        results = []
+        etalon = []
+
+        while wi < len(items):
+            bi = 0
+            xs = []
+            indexes = []
+            seq_lens = []
+            max_len = 0
+
+            while bi < self.batch_size and wi < len(items):
+                word = items[wi]['src']
+                etalon.append(items[wi]['y'])
+                for c_index, char in enumerate(word):
+                    xs.append(self.chars[char] if char in self.chars else self.chars['UNDEFINED'])
+                    indexes.append([bi, c_index])
+                cur_len = len(word)
+                if cur_len > max_len:
+                    max_len = cur_len
+                seq_lens.append(cur_len)
+                bi += 1
+                wi += 1
+                pbar.update(1)
+
+            lnch = [graph_part.probs[0]]
+            nn_results = sess.run(
+                lnch,
+                {
+                    self.rnn.batch_size: bi,
+                    self.rnn.x_seq_lens[0]: np.asarray(seq_lens),
+                    self.rnn.x_vals[0]: np.asarray(xs),
+                    self.rnn.x_inds[0]: np.asarray(indexes),
+                    self.rnn.x_shape[0]: np.asarray([bi, max_len])
+                }
+            )
+            results.extend(nn_results[0])
+
+        return results, etalon
+
+    def __get_lemma_items__(self, sess, items):
+        wi = 0
+        pbar = tqdm(total=len(items))
+        while wi < len(items):
+            bi = 0
+            xs = []
+            clss = []
+            indexes = []
+            seq_lens = []
+            max_len = 0
+
+            while bi < self.batch_size and wi < len(items):
+                item = items[wi]
+                word = item['x_src']
+                x_cls = item['main_cls']
+                for c_index, char in enumerate(word):
+                    xs.append(self.chars[char])
+                    indexes.append([bi, c_index])
+                cur_len = len(word)
+                clss.append(x_cls)
+                if cur_len > max_len:
+                    max_len = cur_len
+                seq_lens.append(cur_len)
+                bi += 1
+                wi += 1
+                pbar.update(1)
+
+            lnch = [self.rnn.lem_result]
+            results = sess.run(
+                lnch,
+                {
+                    self.rnn.batch_size: bi,
+                    self.rnn.x_seq_lens[0]: np.asarray(seq_lens),
+                    self.rnn.x_vals[0]: np.asarray(xs),
+                    self.rnn.x_inds[0]: np.asarray(indexes),
+                    self.rnn.lem_class_pl: np.asarray(clss),
+                    self.rnn.x_shape[0]: np.asarray([bi, max_len])
+                }
+            )
+            for word_src in results[0]:
+                yield decode_word(word_src[0])
+
+    def __get_inflect_items__(self, sess, items):
+        wi = 0
+        pbar = tqdm(total=len(items))
+        while wi < len(items):
+            bi = 0
+            xs = []
+            x_clss = []
+            y_clss = []
+            indexes = []
+            seq_lens = []
+            max_len = 0
+
+            while bi < self.batch_size and wi < len(items):
+                item = items[wi]
+                word = item['x_src']
+                x_cls = item['x_cls']
+                y_cls = item['y_cls']
+                for c_index, char in enumerate(word):
+                    xs.append(self.chars[char])
+                    indexes.append([bi, c_index])
+                cur_len = len(word)
+                x_clss.append(x_cls)
+                y_clss.append(y_cls)
+                if cur_len > max_len:
+                    max_len = cur_len
+                seq_lens.append(cur_len)
+                bi += 1
+                wi += 1
+                pbar.update(1)
+
+            lnch = [self.rnn.inflect_graph_part.results[0]]
+            results = sess.run(
+                lnch,
+                {
+                    self.rnn.batch_size: bi,
+                    self.rnn.x_seq_lens[0]: np.asarray(seq_lens),
+                    self.rnn.x_vals[0]: np.asarray(xs),
+                    self.rnn.x_inds[0]: np.asarray(indexes),
+                    self.rnn.inflect_graph_part.x_cls[0]: np.asarray(x_clss),
+                    self.rnn.inflect_graph_part.y_cls[0]: np.asarray(y_clss),
+                    self.rnn.x_shape[0]: np.asarray([bi, max_len])
+                }
+            )
+
+            for word_src in results[0]:
+                yield decode_word(word_src)
+
+    def __test_classification__(self, sess, key, graph_part, *ds_types):
+        et_items = load_datasets(key, *ds_types)
+        results, etalon = self.__get_classification_items__(sess, et_items, graph_part)
+        total = len(etalon)
+        total_classes = 0
+        full_correct = 0
+        part_correct = 0
+        bad_items = []
+
+        for index, et in enumerate(etalon):
+            classes_count = et.sum()
+            good_classes = np.argwhere(et == 1).ravel()
+            rez_classes = np.argsort(results[index])[-classes_count:]
+
+            total_classes += classes_count
+            correct = True
+            for cls in rez_classes:
+                if cls in good_classes:
+                    part_correct += 1
+                else:
+                    correct = False
+
+            if correct:
+                full_correct += 1
+            else:
+                bad_items.append((et_items[index], rez_classes))
+
+        full_acc = full_correct / total
+        cls_correct = part_correct / total_classes
+        return full_acc, cls_correct, bad_items
+
+    def __test_lemmas__(self, sess, *ds_types):
+        good_items = load_datasets("lemma", *ds_types)
+        good_items = [
+            word
+            for word in good_items
+            if all([c in self.config['chars'] for c in word['x_src']])
+        ]
+        results = list(self.__get_lemma_items__(sess, good_items))
+
+        bad_words = []
+        total = len(good_items)
+        wrong = 0
+        for index, rez in enumerate(results):
+            et_word = good_items[index]
+            if rez != et_word['y_src']:
+                wrong += 1
+                bad_words.append((et_word, rez))
+
+        correct = total - wrong
+        acc = correct / total
+        return acc, bad_words
+
+    def __test_inflect__(self, sess, *ds_types):
+        good_items = load_datasets("inflect", *ds_types)
+        good_items = [
+            word
+            for word in good_items
+            if all([c in self.config['chars'] for c in word['x_src']])
+        ]
+
+        bad_items = []
+        rez_words = list(self.__get_inflect_items__(sess, good_items))
+        total = len(good_items)
+        wrong = 0
+        for index, rez in enumerate(rez_words):
+            et_word = good_items[index]
+            if rez != et_word['y_src']:
+                wrong += 1
+                bad_items.append((et_word, rez))
+
+        correct = total - wrong
+        acc = correct / total
+        return acc, bad_items
+
+
+if __name__ == "__main__":
+    tester = Tester()
+    tester.test()

tf_utils.py

@@ -0,0 +1,495 @@
+import os
+import pickle
+import numpy as np
+import tensorflow as tf
+
+
+def rnn_top(scope_name,
+             rnn_rez,
+             settings,
+             classes_number):
+    with tf.variable_scope(scope_name, reuse=tf.AUTO_REUSE) as scope:
+        rez_size = settings['rnn_state_size']
+        w_softmax = tf.get_variable("W", (rez_size, classes_number))
+        b_softmax = tf.get_variable("b", [classes_number])
+        logits = tf.matmul(rnn_rez, w_softmax) + b_softmax
+
+    return logits
+
+
+def rnn_cell_unit(settings, for_usage, keep_drop=None):
+
+    cell = tf.contrib.rnn.GRUCell(
+        num_units=settings['rnn_state_size']
+    )
+    if not for_usage and keep_drop is not None:
+        cell = tf.contrib.rnn.DropoutWrapper(
+            cell,
+            input_keep_prob = keep_drop
+        )
+    if 'use_residual' in settings and settings['use_residual']:
+        cell = tf.nn.rnn_cell.ResidualWrapper(cell)
+
+    return cell
+
+
+def rnn_cell(settings, for_usage, keep_drop=None):
+    if settings['rnn_layers_count'] > 1:
+        cells = []
+        for i in range(settings['rnn_layers_count']):
+            with tf.variable_scope('RnnUnit_%s' % i, reuse=tf.AUTO_REUSE) as scope:
+                cell = rnn_cell_unit(settings, for_usage, keep_drop)
+            cells.append(cell)
+        cell = tf.nn.rnn_cell.MultiRNNCell(cells)
+    else:
+        with tf.variable_scope('RnnUnit', reuse=tf.AUTO_REUSE) as scope:
+            cell = rnn_cell_unit(settings, for_usage, keep_drop)
+
+    return cell
+
+
+def build_rnn(rnn_input, keep_drop, seq_len, settings, initial_state_fw=None, initial_state_bw=None, for_usage=False, with_seq=False, top_concat=False):
+    if settings['rnn_layers_count'] > 1 and initial_state_fw is not None:
+        initial_state_fw = tuple([initial_state_fw for i in range(settings['rnn_layers_count'])])
+
+    if settings['rnn_layers_count'] > 1 and initial_state_bw is not None:
+        initial_state_bw = tuple([initial_state_bw for i in range(settings['rnn_layers_count'])])
+
+    if settings['rnn_bidirectional']:
+        with tf.variable_scope('Rnn', reuse=tf.AUTO_REUSE) as scope:
+            with tf.variable_scope('FCell', reuse=tf.AUTO_REUSE) as scope:
+                fw_cell = rnn_cell(settings, for_usage, keep_drop)
+            with tf.variable_scope('BCell', reuse=tf.AUTO_REUSE) as scope:
+                bw_cell = rnn_cell(settings, for_usage, keep_drop)
+            seq_val, (final_fw, final_bw) = tf.nn.bidirectional_dynamic_rnn(cell_fw=fw_cell,
+                                                                   cell_bw=bw_cell,
+                                                                   sequence_length=seq_len,
+                                                                   initial_state_fw = initial_state_fw,
+                                                                   initial_state_bw = initial_state_bw,
+                                                                   inputs=rnn_input,
+                                                                   dtype=tf.float32)
+
+            if settings['rnn_layers_count'] > 1:
+                final_fw = final_fw[-1]
+                final_bw = final_bw[-1]
+
+            if top_concat:
+                final_state = tf.concat([final_fw, final_bw], axis=1)
+                final_state = tf.layers.dense(final_state, settings['rnn_state_size'])
+                seq_val = tf.concat([seq_val[0], seq_val[1]], axis=2)
+                seq_val = tf.layers.dense(seq_val, settings['rnn_state_size'])
+            else:
+                final_state = tf.add(final_fw, final_bw)
+                seq_val = tf.add(seq_val[0], seq_val[1])
+
+            if with_seq:
+                return final_state, seq_val
+            else:
+                return final_state
+
+    else:
+        with tf.variable_scope('Rnn', reuse=tf.AUTO_REUSE) as scope:
+            cell = rnn_cell(settings, for_usage, keep_drop)
+            rnn_rez, final_state = tf.nn.dynamic_rnn(cell=cell,
+                                                     sequence_length=seq_len,
+                                                     initial_state=initial_state_fw,
+                                                     inputs=rnn_input,
+                                                     dtype=tf.float32)
+            return final_state[-1]
+
+
+def average_gradients(tower_grads):
+    average_grads = []
+    for grad_and_vars in zip(*tower_grads):
+
+        if any(x for x in grad_and_vars if x[0] is None):
+            continue
+
+        # Note that each grad_and_vars looks like the following:
+        #   ((grad0_gpu0, var0_gpu0), ... , (grad0_gpuN, var0_gpuN))
+        grads = []
+        for g, _ in grad_and_vars:
+            # Add 0 dimension to the gradients to represent the tower.
+            expanded_g = tf.expand_dims(g, 0)
+
+            # Append on a 'tower' dimension which we will average over below.
+            grads.append(expanded_g)
+
+        # Average over the 'tower' dimension.
+        grad = tf.concat(grads, 0)
+        grad = tf.reduce_mean(grad, 0)
+
+        # Keep in mind that the Variables are redundant because they are shared
+        # across towers. So .. we will just return the first tower's pointer to
+        # the Variable.
+        v = grad_and_vars[0][1]
+        grad_and_var = (grad, v)
+        average_grads.append(grad_and_var)
+
+    return average_grads
+
+
+def create_reset_metric(metric, scope='reset_metrics', *args, **metric_args):
+    """
+    Source: https://github.com/tensorflow/tensorflow/issues/4814#issuecomment-314801758
+
+    Usage:
+
+    epoch_loss, epoch_loss_update, epoch_loss_reset = create_reset_metric(
+    tf.contrib.metrics.streaming_mean_squared_error, 'epoch_loss',
+    predictions=output, labels=target)
+
+    :param scope:
+    :param metric_args:
+    :return:
+    """
+    with tf.variable_scope(scope) as scope:
+        metric_op, update_op = metric(*args, **metric_args)
+        vars = tf.contrib.framework.get_variables(
+            scope, collection=tf.GraphKeys.LOCAL_VARIABLES)
+        reset_op = tf.variables_initializer(vars)
+    return metric_op, update_op, reset_op
+
+
+def load_cls_dataset(dataset_path, devices_count, type, batch_size, use_weights, gram="main"):
+
+    path = os.path.join(dataset_path, f"{gram}_{type}_dataset.pkl")
+    with open(path, 'rb') as f:
+        items = pickle.load(f)
+
+    batches = [items[i:i + batch_size] for i in range(0, len(items), batch_size)]
+    cur_step = []
+    for batch in batches:
+        x = np.stack([item['x'][0] for item in batch])
+        seq_len = np.asarray([item['x'][1] for item in batch], np.int)
+        y = np.asarray([item['y'] for item in batch])
+        weight = np.asarray([item['weight'] for item in batch]) if use_weights else [1 for _ in range(batch)]
+        if len(cur_step) == devices_count:
+            yield cur_step
+            cur_step = []
+
+        cur_step.append(dict(
+            x=x,
+            x_seq_len=seq_len,
+            y=y,
+            weight=weight
+        ))
+
+    if len(cur_step)==devices_count:
+        yield cur_step
+
+
+def load_lemma_dataset(dataset_path, devices_count, type, batch_size):
+    path = os.path.join(dataset_path, f"lemma_{type}_dataset.pkl")
+    with open(path, 'rb') as f:
+        items = pickle.load(f)
+
+    batches = [items[i:i + batch_size] for i in range(0, len(items), batch_size)]
+    cur_step = []
+    for batch in batches:
+        x = np.stack([item['x'] for item in batch])
+        x_seq_len = np.asarray([item['x_len'] for item in batch], np.int)
+        x_cls = np.asarray([item['main_cls'] for item in batch], np.int)
+        y_seq_len = np.asarray([item['y_len'] for item in batch], np.int)
+        max_len = y_seq_len.max()
+        y = np.asarray([item['y'][:max_len] for item in batch])
+
+        x_src = [item['x_src'] for item in batch]
+        y_src = [item['y_src'] for item in batch]
+
+        if len(cur_step) == devices_count:
+            yield cur_step
+            cur_step = []
+
+        cur_step.append(dict(
+            x=x,
+            x_seq_len=x_seq_len,
+            x_cls=x_cls,
+            y=y,
+            y_seq_len=y_seq_len,
+            x_src=x_src,
+            y_src=y_src
+        ))
+
+    if len(cur_step) == devices_count and all([len(step['x']) == batch_size for step in cur_step]):
+        yield cur_step
+
+
+def load_inflect_dataset(dataset_path, devices_count, type, batch_size):
+    path = os.path.join(dataset_path, f"inflect_{type}_dataset.pkl")
+    with open(path, 'rb') as f:
+        items = pickle.load(f)
+
+    batches = [items[i:i + batch_size] for i in range(0, len(items), batch_size)]
+    cur_step = []
+    for batch in batches:
+        x = np.stack([item['x'] for item in batch])
+        x_seq_len = np.asarray([item['x_len'] for item in batch], np.int)
+        x_cls = np.asarray([item['x_cls'] for item in batch], np.int)
+        y_seq_len = np.asarray([item['y_len'] for item in batch], np.int)
+        y_cls = np.asarray([item['y_cls'] for item in batch], np.int)
+        max_len = y_seq_len.max()
+        y = np.asarray([item['y'][:max_len] for item in batch])
+        x_src = [item['x_src'] for item in batch]
+        y_src = [item['y_src'] for item in batch]
+
+        if len(cur_step) == devices_count:
+            yield cur_step
+            cur_step = []
+
+        cur_step.append(dict(
+            x=x,
+            x_seq_len=x_seq_len,
+            x_cls=x_cls,
+            y=y,
+            y_seq_len=y_seq_len,
+            y_cls=y_cls,
+            x_src=x_src,
+            y_src=y_src
+        ))
+
+    if len(cur_step) == devices_count and all([len(step['x']) == batch_size for step in cur_step]):
+        yield cur_step
+
+
+def seq2seq(graph_part,
+            batch_size,
+            x,
+            x_init,
+            x_seq_len,
+            y,
+            y_init,
+            y_seq_len):
+    start_tokens = tf.fill([batch_size], graph_part.start_char_index)
+    initializer = tf.random_normal_initializer()
+    y_seq_len += 1
+
+    with tf.variable_scope("Encoder", reuse=tf.AUTO_REUSE):
+        encoder_char_embeddings = tf.get_variable(
+            "CharEmbeddings",
+            [graph_part.chars_count, graph_part.settings['char_vector_size']],
+            initializer=initializer
+        )
+
+        encoder_init_state = ClsGramEmbedder(graph_part.main_cls_dic,
+                                          graph_part.settings['encoder']['gram_vector_size'],
+                                          graph_part.settings['encoder']['ad_cls_vector_size'],
+                                          graph_part.settings['encoder']['rnn_state_size'])(x_init, batch_size)
+        encoder_input = tf.nn.embedding_lookup(encoder_char_embeddings, x)
+
+    with tf.variable_scope("Decoder", reuse=tf.AUTO_REUSE):
+        decoder_char_embeddings = tf.get_variable(
+            "CharEmbeddings",
+            [graph_part.chars_count, graph_part.settings['char_vector_size']],
+            initializer=initializer
+        )
+        decoder_init_state = ClsGramEmbedder(graph_part.main_cls_dic,
+                                          graph_part.settings['decoder']['gram_vector_size'],
+                                          graph_part.settings['decoder']['ad_cls_vector_size'],
+                                          graph_part.settings['decoder']['rnn_state_size'])(y_init, batch_size)
+        decoder_output = tf.nn.embedding_lookup(decoder_char_embeddings, y)
+
+    if graph_part.for_usage:
+        keep_drop = tf.constant(1, dtype=tf.float32, name='KeepDrop')
+        decoder_keep_drop = tf.constant(1, dtype=tf.float32, name='DecoderKeepDrop')
+    else:
+        keep_drop = tf.placeholder(dtype=tf.float32, name='KeepDrop')
+        decoder_keep_drop = tf.placeholder(dtype=tf.float32, name='DecoderKeepDrop')
+
+    with tf.variable_scope('Encoder', reuse=tf.AUTO_REUSE) as scope:
+        _, encoder_output = build_rnn(
+            encoder_input,
+            keep_drop,
+            x_seq_len,
+            graph_part.settings['encoder'],
+            encoder_init_state,
+            encoder_init_state,
+            top_concat=False,
+            for_usage=graph_part.for_usage,
+            with_seq=True
+        )
+
+    with tf.variable_scope('Decoder', reuse=tf.AUTO_REUSE) as scope:
+        if not graph_part.for_usage:
+            start_tokens_emd = tf.nn.embedding_lookup(decoder_char_embeddings, start_tokens)
+            start_tokens_emd = tf.reshape(start_tokens_emd, (batch_size, -1, graph_part.settings['char_vector_size']))
+            decoder_output = tf.concat(values=[start_tokens_emd, decoder_output], axis=1)
+
+            end_tokens = tf.fill([batch_size], graph_part.end_char_index)
+            end_tokens_emd = tf.nn.embedding_lookup(decoder_char_embeddings, end_tokens)
+            end_tokens_emd = tf.reshape(end_tokens_emd, (batch_size, -1, graph_part.settings['char_vector_size']))
+            decoder_output = tf.concat([decoder_output, end_tokens_emd], axis=1)
+
+            end_tokens = tf.reshape(end_tokens, (batch_size, 1))
+            y = tf.concat([y, end_tokens], axis=1)
+
+        if graph_part.for_usage:
+            helper = tf.contrib.seq2seq.GreedyEmbeddingHelper(decoder_char_embeddings,
+                                                              start_tokens=start_tokens,
+                                                              end_token=graph_part.end_char_index)
+        else:
+            helper = tf.contrib.seq2seq.TrainingHelper(decoder_output, y_seq_len)
+
+        attention_mechanism = tf.contrib.seq2seq.BahdanauAttention(
+            num_units=graph_part.settings['decoder']['rnn_state_size'],
+            memory=encoder_output,
+            memory_sequence_length=x_seq_len,
+            normalize=False
+        )
+
+        cell = rnn_cell(graph_part.settings['decoder'],
+                            graph_part.for_usage,
+                            decoder_keep_drop)
+
+        cell = tf.contrib.seq2seq.AttentionWrapper(
+            cell,
+            attention_mechanism,
+            attention_layer_size=graph_part.settings['decoder']['rnn_state_size'] / 2
+        )
+
+        out_cell = tf.contrib.rnn.OutputProjectionWrapper(
+            cell,
+            graph_part.chars_count
+        )
+
+        init_state = cell.zero_state(dtype=tf.float32, batch_size=batch_size).clone(cell_state=decoder_init_state)
+
+        decoder = tf.contrib.seq2seq.BasicDecoder(
+            cell=out_cell,
+            helper=helper,
+            initial_state=init_state
+        )
+
+        max_len = graph_part.settings['max_length'] if graph_part.for_usage else tf.reduce_max(y_seq_len)
+        outputs = tf.contrib.seq2seq.dynamic_decode(
+            decoder=decoder,
+            impute_finished=True,
+            output_time_major=False,
+            maximum_iterations=max_len
+        )
+        decoder_ids = outputs[0].sample_id
+
+        if not graph_part.for_usage:
+            decoder_logits = outputs[0].rnn_output
+            masks = tf.sequence_mask(
+                lengths=y_seq_len,
+                dtype=tf.float32,
+                maxlen=tf.reduce_max(y_seq_len)
+            )
+            seq_mask_int = tf.cast(masks, tf.int32)
+            seq_mask_flat = tf.cast(tf.reshape(masks, (-1,)), tf.int32)
+
+            # seq loss
+            loss = tf.contrib.seq2seq.sequence_loss(
+                decoder_logits,
+                y,
+                masks,
+                name="SeqLoss"
+            )
+            vars = tf.get_collection(tf.GraphKeys.GLOBAL_VARIABLES, scope=graph_part.main_scope_name)
+            grads = graph_part.optimiser.compute_gradients(loss, var_list=vars)
+            graph_part.create_mean_metric(0, loss)
+
+            labels_flat = tf.reshape(y, (-1,))
+            predictions_flat = tf.reshape(decoder_ids, (-1,))
+            # char accuracy
+            nonzero_indices = tf.where(tf.not_equal(seq_mask_flat, 0))
+            labels_flat = tf.gather(labels_flat, nonzero_indices)
+            labels_flat = tf.reshape(labels_flat, (-1,))
+            predictions_flat = tf.gather(predictions_flat, nonzero_indices)
+            predictions_flat = tf.reshape(predictions_flat, (-1,))
+
+            # remove -1 items where no sampling took place
+            sample_indexes = tf.where(tf.not_equal(predictions_flat, -1))
+            labels_flat = tf.gather(labels_flat, sample_indexes)
+            predictions_flat = tf.gather(predictions_flat, sample_indexes)
+
+            graph_part.create_accuracy_metric(1, labels_flat, predictions_flat)
+
+            # seq accuracy
+            labels = y * seq_mask_int
+            predictions = decoder_ids * seq_mask_int
+            labels_flat = tf.reshape(labels, (-1,))
+            predictions_flat = tf.reshape(predictions, (-1,))
+            sample_zeros = tf.cast(tf.not_equal(predictions_flat, -1), tf.int32)
+            predictions = predictions_flat * sample_zeros
+            labels = labels_flat * sample_zeros
+            predictions = tf.reshape(predictions, (batch_size, max_len))
+            labels = tf.reshape(labels, (batch_size, max_len))
+            delta = labels - predictions
+            labels = tf.reduce_sum(delta * delta, 1)
+            predictions = tf.zeros(batch_size)
+            graph_part.create_accuracy_metric(2, labels, predictions)
+
+            graph_part.dev_grads.append(grads)
+            graph_part.losses.append(loss)
+            graph_part.keep_drops.append(keep_drop)
+            graph_part.decoder_keep_drops.append(decoder_keep_drop)
+
+        graph_part.results.append(decoder_ids)
+
+
+class ClsGramEmbedder:
+    def __init__(self, cls_dic, gram_vector_size, ad_cls_vector_size, rez_size):
+        gram_rez_dict = {}
+        tpls = sorted([(key, cls_dic[key]) for key in cls_dic], key=lambda x: x[1])
+        cls_vectors = []
+        for cls_key, cls_index in tpls:
+            cls_vector = []
+            for gram, gram_index in enumerate(list(cls_key)):
+                gram_key = (gram_index, gram)
+                if gram_key not in gram_rez_dict:
+                    gram_rez_dict[gram_key] = len(gram_rez_dict)
+
+                cls_vector.append(gram_rez_dict[gram_key])
+
+            cls_vectors.append(cls_vector)
+
+        tpls = sorted([(key, gram_rez_dict[key]) for key in gram_rez_dict], key=lambda x: x[1])
+        gram_vectors = []
+        for gram_key, gram_index in tpls:
+            if gram_key[0] is None:
+                val = np.zeros(gram_vector_size, dtype=np.float32)
+            else:
+                val = np.random.rand(gram_vector_size).astype(np.float32)
+
+            gram_vectors.append(val)
+
+        self.rez_size = rez_size
+        self.classes_count = len(cls_dic)
+        self.cls_vectors = np.asarray(cls_vectors, dtype=np.int)
+
+        self.grams_count = len(gram_vectors)
+        self.gram_vector_size = gram_vector_size
+        self.gram_vectors = np.stack(gram_vectors)
+
+        self.ad_cls_vector_size = ad_cls_vector_size
+
+    def __call__(self, cls_pl, batch_size):
+        # [self.grams_count, self.gram_vector_size],
+        gram_embeddings = tf.get_variable(
+            "GramEmbeddings",
+            initializer=tf.constant(self.gram_vectors),
+            dtype=tf.float32
+        )
+        # [self.classes_count, self.grams_count],
+        cls_embeddings = tf.get_variable(
+            "ClsEmbeddings",
+            initializer=tf.constant(self.cls_vectors),
+            dtype=tf.int64
+        )
+
+        gram_rez = tf.nn.embedding_lookup(cls_embeddings, cls_pl)
+        gram_rez = tf.reshape(gram_rez, (-1, ))
+        gram_rez = tf.nn.embedding_lookup(gram_embeddings, gram_rez)
+        gram_rez = tf.reshape(gram_rez, (batch_size, -1))
+
+        ad_cls_embeddings = tf.get_variable(
+            "AdClsEmbeddings",
+            [self.classes_count, self.ad_cls_vector_size],
+            initializer=tf.contrib.layers.xavier_initializer()
+        )
+        ad_cls_rez = tf.nn.embedding_lookup(ad_cls_embeddings, cls_pl)
+
+        result = tf.concat([gram_rez, ad_cls_rez], axis=1)
+        return result

utils.py

@@ -0,0 +1,154 @@
+import os
+import yaml
+import random
+import pickle
+import logging
+from collections import defaultdict
+from tqdm import tqdm
+
+
+def _get_config():
+    with open('config.yml', 'r') as f:
+        config = yaml.load(f)
+
+    classes_path = config['cls_classes_path']
+    if os.path.exists(classes_path):
+        with open(classes_path, 'rb') as f:
+            config['main_classes'] = pickle.load(f)
+            config['main_classes_count'] = len(config['main_classes'])
+
+    return config
+
+
+CONFIG = _get_config()
+logging.basicConfig(
+    level=logging.INFO,
+    format='%(asctime)s %(levelname)s: %(message)s',
+    datefmt='%d.%m.%Y %H:%M:%S'
+)
+DATASET_PATH = CONFIG['dataset_path']
+DICS_PATH = CONFIG['dics_path']
+RANDOM = random.Random(CONFIG['random_seed'])
+GRAMMEMES_TYPES = CONFIG['grammemes_types']
+
+
+class MyDefaultDict(defaultdict):
+    def __missing__(self, key):
+
+        if self.default_factory is None:
+            raise KeyError( key )
+        else:
+            ret = self[key] = self.default_factory(key)
+            return ret
+
+
+def get_grams_info(config):
+    dict_post_types = config['dict_post_types']
+    grammemes_types = config['grammemes_types']
+    src_convert = {}
+    classes_indexes = {}
+
+    for gram_key in grammemes_types:
+        gram = grammemes_types[gram_key]
+        cls_dic = gram['classes']
+        classes_indexes[gram_key] = {}
+        for cls_key in cls_dic:
+            cls_obj = cls_dic[cls_key]
+            classes_indexes[gram_key][cls_key] = cls_obj['index']
+
+            for key in cls_obj['keys']:
+                src_convert[key.lower()] = (gram_key, cls_key)
+
+    for post_key in dict_post_types:
+        cls_obj = dict_post_types[post_key.lower()]
+        p_key = ('post', post_key.lower())
+        for key in cls_obj['keys']:
+            src_convert[key.lower()] = p_key
+
+    return src_convert, classes_indexes
+
+
+def decode_word(vect_mas):
+    conf = CONFIG
+    word = []
+    for ci in vect_mas:
+        if ci == conf['end_token']:
+            return "".join(word)
+        elif ci < len((conf['chars'])):
+            word.append(conf['chars'][ci])
+        else:
+            word.append("0")
+
+    return "".join(word)
+
+
+def select_uniform_items(items_dict, persent, ds_info):
+    for cls in tqdm(items_dict, desc=f"Selecting {ds_info} dataset"):
+        i = 0
+        items = items_dict[cls]
+        per_group_count = persent * len(items) / 100
+
+        while i <= per_group_count and len(items) > 0:
+            item = items[0]
+            items.remove(item)
+            yield item
+            i += 1
+
+
+def save_dataset(items_dict, file_prefix):
+    if not os.path.isdir(DATASET_PATH):
+        os.mkdir(DATASET_PATH)
+
+    total_count = sum([len(items_dict[key]) for key in items_dict])
+    logging.info(f"Class '{file_prefix}': {total_count}")
+    for key in tqdm(items_dict, desc=f"Shuffling {file_prefix} items"):
+        random.shuffle(items_dict[key])
+
+    test_items = list(select_uniform_items(items_dict, CONFIG['test_persent'], f"test {file_prefix}"))
+    valid_items = list(select_uniform_items(items_dict, CONFIG['validation_persent'], f"valid {file_prefix}"))
+    items = []
+    for key in items_dict:
+        items.extend(items_dict[key])
+    random.shuffle(items)
+
+    logging.info(f"Saving '{file_prefix}' train dataset")
+    with open(os.path.join(DATASET_PATH, f"{file_prefix}_train_dataset.pkl"), 'wb+') as f:
+        pickle.dump(items, f)
+
+    logging.info(f"Saving '{file_prefix}' valid dataset")
+    with open(os.path.join(DATASET_PATH, f"{file_prefix}_valid_dataset.pkl"), 'wb+') as f:
+        pickle.dump(valid_items, f)
+
+    logging.info(f"Saving '{file_prefix}' test dataset")
+    with open(os.path.join(DATASET_PATH, f"{file_prefix}_test_dataset.pkl"), 'wb+') as f:
+        pickle.dump(test_items, f)
+
+
+def get_dict_path(file_prefix):
+    return os.path.join(DICS_PATH, f"{file_prefix}_dict_items.pkl")
+
+
+def save_dictionary_items(items, file_prefix):
+    with open(get_dict_path(file_prefix), 'wb+') as f:
+        pickle.dump(items, f)
+
+
+def create_cls_tuple(item):
+    return tuple(
+        item[key] if key in item else None
+        for key in GRAMMEMES_TYPES
+    )
+
+
+def load_datasets(main_type, *ds_type):
+    words = []
+
+    def load_words(type):
+        path = os.path.join(CONFIG['dataset_path'], f"{main_type}_{type}_dataset.pkl")
+        with open(path, 'rb') as f:
+            words.extend(pickle.load(f))
+
+    for key in ds_type:
+        load_words(key)
+
+    return words