коммуникация с сервером без очищения от лишних сеансов

.gitignore

@@ -0,0 +1,143 @@
+# ---> Python
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+# C extensions
+*.so
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+# Translations
+*.mo
+*.pot
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+# Flask stuff:
+instance/
+.webassets-cache
+# Scrapy stuff:
+.scrapy
+# Sphinx documentation
+docs/_build/
+# PyBuilder
+.pybuilder/
+target/
+# Jupyter Notebook
+.ipynb_checkpoints
+# IPython
+profile_default/
+ipython_config.py
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow
+__pypackages__/
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+# SageMath parsed files
+*.sage.py
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+# Spyder project settings
+.spyderproject
+.spyproject
+# Rope project settings
+.ropeproject
+# mkdocs documentation
+/site
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+# Pyre type checker
+.pyre/
+# pytype static type analyzer
+.pytype/
+# Cython debug symbols
+cython_debug/
+# PyCharm
+#  JetBrains specific template is maintainted in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+/gyno_videos/
+/result_gyno/
+/images/
+/test/
+/Ubuntu_CFG/
+/distPirogovDesktop/
+app_config.json
+color_profile.json
+users.json
+data/2024-08-22_00-25-03-343367.png
+data/2024-08-22_00-58-24-039690.png
+data/2024-08-22_00-59-26.webm
+data/2024-08-22_00-59-34-870157.png
+data/2024-08-22_00-59-50-575029.png
+data/2024-08-22_00-59-06-827412.png
+data/2024-08-22_00-29-12-759857.png
+/plugins/
+.idea/

ai/d.py

@@ -0,0 +1,96 @@
+import os
+import time
+import recognition
+from cleantemp import clean
+from datetime import datetime
+import paramiko
+path = '/home/dev/ExampleApp/audiofiles/'
+#n = os.listdir(path)
+
+def connect_to_sftp(hostname, username, password, remote_path):
+    try:
+        ssh = paramiko.SSHClient()
+        ssh.set_missing_host_key_policy(paramiko.AutoAddPolicy())  # Добавляет неизвестный хост в известные
+        ssh.connect(hostname, username=username, password=password)
+        sftp = ssh.open_sftp()
+        sftp.chdir(remote_path)  # Переходим в указанную директорию.  Обработка ошибок ниже.
+        return sftp
+    except paramiko.AuthenticationException:
+        print("Ошибка аутентификации. Проверьте имя пользователя и пароль.")
+        return None
+    except paramiko.SSHException as e:
+        print(f"Ошибка SSH: {e}")
+        return None
+    except FileNotFoundError:
+        print(f"{str(datetime.now())[:-7]} - директория {remote_path} ещё не создана. Повторная попытка через 30 минут.")
+        return None
+    except IOError as e:
+        print(f"Ошибка ввода-вывода: {e}")
+        return None
+
+
+remote_path = '/2025/02/'
+'''for i in range(24, 26):
+    sftp_client = connect_to_sftp("192.168.1.150", "monitor", "Audio4analy6!6", remote_path + str(i))
+    files = sftp_client.listdir()
+    temp=remote_path+str(i)+'/'
+    for file in files:
+        sftp_client.get(temp+file,f"/home/dev/ExampleApp/audiofiles{temp}{file}")'''
+
+
+
+
+
+def is_file_fully_loaded(file_path, name):
+    print(name)
+    try:
+        sftp_client = connect_to_sftp("192.168.1.150", "monitor", "Audio4analy6!6", file_path)
+        info = sftp_client.stat(name)
+        size = info.st_size
+        time.sleep(1)
+        sftp_client = connect_to_sftp("192.168.1.150", "monitor", "Audio4analy6!6", file_path)
+        info = sftp_client.stat(name)
+        new_size = info.st_size
+
+        if size != new_size: return False
+        else: return True
+    except FileNotFoundError: return False
+
+
+
+
+while True:
+    time.sleep(1)
+    remote_path = f"/{str(datetime.now())[:10].replace('-', '/')}/"
+    try: os.makedirs('/home/dev/ExampleApp/textfiles'+remote_path)
+    except FileExistsError: print('')
+    try:
+        sftp_client = connect_to_sftp("192.168.1.150", "monitor", "Audio4analy6!6", remote_path)
+        files = sftp_client.listdir()
+        print('Файлов на сервере за сегодня -', len(files))
+        texts = os.listdir('/home/dev/ExampleApp/textfiles'+remote_path)
+        texts=[i[:-4] for i in texts]
+        while len(set(files)-set(texts))==0:
+            print(datetime.now())
+            time.sleep(60)
+            sftp_client = connect_to_sftp("192.168.1.150", "monitor", "Audio4analy6!6", remote_path)
+            files=sftp_client.listdir()
+        for i in list(set(files) - set(texts)):
+            while not is_file_fully_loaded(remote_path,i):
+                time.sleep(5)
+            info = sftp_client.stat(i)
+            size = info.st_size
+            if size>=1024:
+                try:
+                    sftp_client.get(remote_path+i,'/home/dev/ExampleApp/temp/'+i)
+                    recognition.to_txt_wfp('/home/dev/ExampleApp/temp/'+i,i,f'/home/dev/ExampleApp/textfiles{remote_path}')
+                    os.remove('/home/dev/ExampleApp/temp/'+i)
+                except IndexError:
+                    fin = open(f'/home/dev/ExampleApp/textfiles/{remote_path}/{i}.txt', 'w')
+                    fin.write('Аудиозапись пуста (IndexError).')
+                    fin.close()
+            else:
+                fin=open(f'/home/dev/ExampleApp/textfiles/{remote_path}/{i}.txt','w')
+                fin.write('Аудиозапись пуста.')
+                fin.close()
+    except AttributeError: time.sleep(1800)

ai/old_calls_rec.py

@@ -0,0 +1,54 @@
+from rec_all_for import rec_all_from
+import os
+import paramiko
+
+'''def connect_to_sftp(hostname, username, password, remote_path):
+    try:
+        ssh = paramiko.SSHClient()
+        ssh.set_missing_host_key_policy(paramiko.AutoAddPolicy())  # Добавляет неизвестный хост в известные
+        ssh.connect(hostname, username=username, password=password)
+        sftp = ssh.open_sftp()
+        sftp.chdir(remote_path)  # Переходим в указанную директорию.  Обработка ошибок ниже.
+        return sftp
+    except paramiko.AuthenticationException:
+        print("Ошибка аутентификации. Проверьте имя пользователя и пароль.")
+        return None
+    except paramiko.SSHException as e:
+        print(f"Ошибка SSH: {e}")
+        return None
+    except FileNotFoundError:
+        print(f"Директория {remote_path} не найдена на сервере.")
+        return None
+    except IOError as e:
+        print(f"Ошибка ввода-вывода: {e}")
+        return None'''
+
+
+'''remote_path = '/2025/02/'
+for i in range(26, 27):
+    sftp_client = connect_to_sftp("192.168.1.150", "monitor", "Audio4analy6!6", remote_path + str(i))
+    files = sftp_client.listdir()
+    temp=remote_path+str(i)+'/'
+    for file in files:
+        sftp_client.get(temp+file,f"/home/dev/ExampleApp/audiofiles{temp}{file}")'''
+
+days01= ['30']#[str(i) for i in range(24,32)]
+#[24,25,26,27,28,29,30,31]
+days03= ['0'+str(i) for i in range(1,16)]+[str(i) for i in range(10,14)]
+#[01,02,03,04,05,06,07,08,09,10,11,12...21]
+
+'''
+#for day in days01:
+#    os.makedirs(f'/home/dev/ExampleApp/textfiles/2025/01/{day}')
+    rec_all_from(f'/2025/01/{str(day)}')
+    print(f'/2025/01/{str(day)}',len(os.listdir(f'/home/dev/ExampleApp/textfiles/2025/01/{day}')))
+'''
+
+for day in days03:
+#    os.makedirs(f'/home/dev/ExampleApp/textfiles/2025/02/{day}')
+    rec_all_from(f'/2025/03/{str(day)}')
+    try:
+        os.makedirs('/home/dev/ExampleApp/textfiles' + f'/2025/03/{str(day)}')
+    except FileExistsError:
+        print('ok')
+    print(f'/2025/03/{str(day)}',len(os.listdir(f'/home/dev/ExampleApp/textfiles/2025/03/{day}')))

ai/preprocessing.py

@@ -0,0 +1,10 @@
+
+
+def preprocessing(file_path, channel):
+    """
+    Разделение аудиозаписи на реплики
+    yield - функция генератор, возвращет N количество раз в цикле FOR
+    """
+    # TODO: перенести функцию разделения текста
+    for i in range(0,10):
+        yield i

ai/rec_all_for.py

@@ -0,0 +1,41 @@
+import os
+import time
+import recognition
+import paramiko
+
+#from apiApp.ai.bot import remote_path
+from cleantemp import clean
+path = '/home/dev/ExampleApp/audiofiles'
+'''
+def is_file_fully_loaded(file_path):
+        size = os.path.getsize(file_path)
+        time.sleep(1)
+        new_size = os.path.getsize(file_path)
+        if size != new_size: return False
+        else: return True
+
+'''
+
+import torch
+import soundfile as sf
+import librosa
+
+def rec_all_from(remote_path):
+    try: os.makedirs('/home/dev/ExampleApp/audiofiles'+remote_path)
+    except FileExistsError: print('ok')
+    for i in os.listdir('/home/dev/ExampleApp/audiofiles'+remote_path):
+        try:
+            print(i)
+            if os.path.getsize(path + remote_path + '/' + i) >= 4096:
+                recognition.to_txt(path + remote_path + '/'+i,remote_path)
+            else:
+                fin = open('/home/dev/ExampleApp/textfiles'+remote_path+f'/{i}.txt', 'w')
+                fin.write('Аудиозапись пуста.')
+                fin.close()
+            clean()
+        except torch.OutOfMemoryError:
+            with open("/home/dev/ExampleApp/govno/govno.txt", "a") as govno:
+                govno.write(remote_path+"/"+i+"\n")
+        except IndexError:
+            with open("/home/dev/ExampleApp/govno/govno.txt", "a") as govno:
+                govno.write(remote_path+"/"+i+" IndexError\n")

ai/recognition.py

@@ -0,0 +1,380 @@
+
+
+#TODO: Убрать комментарии
+import wave
+import torch, torchaudio
+from nemo.collections.asr.models import EncDecRNNTBPEModel
+from nemo.collections.asr.modules.audio_preprocessing import (AudioToMelSpectrogramPreprocessor as NeMoAudioToMelSpectrogramPreprocessor)
+from nemo.collections.asr.parts.preprocessing.features import (FilterbankFeaturesTA as NeMoFilterbankFeaturesTA)
+import locale
+
+#from rec_all_for_day import flush
+
+locale.getpreferredencoding = lambda: "UTF-8"
+import mutagen
+import os
+import librosa
+import soundfile as sf
+import noisereduce as nr
+import pydub
+from pydub import AudioSegment, silence
+model = None
+ckpt=None
+thresh=-90.0
+torch.cuda.empty_cache()
+max_memory = 14 * 1024**3  # 20GB
+torch.cuda.set_per_process_memory_fraction(14/24)
+import torch
+import gc
+def flush():
+    gc.collect()
+    torch.cuda.empty_cache()
+    torch.cuda.reset_peak_memory_stats()
+    # Дополнительная очистка если используется multiprocessing
+    '''if torch.multiprocessing.is_started():
+        torch.multiprocessing.reducer._python_exit()if torch.multiprocessing.is_started():
+        torch.multiprocessing.reducer._python_exit()'''
+model = None
+
+def init_model():
+    global model
+    if model is not None:  # Предотвращаем повторную загрузку
+        return
+
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+
+    try:
+        with torch.no_grad():
+            torch.cuda.empty_cache()
+            model = EncDecRNNTBPEModel.from_config_file("/home/dev/GigaAM_API/rnnt_model_config.yaml")
+            ckpt = torch.load("/home/dev/GigaAM_API/rnnt_model_weights.ckpt", map_location="cpu")
+            model.load_state_dict(ckpt, strict=False)
+            model.eval().to(device)
+    except torch.cuda.OutOfMemoryError:
+        handle_oom_error()
+
+
+def handle_oom_error():
+    torch.cuda.empty_cache()
+    if 'model' in globals():
+        del model
+    gc.collect()
+    raise RuntimeError("Out of Memory. Model initialization failed.")
+
+
+init_model()
+
+def clean():
+    if len(os.listdir('/home/dev/ExampleApp/temp/splitted/segments/'))>0:
+        for i in os.listdir('/home/dev/ExampleApp/temp/splitted/segments/'):
+            os.remove(os.path.join('/home/dev/ExampleApp/temp/splitted/segments/',i))
+
+class FilterbankFeaturesTA(NeMoFilterbankFeaturesTA):
+    def __init__(self, mel_scale: str = "htk", wkwargs=None, **kwargs):
+        if "window_size" in kwargs:
+            del kwargs["window_size"]
+        if "window_stride" in kwargs:
+            del kwargs["window_stride"]
+
+        super().__init__(**kwargs)
+
+        self._mel_spec_extractor: torchaudio.transforms.MelSpectrogram = (
+            torchaudio.transforms.MelSpectrogram(
+                sample_rate=self._sample_rate,
+                win_length=self.win_length,
+                hop_length=self.hop_length,
+                n_mels=kwargs["nfilt"],
+                window_fn=self.torch_windows[kwargs["window"]],
+                mel_scale=mel_scale,
+                norm=kwargs["mel_norm"],
+                n_fft=kwargs["n_fft"],
+                f_max=kwargs.get("highfreq", None),
+                f_min=kwargs.get("lowfreq", 0),
+                wkwargs=wkwargs,
+            )
+        )
+
+class AudioToMelSpectrogramPreprocessor(NeMoAudioToMelSpectrogramPreprocessor):
+    def __init__(self, mel_scale: str = "htk", **kwargs):
+        super().__init__(**kwargs)
+        kwargs["nfilt"] = kwargs["features"]
+        del kwargs["features"]
+        self.featurizer = (
+            FilterbankFeaturesTA(
+                mel_scale=mel_scale,
+                **kwargs,
+            )
+        )
+
+def segment(file_path):
+    clean()
+    dia = {}
+    lraudiosplit(file_path)
+    temp_path = '/home/dev/ExampleApp/temp/splitted/'
+    segments=[]#{}
+    audio, sr = librosa.load(os.path.join(temp_path, 'client.wav'), mono=True)
+    silence = silence_c()
+    if len(silence)>1:
+        for i in range(len(silence)):
+            sf.write(os.path.join(temp_path, 'segments', 'c_' + str(silence[i][0]+1) + '.wav'), audio[silence[i][0]:silence[i][1]], sr)
+            segments.append(os.path.join(temp_path, 'segments', 'c_' + str(silence[i][0]+1) + '.wav'))
+            dia[silence[i][0]+1] = 'П: '
+    else:
+        sf.write(os.path.join(temp_path, 'segments', 'c' + '.wav'), audio, sr)
+        segments.append(os.path.join(temp_path, 'segments', 'c' + '.wav'))
+        dia[0] = 'П: '
+    silence=silence_o()
+    if len(silence) > 1:
+        audio, sr = librosa.load(os.path.join(temp_path, 'operator.wav'), mono=True)
+
+        for i in range(len(silence)):
+            sf.write(os.path.join(temp_path, 'segments', 'o_' + str(silence[i][0]) + '.wav'), audio[silence[i][0]:silence[i][1]], sr)
+            segments.append(os.path.join(temp_path, 'segments', 'o_'+str(silence[i][0]) + '.wav'))
+            dia[silence[i][0]]='О: '
+    else:
+        sf.write(os.path.join(temp_path, 'segments', 'o' + '.wav'), audio, sr)
+        segments.append(os.path.join(temp_path, 'segments', 'o' + '.wav'))
+        dia[1] = 'О: '
+    return segments,dia
+
+def segmentmono(file_path):
+    dia = {}
+    temp_path='/home/dev/ExampleApp/apiApp/ai'
+    if file_path[-3:]=='ogg':
+        sound = AudioSegment.from_ogg(os.path.join(temp_path, file_path))
+        sound.export(f"{file_path}.wav", format="wav")
+        sound = AudioSegment.from_file(os.path.join(temp_path, f'{file_path}.wav'))
+    elif file_path[-3:] == 'mp3':
+        sound = AudioSegment.from_mp3(os.path.join(temp_path, file_path))
+        sound.export(f"{file_path}.wav", format="wav")
+        sound = AudioSegment.from_file(os.path.join(temp_path, f'{file_path}.wav'))
+    elif file_path[-3:] == 'wav':
+        sound = AudioSegment.from_file(os.path.join(temp_path, file_path))
+    else:
+        return TypeError
+    segments=[]#{}
+    audio, sr = librosa.load(os.path.join(temp_path, file_path), mono=True)
+    silence = pydub.silence.detect_silence(sound, silence_thresh=-60.0)
+    j=1
+    if len(silence)>1:
+        sf.write(os.path.join(temp_path, 'segments', 'c_' + str(str(j) + '.wav')), audio[:silence[1][0]], sr)
+        segments.append(os.path.join(temp_path, 'segments', 'c_' + str(str(j) + '.wav')))
+        dia[silence[0][1]] = 'П: '
+        for i in range(len(silence)):
+            j+=1
+            sf.write(os.path.join(temp_path, 'segments', 'c_' + str(j) + '.wav'), audio[silence[i][1]:silence[i + 1][0]], sr)
+            segments.append(os.path.join(temp_path, 'segments', 'c_' + str(j) + '.wav'))
+            dia[silence[i][1]] = ''
+    else: segments='123.wav'
+    return segments,dia
+
+def lraudiosplit(file_path):
+    incoming=True
+    s_audio_dir = '/home/dev/ExampleApp/temp/splitted/'
+    opath,cpath = s_audio_dir + 'client.wav',s_audio_dir + 'operator.wav'
+    audio, sr = librosa.load(file_path, mono=False)
+    if 'out' in file_path: incoming=False
+    if incoming: sf.write(opath, audio[0], sr); sf.write(cpath, audio[1], sr)
+    else: sf.write(cpath, audio[0], sr); sf.write(opath, audio[1], sr)
+    return cpath, opath
+
+def silence_c():
+    global thresh
+    saudio_dir = '/home/dev/ExampleApp/temp/splitted/'
+    cpath = saudio_dir + 'client.wav'
+    audio, sr = librosa.load(cpath, mono=False)
+    caudio,csr=librosa.load(cpath,mono=True)
+    client_reduced_noise = nr.reduce_noise(y=caudio, sr=csr, prop_decrease=1, time_constant_s=0.1)
+    sf.write(cpath, client_reduced_noise, csr)
+    sound = AudioSegment.from_file(os.path.join(saudio_dir, 'client.wav'))
+    ret = pydub.silence.detect_silence(sound, silence_thresh=thresh)
+    for s in ret:
+        s[0] = int(round(s[0] * sr / 1000, 0))
+        s[1] = int(round(s[1] * sr / 1000, 0))
+    if len(ret)>1:
+        ret2=[[0,ret[0][0]]]
+        for i in range(len(ret)-1):
+            ret2.append([ret[i][1],ret[i+1][0]])
+        ret2.append([ret[-1][1], int(round(len(sound) * sr / 1000,0))])
+        while [0,0] in ret2:
+            ret2.pop(ret2.index([0,0]))
+        for i in ret2:
+            if i[0]!=0: i[0]-=sr//2
+            if i[1]<=len(sound)-(sr+5): i[1]+=sr//2
+        sf.write(cpath, audio, 22000)
+        sf.write(cpath, audio, 22000)
+        if ret2[-1][0]>ret2[-1][1]:
+            ret2[-1][1]=int(round(ret2[-1][1] * sr / 1000, 0))
+        return ret2
+    else: return ret
+
+def silence_o():
+    global thresh
+    saudio_dir = '/home/dev/ExampleApp/temp/splitted/'
+    opath = saudio_dir + 'operator.wav'
+    audio, sr = librosa.load(opath, mono=False)
+    oaudio,osr=librosa.load(opath,mono=True)
+    client_reduced_noise = nr.reduce_noise(y=oaudio, sr=osr,prop_decrease=1, time_constant_s=0.2)
+    sf.write(opath, client_reduced_noise, osr)
+    sound = AudioSegment.from_file(os.path.join(saudio_dir, 'operator.wav'))
+    ret = pydub.silence.detect_silence(sound, silence_thresh=thresh+15)
+    for s in ret:
+        s[0] = int(round(s[0] * sr / 1000, 0))
+        s[1] = int(round(s[1] * sr / 1000, 0))
+    if len(ret)>1:
+        ret2=[[0,ret[0][0]]]
+        for i in range(len(ret)-1):
+            ret2.append([ret[i][1],ret[i+1][0]])
+        ret2.append([ret[-1][1], int(round(len(sound) * sr / 1000,0))])
+
+        for i in ret2:
+            if i[0]!=0: i[0]-=sr//2
+            if i[1]<=len(sound)-(sr+5): i[1]+=sr
+        sf.write(opath, audio, 22000)
+        if ret2[-1][0]>ret2[-1][1]:
+            ret2[-1][1]=int(round(ret2[-1][1] * sr / 1000, 0))
+        return ret2
+    else: return ret
+
+def run_recognition(file_path):
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+    model = EncDecRNNTBPEModel.from_config_file("/home/dev/ExampleApp/apiApp/ai/rnnt_model_config.yaml")
+    ckpt = torch.load("/home/dev/ExampleApp/apiApp/ai/rnnt_model_weights.ckpt", map_location="cpu")
+    model.load_state_dict(ckpt, strict=False)
+    model.eval()
+    #model = model.half()
+    model = model.to(device)
+    segments,dia=segment(file_path)
+    dia2={}
+    try:
+        call_text=model.transcribe(segments)[0]
+    except torch.OutOfMemoryError:
+        flush()
+        init_model()
+        run_recognition(file_path)
+    for i in sorted(dia):
+        j=list(dia.keys()).index(i)
+        if call_text[j]!='':
+            dia2[i] = dia[i]+call_text[j]
+
+    time_positions = sorted(dia2)
+    resul = []
+    for i in time_positions:
+        resul.append({"timestamp": i, "person": dia2[i][:3], "text": dia2[i][3:]})
+        '''if dia2[i][:3]=='О: ':
+            resul.append({"timestamp": i, "person": 'Опеатор:', "text": dia2[i][3:]})
+        else:
+            resul.append({"timestamp": i, "person": 'Пациент:', "text": dia2[i][3:]})'''
+    result = {"data":resul}
+    return result
+
+def to_txt_wdb(file_path):
+    flush()
+    file_name = file_path.split('/')[-1]
+    global model
+    if os.path.getsize(file_path)>=1024:
+        text = ''
+        global device,model,ckpt
+        if model is None: init_model()
+        segments, dia = segment(file_path)
+        dia2 = {}
+        if file_path[-1]=='3':
+            sound = AudioSegment.from_mp3(file_path)
+            sound.export(f"{file_path}.wav", format="wav")
+        filename = file_path.replace('/home/dev/ExampleApp/temp/', '')
+        fin = open(f'{filename}.txt'.replace('audiofiles/',''), 'w')
+        call_text = model.transcribe(segments)[0]
+        for i in sorted(dia):
+            j = list(dia.keys()).index(i)
+            if call_text[j] != '':
+                dia2[i] = dia[i] + call_text[j]
+        time_positions = sorted(dia2)
+        resul = []
+        for i in time_positions:
+            resul.append({"timestamp": i, "person": dia2[i][:3], "text": dia2[i][3:]})
+     #   fin=open(f'/home/dev/ExampleApp/textfiles{remote_path}/{filename}.txt','w')
+        for i in time_positions:
+            text+=f"{dia2[i]}\n"
+        fin.write(text)
+        fin.close()
+        return {'file_name' : file_name,
+                'text' : text}
+    elif os.path.getsize(file_path)<1024:
+        filename = file_path
+        fin = open(f'{filename}.txt'.replace('audiofiles/',''), 'w')
+        fin.write("Аудиозапись пуста.")
+        fin.close()
+        return {'file_name' : file_name,
+                'text' : "Аудиозапись пуста."}
+    else: return 'Аудиозапись сжата.'
+
+def stereo_or_mono(file_path):
+    if file_path[-3:]=='wav':
+        with wave.open(file_path, "rb") as wav_file:
+            num_channels = wav_file.getnchannels()
+        if num_channels == 1: return to_message(file_path)
+        elif num_channels == 2: return call_to_message(file_path)
+    elif file_path[-3:]=='mp3':
+        temp_path = '/home/dev/ExampleApp/apiApp/ai'
+        audio = mutagen.File(file_path)
+        sound = AudioSegment.from_mp3(os.path.join(temp_path, file_path))
+        sound.export(f"{file_path}.wav", format="wav")
+        if audio.info.channels == 1: return to_message(file_path)
+        elif audio.info.channels == 2: return call_to_message(file_path)
+    elif file_path[-3:] == 'ogg': return to_message(file_path)
+    else: return 'Неподдерживаемый тип файла'
+
+def call_to_message(file_path):
+    print(file_path)
+    global model, ckpt
+    if model is None: init_model()
+    segments, dia = segment(file_path)
+    dia2 = {}
+    try:
+        if model is None: init_model()
+        call_text = model.transcribe(segments)[0]
+    except torch.OutOfMemoryError:
+        model = None
+        init_model()
+        call_text = model.transcribe(segments)[0]
+    if call_text.count('')==len(call_text):
+        return 'Текст не был распознан, попробуйте другой файл'
+    elif call_text[:int(len(call_text)/2)-1]==call_text[int(len(call_text)/2):len(call_text)-1]:
+        result = str(call_text[:int(len(call_text)/2)])
+        result = result.replace("['",'')
+        result = result.replace("']", '')
+        result = result.replace("', '", ' ')
+        return result
+    else:
+
+        for i in sorted(dia):
+            j = list(dia.keys()).index(i)
+            if call_text[j] != '':
+                dia2[i] = dia[i] + call_text[j]
+        time_positions = sorted(dia2)
+        resul = []
+        for i in time_positions:
+            resul.append({"timestamp": i, "person": dia2[i][:3], "text": dia2[i][3:]})
+        result = ''
+        for i in time_positions:
+            result+=f"{dia2[i]}\n"
+        result = result.replace('О: ', '\n<b>Оператор:</b> \n')
+        result = result.replace('П: ', '\n<b>Пациент:</b> \n')
+        return result
+
+def to_message(file_path):
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+    model = EncDecRNNTBPEModel.from_config_file("/home/dev/ExampleApp/apiApp/ai/rnnt_model_config.yaml")
+    ckpt = torch.load("/home/dev/ExampleApp/apiApp/ai/rnnt_model_weights.ckpt", map_location="cpu")
+    model.load_state_dict(ckpt, strict=False)
+    model.eval()
+    model = model.to(device)
+    segments, dia = segmentmono(file_path)
+    if segments=='123.wav': result=model.transcribe([file_path])[0][0]
+    else:
+        call_text = model.transcribe(segments)[0]
+        result=''
+        for i in call_text:
+            if i != '':
+                result+=(i+'  ')
+    return result

ai/rnnt_model_config.yaml

@@ -0,0 +1,109 @@
+model_class: enc_dec_rnnt_bpe
+sample_rate: 16000
+log_prediction: true
+model_defaults:
+  enc_hidden: 768
+  pred_hidden: 320
+  join_hidden: 320
+
+preprocessor:
+  _target_: nemo.collections.asr.modules.AudioToMelSpectrogramPreprocessor
+  sample_rate: 16000
+  n_fft: 400
+  n_window_size: 400
+  window_size: null
+  n_window_stride: 160
+  window_stride: null
+  features: 64
+  dither: 0.0
+  preemph: null
+  log: true
+  log_zero_guard_type: clamp
+  normalize: null
+  pad_to: 0
+  mel_norm: null
+  window: hann
+  log_zero_guard_value: 1e-9
+
+tokenizer:
+  dir: tokenizer_all_sets/
+  type: bpe
+
+validation_ds:
+  shuffle: False
+  manifest_filepath: null
+
+encoder:
+  _target_: nemo.collections.asr.modules.ConformerEncoder
+  feat_in: 64
+  feat_out: -1
+  n_layers: 16
+  d_model: 768
+  subsampling: striding
+  subsampling_factor: 4
+  subsampling_conv_channels: 768
+  ff_expansion_factor: 4
+  self_attention_model: rel_pos
+  pos_emb_max_len: 5000
+  n_heads: 16
+  xscaling: false
+  untie_biases: true
+  conv_kernel_size: 31
+  dropout: 0.1
+  dropout_emb: 0.1
+  dropout_att: 0.1
+decoder:
+  _target_: nemo.collections.asr.modules.RNNTDecoder
+  normalization_mode: null
+  random_state_sampling: false
+  blank_as_pad: true
+  vocab_size: 512
+  prednet:
+    pred_hidden: 320
+    pred_rnn_layers: 1
+    t_max: null
+    dropout: 0.0
+joint:
+  _target_: nemo.collections.asr.modules.RNNTJoint
+  log_softmax: null
+  fuse_loss_wer: false
+  fused_batch_size: 1
+  jointnet:
+    joint_hidden: 320
+    activation: relu
+    dropout: 0.0
+    encoder_hidden: 768
+optim:
+  name: adamw
+  lr: 5.0e-05
+  betas:
+  - 0.9
+  - 0.98
+  weight_decay: 0.01
+  sched:
+    name: CosineAnnealing
+    warmup_steps: 10000
+    warmup_ratio: null
+    min_lr: 1.0e-07
+nemo_version: 1.12.0
+decoding:
+  strategy: greedy_batch
+  preserve_alignments: false
+  greedy:
+    max_symbols: 3
+  beam:
+    beam_size: 5
+    score_norm: true
+
+
+loss:
+  loss_name: default
+  mwer: false
+  rnnt_reduction: mean_batch
+  wer_coef: false
+  subtract_mean: true
+  warprnnt_numba_kwargs:
+    fastemit_lambda: 0.0
+    clamp: -1.0
+  rnnt_weight: 0.1
+  unique_hyp: true

apiApp/__init__.py

@@ -0,0 +1,43 @@
+# -*- coding: utf-8 -*-
+from flask import Flask
+import logging
+import sys, os
+from flask_compress import Compress
+from apiApp.config import config
+from flask_sqlalchemy import SQLAlchemy
+
+# Disable
+def blockPrint():
+    sys.stdout = open(os.devnull, 'w')
+# blockPrint()
+__version__ = '0.1'
+
+apiApp = Flask(__name__, instance_relative_config=True)
+
+
+
+apiApp.config['SECRET_KEY'] = '2qwtq2'
+apiApp.config['INTERNAL_API_KEY'] = 'lGR9g09RW5cE07rKivRNIH9f'
+apiApp.config['REMOTE_LOGIN_KEY'] = 'oRTiLTHeNTraZEsI'
+apiApp.config['SQLALCHEMY_DATABASE_URI'] = config['DATABASE']['URI']
+apiApp.config['SQLALCHEMY_TRACK_MODIFICATIONS'] = True
+apiApp.config['SQLALCHEMY_POOL_SIZE'] = 10
+apiApp.config['SQLALCHEMY_MAX_OVERFLOW'] = 20
+db = SQLAlchemy(apiApp)
+
+compress = Compress()
+compress.init_app(apiApp)
+
+
+logging.basicConfig(level = logging.INFO)
+
+from apiApp.interfaces import *
+from apiApp.helper import jinjaHelper
+#from apiApp.database import *
+from apiApp.helper.utils import attach_monitor
+attach_monitor(apiApp)
+
+
+
+with apiApp.app_context():
+    db.create_all()

apiApp/config/__init__.py

@@ -0,0 +1,3 @@
+import configparser
+config = configparser.ConfigParser()
+config.read('config.ini')

apiApp/database/Call.py

@@ -0,0 +1,61 @@
+'''from apiApp import db
+from sqlalchemy_serializer import SerializerMixin
+import uuid
+from apiApp import apiApp
+
+class Call(db.Model, SerializerMixin):
+    __tablename__ = "call"
+    id = db.Column(db.Integer, primary_key=True)
+    filename = db.Column(db.String(100), unique=True)
+
+    appointment = db.Column(db.Boolean, default=False)
+    target = db.Column(db.Boolean, default=False)
+    operation = db.Column(db.Boolean, default=False)
+
+    call_text = db.Column(db.Text, unique=True)
+    ai_call_text = db.Column(db.Text)
+    ai_conclusion = db.Column(db.Text)
+
+    operator_id = db.Column(db.UUID, db.ForeignKey('operator.id'))
+
+    def add_call(filename, operator_id, call_text=None, ai_call_text=None,
+                 ai_conclusion=None, appointment=False, target=False, operation=False):
+        try:
+            # Если контекст уже есть (Flask-запрос), работаем в нём
+            if apiApp and hasattr(apiApp, 'app_context'):
+                return Call._add_call_internal(
+                    filename, operator_id, call_text, ai_call_text,
+                    ai_conclusion, appointment, target, operation
+                )
+            # Если контекста нет (скрипт), создаём его
+            else:
+                from apiApp import create_app
+                app = create_app()
+                with app.app_context():
+                    return Call._add_call_internal(
+                        filename, operator_id, call_text, ai_call_text,
+                        ai_conclusion, appointment, target, operation
+                    )
+        except Exception as e:
+            return False, f"Ошибка: {str(e)}"
+
+    def _add_call_internal(filename, operator_id, call_text, ai_call_text,
+                          ai_conclusion, appointment, target, operation):
+        existing_call = Call.query.filter_by(filename=filename, operator_id=operator_id).first()
+        if existing_call:
+            return False, "Такой звонок уже существует."
+
+        new_call = Call(
+            filename=filename,
+            operator_id=operator_id,
+            call_text=call_text,
+            ai_call_text=ai_call_text,
+            ai_conclusion=ai_conclusion,
+            appointment=appointment,
+            target=target,
+            operation=operation
+        )
+
+        db.session.add(new_call)
+        db.session.commit()
+        return True, new_call'''

apiApp/database/Operator.py

@@ -0,0 +1,14 @@
+from apiApp import db
+from sqlalchemy_serializer import SerializerMixin
+import uuid
+
+
+class Operator(db.Model, SerializerMixin):
+    __tablename__ = "operator"
+    id = db.Column(db.UUID(as_uuid=True), primary_key=True, default=uuid.uuid4)
+    fio = db.Column(db.String(100))
+    num = db.Column(db.Integer)
+
+    calls = db.relationship("Call", backref="operator")
+
+

apiApp/database/User.py

@@ -0,0 +1,28 @@
+from apiApp import db
+from sqlalchemy_serializer import SerializerMixin
+import uuid
+from werkzeug.security import check_password_hash
+
+class User(db.Model):
+    __tablename__ = "user"
+    id = db.Column(db.Integer, primary_key=True)
+    name = db.Column(db.String(50))
+    full_name = db.Column(db.String(100))
+    email = db.Column(db.String(100))
+    password = db.Column(db.String(250))  # Увеличено до 250 символов
+    admin = db.Column(db.Boolean, default=False)
+
+    def verify_password(self, password):
+        if check_password_hash(self.password, password):
+            return True
+        else:
+            return False
+
+    def is_active(self):
+        return True
+
+    def is_authenticated(self):
+        return True
+
+    def get_id(self):
+        return self.id

apiApp/database/__init__.py

@@ -0,0 +1,3 @@
+from apiApp.database.Call import Call
+from apiApp.database.Operator import Operator
+from apiApp.database.User import User

apiApp/helper/jinjaHelper.py

@@ -0,0 +1,11 @@
+from apiApp import apiApp
+import random, string
+
+@apiApp.context_processor
+def example():
+    def randomString(stringLength=10):
+        letters = string.ascii_lowercase
+        return ''.join(random.choice(letters) for i in range(stringLength))
+
+
+    return dict(randomString = randomString)

apiApp/helper/utils.py

@@ -0,0 +1,11 @@
+from healthcheck import HealthCheck, EnvironmentDump
+import sys
+
+def app_data():
+    return {
+        "platform": "Pirogov.AI",
+        "python_version": sys.version
+    }
+
+def attach_monitor(app):
+    health = HealthCheck(app, '/status')

apiApp/interfaces/__init__.py

@@ -0,0 +1 @@
+from apiApp.interfaces.apiFunctions import *

apiApp/interfaces/apiFunctions.py

@@ -0,0 +1,36 @@
+from apiApp import apiApp
+from flask import render_template, make_response, send_from_directory, request
+import os
+from apiApp.config import config
+import json
+import requests
+from ai.recognition import run_recognition, stereo_or_mono
+import time
+
+
+
+
+@apiApp.route('/recognize', methods=['POST'])
+def recognize_call():
+    r_files = request.files.getlist("file")
+
+    file = r_files[0]
+    file.save(os.path.join(config['DIST']['PATH'], file.filename))
+
+    call_text = run_recognition(os.path.join(config['DIST']['PATH'], file.filename))
+    time.sleep(10)
+
+    return {'call_text': call_text["data"]}
+
+@apiApp.route('/recognize_bot', methods=['POST'])
+def recognize_cal_bot():
+
+    r_files = request.files.getlist("file")
+
+    file = r_files[0]
+    file.save(os.path.join(config['DIST']['PATH'], file.filename))
+
+    call_text = stereo_or_mono(os.path.join(config['DIST']['PATH'], file.filename))
+    #time.sleep(1)
+
+    return {'call_text': call_text}

config.ini

@@ -0,0 +1,6 @@
+[WEB_APP]
+PORT = 5001
+[DIST]
+PATH = /home/dev/GigaAM_API/temp
+[DATABASE]
+URI = postgresql://postgres:Gxq-8fN7A9@localhost:5432/call_analytics

keep_txts.py

@@ -0,0 +1,17 @@
+import datetime, time
+from rec_all_for_day import today_audios
+
+
+date = str(datetime.datetime.today()).split(' ')[0].split('-')
+remote_path = f'/{date[0]}/{date[1]}/{date[2]}/'
+
+
+
+def permanent_recognition():
+    curr_time = str(datetime.datetime.today()).split(' ')[1].split(':')[0:2]
+    curr_time = [int(i) for i in curr_time]
+    #if curr_time[0]<17 and curr_time[0]>7:
+    #time.sleep((17-curr_time[0])*3600+(17-curr_time[1])*60)
+    today_audios()
+
+permanent_recognition()

makefilefornatalia.py

@@ -0,0 +1,11 @@
+import os
+text=''
+for file in os.listdir('/media/dev/TEXTFILES/2025/06/01/'):
+    with open(f'/media/dev/TEXTFILES/2025/06/01/{file}', 'r') as fil:
+        text += f"{file} (01.06.2025) - \n{fil.read()}\n\n"
+for day in ['31','30', '29','28','27','26']:
+    for file in os.listdir('/media/dev/TEXTFILES/2025/05'+'/'+day):
+        with open(os.path.join('/media/dev/TEXTFILES/2025/05'+'/'+day+'/'+file),'r') as fil:
+            text+=f"{file} ({day}.05.2025) - \n{fil.read()}\n\n"
+with open('calls_for_week_in_single_file.txt','w') as f:
+    f.write(text)

rec_all_for_day.py

@@ -0,0 +1,165 @@
+import os
+import time
+import shutil
+#from apiApp import init_model
+#from apiApp.ai import recognition
+from ai import recognition
+import paramiko
+#from apiApp.database import Call
+from ai.recognition import clean
+import datetime
+path = '/media/dev/TEXTFILES/audiofiles'
+import gc
+import torch
+
+def flush():
+  gc.collect()
+  torch.cuda.empty_cache()
+  torch.cuda.reset_peak_memory_stats()
+
+def is_file_fully_loaded(file_path):
+    #try:
+    size = os.path.getsize(file_path)
+    #except FileNotFoundError:
+    #    print('not found')
+    #    return False
+    new_size = os.path.getsize(file_path)
+    if size != new_size: return False
+    else: return True
+
+def connect_to_sftp(hostname, username, password, remote_path):
+    try:
+        ssh = paramiko.SSHClient()
+        ssh.set_missing_host_key_policy(paramiko.AutoAddPolicy())  # Добавляет неизвестный хост в известные
+        ssh.connect(hostname, username=username, password=password)
+        sftp = ssh.open_sftp()
+        if len(remote_path)>3:
+            if remote_path[1]=='/': remote_path=remote_path[1:]
+        sftp.chdir(remote_path) # Переходим в указанную директорию.  Обработка ошибок ниже.
+        return sftp
+    except paramiko.AuthenticationException:
+        print("Ошибка аутентификации. Проверьте имя пользователя и пароль.")
+        return None
+    except paramiko.SSHException as e:
+        print(f"Ошибка SSH: {e}")
+        return None
+    except FileNotFoundError:
+        print(f"Директория {remote_path} не найдена на сервере.")
+        return None
+    except IOError as e:
+        print(f"Ошибка ввода-вывода: {e}")
+        return None
+
+from datetime import timedelta
+def weekd(s):
+    return [-1,-2,-3,-4,-5,-6,-7]
+"""output = []
+    if s == 3:
+        output=[0]
+    if s > 3:
+        for i in range(0,s-3+1):
+            output.append(-i)
+    if s < 3:
+        for i in range(0,s+5):
+            output.append(-i)
+    return sorted(output, reverse = False)"""
+
+
+def today_audios():
+    while True:
+        try:
+            for i in weekd(datetime.datetime.today().weekday()):
+                date = str(datetime.datetime.today()).split(' ')[0].split('-')
+                hour = str(datetime.datetime.today()).split(' ')[1].split(':')[0]
+                if int(hour)<9:
+                    yesterday = datetime.datetime.now() - timedelta(days=-i+1)
+                    year = yesterday.strftime("%Y")
+                    month = yesterday.strftime("%m")
+                    day = yesterday.strftime("%d")
+                    if int(day) >= 10: remote_path = f"/{year}/{month}/{int(day)}/"
+                    else: remote_path = f"/{year}/{month}/0{int(day)}/"
+                else:
+                    yesterday = datetime.datetime.now() - timedelta(days=-i)
+                    year = yesterday.strftime("%Y")
+                    month = yesterday.strftime("%m")
+                    day = yesterday.strftime("%d")
+                    if int(day) >= 10: remote_path = f"/{year}/{month}/{str(int(day))}/"
+                    else: remote_path = f"/{year}/{month}/0{int(day)}/"
+                os.makedirs('/media/dev/TEXTFILES'+remote_path, exist_ok=True)
+                os.makedirs(f'/media/dev/TEXTFILES/error_audiofiles/OutOfMemoryError{remote_path}', exist_ok=True)
+                os.makedirs(f'/media/dev/TEXTFILES/error_audiofiles/IndexError{remote_path}', exist_ok=True)
+                os.makedirs('/media/dev/TEXTFILES/audiofiles' + remote_path, exist_ok=True)
+
+                sftp_client = connect_to_sftp("192.168.1.150", "monitor", "Audio4analy6!6", remote_path)
+                files = sftp_client.listdir()
+                txtfiles = [i[:-4] for i in os.listdir(f'/media/dev/TEXTFILES{remote_path}')] + os.listdir(
+                    f'/media/dev/TEXTFILES/error_audiofiles/OutOfMemoryError{remote_path}') + os.listdir(
+                    f'/media/dev/TEXTFILES/error_audiofiles/IndexError{remote_path}')
+                files = list(set(files)-set(txtfiles))
+                day_audios(remote_path)
+                time.sleep(900)
+        except AttributeError as e:
+            print('SSH error',e)
+            time.sleep(1800)
+            today_audios()
+            
+archive_path = '/media/dev/TEXTFILES/audiofiles'
+
+def day_audios(remote_path):
+        os.makedirs('/media/dev/TEXTFILES' + remote_path, exist_ok=True)
+        os.makedirs(f'/media/dev/TEXTFILES/error_audiofiles/OutOfMemoryError{remote_path}', exist_ok=True)
+        os.makedirs(f'/media/dev/TEXTFILES/error_audiofiles/IndexError{remote_path}', exist_ok=True)
+        os.makedirs('/media/dev/TEXTFILES/audiofiles' + remote_path, exist_ok=True)
+
+        sftp_client = connect_to_sftp("192.168.1.150", "monitor", "Audio4analy6!6", remote_path)
+        filess = sftp_client.listdir()
+        files = []
+        for i in filess:
+            if 'in-' in i and not '_t' in i and not 'mp3' in i:
+                files.append(i)
+        #print(len(files))
+        txtfiles = [i[:-4] for i in os.listdir(f'/media/dev/TEXTFILES{remote_path}')] + os.listdir(
+            f'/media/dev/TEXTFILES/error_audiofiles/OutOfMemoryError{remote_path}') + os.listdir(
+            f'/media/dev/TEXTFILES/error_audiofiles/IndexError{remote_path}')
+        #print(set(files)-set(txtfiles))
+        files = list(set(files) - set(txtfiles))
+        if len(files)>0: print(files)
+        print(len(files))
+        #print(files)
+        flush()
+        if files:
+            for i in files:
+                if 'in-' in i and not '_t' in i and not 'mp3' in i:
+                    sftp_client.get(remote_path + '/' + i, f"{archive_path}{remote_path}{i}")
+                    print('1ok')
+                    #print(f"{archive_path}{remote_path}{i}")
+                    while not is_file_fully_loaded(os.path.join(f"{archive_path}{remote_path}{i}")):
+                        time.sleep(1)
+                    print('2ok')
+                    try:
+                        print('3ok')
+                        call_text = recognition.to_txt_wdb(archive_path + remote_path + i)
+                        print('4ok')
+                        #Call.add_call(i,'-', call_text, '-', '-', '-', '-', '-')
+                    except torch.OutOfMemoryError:
+                        clean()
+                        recognition.init_model()
+                        try:
+                            recognition.to_txt_wdb(archive_path + remote_path + i)
+                        except torch.OutOfMemoryError:
+                            os.makedirs(f'/media/dev/TEXTFILES/error_audiofiles/OutOfMemoryError{remote_path}{i}',
+                                        exist_ok=True)
+                            shutil.copy(archive_path + remote_path + i,
+                                        f'/media/dev/TEXTFILES/error_audiofiles/OutOfMemoryError{remote_path}{i}')
+                    except IndexError:
+                        os.makedirs(f'/media/dev/TEXTFILES/error_audiofiles/IndexError{remote_path}{i}',
+                                    exist_ok=True)
+                        shutil.copy((archive_path + remote_path + i),
+                                    f'/media/dev/TEXTFILES/error_audiofiles/IndexError{remote_path}{i}')
+
+#day_audios('/2025/05/27/')
+
+
+
+
+

requirements.txt

@@ -0,0 +1,290 @@
+absl-py==2.1.0
+aiofiles==24.1.0
+aiogram==3.17.0
+aiohappyeyeballs==2.4.4
+aiohttp==3.11.10
+aiosignal==1.3.1
+alabaster==0.7.16
+alembic==1.14.0
+aniso8601==9.0.1
+annotated-types==0.7.0
+antlr4-python3-runtime==4.9.3
+appdirs==1.4.4
+asciitree==0.3.3
+asteroid-filterbanks==0.4.0
+asttokens==3.0.0
+async-timeout==5.0.1
+attrdict==2.0.1
+attrs==24.2.0
+audioread==3.0.1
+babel==2.16.0
+bcrypt==4.2.1
+beautifulsoup4==4.12.3
+black==19.10b0
+blinker==1.9.0
+boto3==1.35.76
+botocore==1.35.76
+braceexpand==0.1.7
+Brotli==1.1.0
+cdifflib==1.2.6
+certifi==2024.8.30
+cffi==1.17.1
+charset-normalizer==3.4.0
+click==8.0.2
+colorama==0.4.6
+colorlog==6.9.0
+comm==0.2.2
+contourpy==1.3.0
+cryptography==44.0.1
+cycler==0.12.1
+Cython==3.0.11
+datasets==3.1.0
+DAWG-Python==0.7.2
+decorator==5.1.1
+dill==0.3.8
+Distance==0.1.3
+docker-pycreds==0.4.0
+docopt==0.6.2
+docutils==0.21.2
+editdistance==0.8.1
+einops==0.8.0
+exceptiongroup==1.2.2
+executing==2.1.0
+faiss-cpu==1.8.0.post1
+fasteners==0.19
+fasttext==0.9.3
+filelock==3.16.1
+Flask==2.2.5
+Flask-Compress==1.17
+Flask-Migrate==4.0.7
+Flask-RESTful==0.3.10
+Flask-SQLAlchemy==3.1.1
+fonttools==4.55.2
+frozenlist==1.5.0
+fsspec==2024.9.0
+ftfy==6.3.1
+g2p-en==2.1.0
+gdown==5.2.0
+gitdb==4.0.11
+GitPython==3.1.43
+greenlet==3.1.1
+grpcio==1.68.1
+h5py==3.12.1
+huggingface-hub==0.23.2
+hydra-core==1.3.2
+HyperPyYAML==1.2.2
+idna==3.10
+ijson==3.3.0
+imageio==2.37.0
+imageio-ffmpeg==0.6.0
+imagesize==1.4.1
+importlib_metadata==8.5.0
+importlib_resources==6.4.5
+inflect==7.4.0
+iniconfig==2.0.0
+ipython==8.18.1
+ipywidgets==8.1.5
+isort==5.13.2
+itsdangerous==2.2.0
+jedi==0.19.2
+jieba==0.42.1
+Jinja2==3.1.4
+jiwer==2.5.2
+jmespath==1.0.1
+joblib==1.4.2
+julius==0.2.7
+jupyterlab_widgets==3.0.13
+kaldi-python-io==1.2.2
+kaldiio==2.18.0
+kiwisolver==1.4.7
+kornia==0.7.4
+kornia_rs==0.1.7
+latexcodec==3.0.0
+lazy_loader==0.4
+Levenshtein==0.22.0
+librosa==0.10.2.post1
+lightning==2.4.0
+lightning-utilities==0.11.9
+llvmlite==0.43.0
+loguru==0.7.3
+lxml==5.3.0
+magic-filter==1.0.12
+Mako==1.3.7
+Markdown==3.7
+markdown-it-py==3.0.0
+markdown2==2.5.1
+MarkupSafe==3.0.2
+marshmallow==3.23.1
+matplotlib==3.9.3
+matplotlib-inline==0.1.7
+mdurl==0.1.2
+megatron-core==0.3.0
+more-itertools==10.5.0
+morph==0.1.5
+mpmath==1.3.0
+msgpack==1.1.0
+multidict==6.1.0
+multiprocess==0.70.16
+mutagen==1.47.0
+nemo-text-processing==0.1.8rc0
+nemo-toolkit==1.21.0
+networkx==3.2.1
+nltk==3.9.1
+noisereduce==3.0.3
+numba==0.60.0
+numcodecs==0.12.1
+nvidia-cublas-cu12==12.4.5.8
+nvidia-cuda-cupti-cu12==12.4.127
+nvidia-cuda-nvrtc-cu12==12.4.127
+nvidia-cuda-runtime-cu12==12.4.127
+nvidia-cudnn-cu12==9.1.0.70
+nvidia-cufft-cu12==11.2.1.3
+nvidia-curand-cu12==10.3.5.147
+nvidia-cusolver-cu12==11.6.1.9
+nvidia-cusparse-cu12==12.3.1.170
+nvidia-nccl-cu12==2.21.5
+nvidia-nvjitlink-cu12==12.4.127
+nvidia-nvtx-cu12==12.4.127
+omegaconf==2.3.0
+onnx==1.17.0
+OpenCC==1.1.6
+optuna==4.1.0
+packaging==24.2
+pandas==2.2.3
+pangu==4.0.6.1
+parameterized==0.9.0
+paramiko==3.5.1
+parso==0.8.4
+pathspec==0.12.1
+pexpect==4.9.0
+pillow==11.0.0
+plac==1.4.3
+platformdirs==4.3.6
+pluggy==1.5.0
+pooch==1.8.2
+portalocker==3.0.0
+primePy==1.3
+proglog==0.1.10
+progress==1.6
+prompt_toolkit==3.0.48
+propcache==0.2.1
+protobuf==5.29.1
+psutil==6.1.0
+ptyprocess==0.7.0
+pure_eval==0.2.3
+pyannote.audio==3.2.0
+pyannote.core==5.0.0
+pyannote.database==5.1.0
+pyannote.metrics==3.2.1
+pyannote.pipeline==3.0.1
+pyarrow==18.1.0
+pybind11==2.13.6
+pybtex==0.24.0
+pybtex-docutils==1.0.3
+pycparser==2.22
+pydantic==2.10.5
+pydantic_core==2.27.2
+pydub==0.25.1
+Pygments==2.18.0
+pymorphy2==0.9.1
+pymorphy2-dicts-ru==2.4.417127.4579844
+PyNaCl==1.5.0
+pynini==2.1.5
+pyodbc==5.2.0
+pyparsing==3.2.0
+pypinyin==0.53.0
+pypinyin-dict==0.8.0
+pypugjs==5.11.0
+PySocks==1.7.1
+pytest==8.3.4
+pytest-runner==6.0.1
+python-dateutil==2.9.0.post0
+python-dotenv==1.0.1
+pytorch-lightning==2.0.7
+pytorch-metric-learning==2.7.0
+pytz==2024.2
+PyYAML==6.0.2
+rapidfuzz==2.13.7
+regex==2024.11.6
+requests==2.32.3
+rich==13.9.4
+rouge-score==0.1.2
+ruamel.yaml==0.18.6
+ruamel.yaml.clib==0.2.12
+s3transfer==0.10.4
+sacrebleu==2.4.3
+sacremoses==0.1.1
+safetensors==0.4.5
+scikit-learn==1.5.2
+scipy==1.13.1
+semver==3.0.2
+sentence-transformers==2.6.1
+sentencepiece==0.2.0
+sentry-sdk==2.19.2
+setproctitle==1.3.4
+shellingham==1.5.4
+six==1.17.0
+smmap==5.0.1
+snowballstemmer==2.2.0
+sortedcontainers==2.4.0
+soundfile==0.12.1
+soupsieve==2.6
+numpy==1.23.5
+sox==1.5.0
+soxr==0.5.0.post1
+speechbrain==1.0.2
+Sphinx==7.4.7
+sphinxcontrib-applehelp==2.0.0
+sphinxcontrib-bibtex==2.6.3
+sphinxcontrib-devhelp==2.0.0
+sphinxcontrib-htmlhelp==2.1.0
+sphinxcontrib-jsmath==1.0.1
+sphinxcontrib-qthelp==2.0.0
+sphinxcontrib-serializinghtml==2.0.0
+SQLAlchemy==2.0.36
+stack-data==0.6.3
+sympy==1.13.1
+tabulate==0.9.0
+tensorboard==2.18.0
+tensorboard-data-server==0.7.2
+tensorboardX==2.6.2.2
+tensorstore==0.1.45
+termcolor==2.5.0
+text-unidecode==1.3
+textdistance==4.6.3
+texterrors==0.5.1
+threadpoolctl==3.5.0
+tokenizers==0.13.3
+toml==0.10.2
+tomli==2.2.1
+torch==2.5.1
+torch-audiomentations==0.11.1
+torch_pitch_shift==1.2.5
+torchaudio==2.5.1
+torchmetrics==1.6.0
+tqdm==4.67.1
+traitlets==5.14.3
+transformers==4.33.3
+triton==3.1.0
+typed-ast==1.5.5
+typeguard==4.4.1
+typer==0.15.1
+typing_extensions==4.12.2
+tzdata==2024.2
+urllib3==1.26.20
+uWSGI==2.0.28
+waitress==3.0.2
+wandb==0.18.7
+wcwidth==0.2.13
+webdataset==0.1.62
+Werkzeug==3.1.3
+wget==3.2
+widgetsnbextension==4.0.13
+wrapt==1.17.0
+xxhash==3.5.0
+yargy==0.16.0
+yarl==1.18.3
+youtokentome==1.0.6
+zarr==2.18.2
+zipp==3.21.0
+zstandard==0.23.0

rnnt_model_config.yaml

@@ -0,0 +1,109 @@
+model_class: enc_dec_rnnt_bpe
+sample_rate: 16000
+log_prediction: true
+model_defaults:
+  enc_hidden: 768
+  pred_hidden: 320
+  join_hidden: 320
+
+preprocessor:
+  _target_: nemo.collections.asr.modules.AudioToMelSpectrogramPreprocessor
+  sample_rate: 16000
+  n_fft: 400
+  n_window_size: 400
+  window_size: null
+  n_window_stride: 160
+  window_stride: null
+  features: 64
+  dither: 0.0
+  preemph: null
+  log: true
+  log_zero_guard_type: clamp
+  normalize: null
+  pad_to: 0
+  mel_norm: null
+  window: hann
+  log_zero_guard_value: 1e-9
+
+tokenizer:
+  dir: tokenizer_all_sets/
+  type: bpe
+
+validation_ds:
+  shuffle: False
+  manifest_filepath: null
+
+encoder:
+  _target_: nemo.collections.asr.modules.ConformerEncoder
+  feat_in: 64
+  feat_out: -1
+  n_layers: 16
+  d_model: 768
+  subsampling: striding
+  subsampling_factor: 4
+  subsampling_conv_channels: 768
+  ff_expansion_factor: 4
+  self_attention_model: rel_pos
+  pos_emb_max_len: 5000
+  n_heads: 16
+  xscaling: false
+  untie_biases: true
+  conv_kernel_size: 31
+  dropout: 0.1
+  dropout_emb: 0.1
+  dropout_att: 0.1
+decoder:
+  _target_: nemo.collections.asr.modules.RNNTDecoder
+  normalization_mode: null
+  random_state_sampling: false
+  blank_as_pad: true
+  vocab_size: 512
+  prednet:
+    pred_hidden: 320
+    pred_rnn_layers: 1
+    t_max: null
+    dropout: 0.0
+joint:
+  _target_: nemo.collections.asr.modules.RNNTJoint
+  log_softmax: null
+  fuse_loss_wer: false
+  fused_batch_size: 1
+  jointnet:
+    joint_hidden: 320
+    activation: relu
+    dropout: 0.0
+    encoder_hidden: 768
+optim:
+  name: adamw
+  lr: 5.0e-05
+  betas:
+  - 0.9
+  - 0.98
+  weight_decay: 0.01
+  sched:
+    name: CosineAnnealing
+    warmup_steps: 10000
+    warmup_ratio: null
+    min_lr: 1.0e-07
+nemo_version: 1.12.0
+decoding:
+  strategy: greedy_batch
+  preserve_alignments: false
+  greedy:
+    max_symbols: 3
+  beam:
+    beam_size: 5
+    score_norm: true
+
+
+loss:
+  loss_name: default
+  mwer: false
+  rnnt_reduction: mean_batch
+  wer_coef: false
+  subtract_mean: true
+  warprnnt_numba_kwargs:
+    fastemit_lambda: 0.0
+    clamp: -1.0
+  rnnt_weight: 0.1
+  unique_hyp: true

run.py

@@ -0,0 +1,19 @@
+# -*- coding: utf-8 -*-
+from apiApp import apiApp
+from apiApp.config import config
+from waitress import serve
+model = None
+
+'''
+from pathlib import Path
+import os
+BASE_DIR = Path(__file__).parent.absolute()
+CONFIG_PATH = '/home/dev/GigaAM_API/config.ini'
+'''
+
+
+#config.read(CONFIG_PATH)
+
+if __name__ == '__main__':
+    #apiApp.run(debug=True, port=config['WEB_APP']['PORT'], use_reloader=False)
+    serve(apiApp, host='0.0.0.0', port=config['WEB_APP']['PORT'])