Update app.py
Browse files
app.py
CHANGED
@@ -5,68 +5,68 @@ import torch
|
|
5 |
from utils import convert_segments_object_to_text, check_password
|
6 |
from gigiachat_requests import get_access_token, get_completion_from_gigachat
|
7 |
|
8 |
-
|
9 |
-
st.title('Audio Transcription App')
|
10 |
-
st.sidebar.title("Settings")
|
11 |
-
# Sidebar inputs
|
12 |
-
device = st.sidebar.selectbox("Device", ["cpu", "cuda"], index=1)
|
13 |
-
batch_size = st.sidebar.number_input("Batch Size", min_value=1, value=16)
|
14 |
-
compute_type = st.sidebar.selectbox("Compute Type", ["float16", "int8"], index=0)
|
15 |
|
16 |
-
initial_giga_base_prompt = os.getenv('GIGA_BASE_PROMPT')
|
17 |
-
initial_giga_processing_prompt = os.getenv('GIGA_PROCCESS_PROMPT')
|
18 |
|
19 |
-
giga_base_prompt = st.sidebar.text_area("Промпт ГигаЧата для резюмирования", value=initial_giga_base_prompt)
|
20 |
-
giga_max_tokens = st.sidebar.number_input("Максимальное количество токенов при резюмировании", min_value=1, value=1024)
|
21 |
|
22 |
-
enable_summarization = st.sidebar.checkbox("Добавить обработку транскрибации", value=False)
|
23 |
-
giga_processing_prompt = st.sidebar.text_area("Промпт ГигаЧата для обработки транскрибации", value=initial_giga_processing_prompt)
|
24 |
|
25 |
-
ACCESS_TOKEN = st.secrets["HF_TOKEN"]
|
26 |
|
27 |
-
uploaded_file = st.file_uploader("Загрузите аудиофайл", type=["mp4", "wav", "m4a"])
|
28 |
|
29 |
-
if uploaded_file is not None:
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
|
34 |
-
|
35 |
-
|
36 |
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
|
57 |
-
|
58 |
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
|
63 |
-
|
64 |
-
|
65 |
|
66 |
|
67 |
|
68 |
-
|
69 |
-
|
70 |
|
71 |
-
|
72 |
-
|
|
|
5 |
from utils import convert_segments_object_to_text, check_password
|
6 |
from gigiachat_requests import get_access_token, get_completion_from_gigachat
|
7 |
|
8 |
+
if check_password():
|
9 |
+
st.title('Audio Transcription App')
|
10 |
+
st.sidebar.title("Settings")
|
11 |
+
# Sidebar inputs
|
12 |
+
device = st.sidebar.selectbox("Device", ["cpu", "cuda"], index=1)
|
13 |
+
batch_size = st.sidebar.number_input("Batch Size", min_value=1, value=16)
|
14 |
+
compute_type = st.sidebar.selectbox("Compute Type", ["float16", "int8"], index=0)
|
15 |
|
16 |
+
initial_giga_base_prompt = os.getenv('GIGA_BASE_PROMPT')
|
17 |
+
initial_giga_processing_prompt = os.getenv('GIGA_PROCCESS_PROMPT')
|
18 |
|
19 |
+
giga_base_prompt = st.sidebar.text_area("Промпт ГигаЧата для резюмирования", value=initial_giga_base_prompt)
|
20 |
+
giga_max_tokens = st.sidebar.number_input("Максимальное количество токенов при резюмировании", min_value=1, value=1024)
|
21 |
|
22 |
+
enable_summarization = st.sidebar.checkbox("Добавить обработку транскрибации", value=False)
|
23 |
+
giga_processing_prompt = st.sidebar.text_area("Промпт ГигаЧата для обработки транскрибации", value=initial_giga_processing_prompt)
|
24 |
|
25 |
+
ACCESS_TOKEN = st.secrets["HF_TOKEN"]
|
26 |
|
27 |
+
uploaded_file = st.file_uploader("Загрузите аудиофайл", type=["mp4", "wav", "m4a"])
|
28 |
|
29 |
+
if uploaded_file is not None:
|
30 |
+
st.audio(uploaded_file)
|
31 |
+
file_extension = uploaded_file.name.split(".")[-1] # Получаем расширение файла
|
32 |
+
temp_file_path = f"temp_file.{file_extension}" # Создаем временное имя файла с правильным расширением
|
33 |
|
34 |
+
with open(temp_file_path, "wb") as f:
|
35 |
+
f.write(uploaded_file.getbuffer())
|
36 |
|
37 |
+
with st.spinner('Транскрибируем...'):
|
38 |
+
# Load model
|
39 |
+
model = whisperx.load_model(os.getenv('WHISPER_MODEL_SIZE'), device, compute_type=compute_type)
|
40 |
+
# Load and transcribe audio
|
41 |
+
audio = whisperx.load_audio(temp_file_path)
|
42 |
+
result = model.transcribe(audio, batch_size=batch_size, language="ru")
|
43 |
+
print('Transcribed, now aligning')
|
44 |
|
45 |
+
model_a, metadata = whisperx.load_align_model(language_code=result["language"], device=device)
|
46 |
+
result = whisperx.align(result["segments"], model_a, metadata, audio, device, return_char_alignments=False)
|
47 |
+
print('Aligned, now diarizing')
|
48 |
|
49 |
+
diarize_model = whisperx.DiarizationPipeline(use_auth_token=st.secrets["HF_TOKEN"], device=device)
|
50 |
+
diarize_segments = diarize_model(audio)
|
51 |
+
result_diar = whisperx.assign_word_speakers(diarize_segments, result)
|
52 |
|
53 |
+
st.write("Результат транскрибации:")
|
54 |
+
transcript = convert_segments_object_to_text(result_diar)
|
55 |
+
st.text(transcript)
|
56 |
|
57 |
+
access_token = get_access_token()
|
58 |
|
59 |
+
if (enable_summarization):
|
60 |
+
with st.spinner('Обрабатываем транскрибацию...'):
|
61 |
+
transcript = get_completion_from_gigachat(giga_processing_prompt + transcript, 32768, access_token)
|
62 |
|
63 |
+
st.write("Результат обработки:")
|
64 |
+
st.text(transcript)
|
65 |
|
66 |
|
67 |
|
68 |
+
with st.spinner('Резюмируем...'):
|
69 |
+
summary_answer = get_completion_from_gigachat(giga_base_prompt + transcript, giga_max_tokens, access_token)
|
70 |
|
71 |
+
st.write("Результат резюмирования:")
|
72 |
+
st.text(summary_answer)
|