Update app.py
Browse files
app.py
CHANGED
@@ -6,11 +6,6 @@ import librosa
|
|
6 |
from moviepy.editor import VideoFileClip
|
7 |
import os
|
8 |
import tempfile
|
9 |
-
import logging
|
10 |
-
|
11 |
-
# Set up logging
|
12 |
-
logging.basicConfig(level=logging.INFO)
|
13 |
-
logger = logging.getLogger(__name__)
|
14 |
|
15 |
# Load Whisper base model and processor
|
16 |
whisper_model_name = "openai/whisper-base"
|
@@ -20,26 +15,8 @@ whisper_model = WhisperForConditionalGeneration.from_pretrained(whisper_model_na
|
|
20 |
# Load RAG sequence model and tokenizer
|
21 |
rag_model_name = "facebook/rag-sequence-nq"
|
22 |
rag_tokenizer = RagTokenizer.from_pretrained(rag_model_name)
|
23 |
-
|
24 |
-
|
25 |
-
try:
|
26 |
-
rag_retriever = RagRetriever.from_pretrained(
|
27 |
-
rag_model_name,
|
28 |
-
index_name="exact",
|
29 |
-
use_dummy_dataset=True,
|
30 |
-
dataset_path=local_dataset_path,
|
31 |
-
trust_remote_code=True
|
32 |
-
)
|
33 |
-
logger.info("Successfully loaded RagRetriever with trust_remote_code=True")
|
34 |
-
except ValueError as e:
|
35 |
-
logger.error(f"Error loading RagRetriever: {e}")
|
36 |
-
rag_retriever = None
|
37 |
-
|
38 |
-
if rag_retriever is not None:
|
39 |
-
rag_model = RagSequenceForGeneration.from_pretrained(rag_model_name, retriever=rag_retriever)
|
40 |
-
else:
|
41 |
-
logger.error("RagRetriever is not available, unable to proceed with loading RAG model.")
|
42 |
-
st.error("RagRetriever is not available, unable to proceed with loading RAG model.")
|
43 |
|
44 |
def transcribe_audio(audio_path, language="ru"):
|
45 |
speech, rate = librosa.load(audio_path, sr=16000)
|
@@ -50,8 +27,6 @@ def transcribe_audio(audio_path, language="ru"):
|
|
50 |
return transcription
|
51 |
|
52 |
def translate_and_summarize(text):
|
53 |
-
if rag_retriever is None:
|
54 |
-
return ["Translation and summarization feature is not available due to RAG retriever loading issue."]
|
55 |
inputs = rag_tokenizer(text, return_tensors="pt")
|
56 |
input_ids = inputs["input_ids"]
|
57 |
attention_mask = inputs["attention_mask"]
|
|
|
6 |
from moviepy.editor import VideoFileClip
|
7 |
import os
|
8 |
import tempfile
|
|
|
|
|
|
|
|
|
|
|
9 |
|
10 |
# Load Whisper base model and processor
|
11 |
whisper_model_name = "openai/whisper-base"
|
|
|
15 |
# Load RAG sequence model and tokenizer
|
16 |
rag_model_name = "facebook/rag-sequence-nq"
|
17 |
rag_tokenizer = RagTokenizer.from_pretrained(rag_model_name)
|
18 |
+
rag_retriever = RagRetriever.from_pretrained(rag_model_name, index_name="exact", use_dummy_dataset=True, trust_remote_code=True)
|
19 |
+
rag_model = RagSequenceForGeneration.from_pretrained(rag_model_name, retriever=rag_retriever)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
|
21 |
def transcribe_audio(audio_path, language="ru"):
|
22 |
speech, rate = librosa.load(audio_path, sr=16000)
|
|
|
27 |
return transcription
|
28 |
|
29 |
def translate_and_summarize(text):
|
|
|
|
|
30 |
inputs = rag_tokenizer(text, return_tensors="pt")
|
31 |
input_ids = inputs["input_ids"]
|
32 |
attention_mask = inputs["attention_mask"]
|