add transformer get transcript

app.py

@@ -3,7 +3,7 @@ import os
 from fastapi import FastAPI
 from fastapi.middleware.cors import CORSMiddleware
 
-from routers import get_transcript, get_chatrespone
+from routers import get_transcript, get_chatrespone, get_transcript_transformer
 
 os.environ['HF_HOME'] = "./cached/"
 
@@ -13,6 +13,7 @@ app.add_middleware(CORSMiddleware, allow_origins=["*"], allow_credentials=True,
 
 app.include_router(get_transcript.router)
 app.include_router(get_chatrespone.router)
+app.include_router(get_transcript_transformer.router)
 
 @app.get("/")
 def read_root():

libs/transformer/get_transcript.py

@@ -0,0 +1,31 @@
+import torch
+from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
+
+
+
+device = "cuda:0" if torch.cuda.is_available() else "cpu"
+torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
+
+model_id = "distil-whisper/distil-large-v3"
+
+model = AutoModelForSpeechSeq2Seq.from_pretrained(
+    model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True
+)
+model.to(device)
+
+processor = AutoProcessor.from_pretrained(model_id)
+
+pipe = pipeline(
+    "automatic-speech-recognition",
+    model=model,
+    tokenizer=processor.tokenizer,
+    feature_extractor=processor.feature_extractor,
+    max_new_tokens=128,
+    torch_dtype=torch_dtype,
+    device=device,
+    return_timestamps=True
+)
+
+
+result = pipe("https://static.langkingdom.com/user_playlist_practice_videos/2114103294b5c15605fd59773e948e58.mp3")
+print(result)

libs/transformer/get_transcript_2.py

@@ -0,0 +1,12 @@
+from transformers import pipeline
+
+def get_transcribe_transformers(url:str):
+    pipe = pipeline("automatic-speech-recognition", model="distil-whisper/distil-large-v3", return_timestamps=True)
+
+    result = pipe(url)
+
+    return result.get("text"), result.get("chunks")
+
+text, chunks = get_transcribe_transformers("https://static.langkingdom.com/user_playlist_practice_videos/0a871d06c1e8e1174da09d5aad6ec550.x-m4a")
+
+print(chunks[0].get("timestamp")[1])

routers/get_chatrespone.py

@@ -24,7 +24,7 @@ class ChatInputForm(BaseModel):
     prompt: str
 
 @router.post("/")
-async def get_chat_respone(body: ChatInputForm):
+async def get_chat_respone(body: ChatInputForm, api_key: str = Depends(get_api_key)):
 
     prompt = get_prompt(body.prompt)
 
@@ -57,17 +57,16 @@ def get_response(response: Iterator[str]):
         yield chunk
 
 
-checkWritting = """You'll be provided with a text: {prompt}
+checkWritting = """You'll be provided with a text. Convert the text to standard English.
 ---------------
 IMPORTANT:
 - If the text is empty, do nothing.
 - If the given text maintains grammatical accuracy, no suggestions are needed.
 - If the given text is empty, do nothing.
 - If the given text contains any errors in grammatical accuracy, provide the corrected text.
-
 """
 
-template = """You are a helpful English teacher. Chat and do user requirement.
+template = """You are a helpful assistant. Do whatever user require. Response in markdown format.
     Answer: Let's think step by step."""
 baiGiang = """Provide the given phrase in English. Provide the correct and popularly used English phrase along with its American IPA pronunciation and a brief explanation for it. Use the correct English phrase to create 4 example sentences along with the example IPA and brief meanings. Finally, suggest 4 similar English phrases with the correct English version, along with American IPA and their brief meanings.
 Provie your response in markdown format"""

routers/get_transcript.py

@@ -105,4 +105,4 @@ def get_transcript(audio_path: str, model_size: str = "distil-large-v3", api_key
 #             "text": " She can't help but smile wider."
 #         }
 #     ]
-# }
+# }

routers/get_transcript_transformer.py

@@ -0,0 +1,35 @@
+import time
+from fastapi import APIRouter, Depends, HTTPException, status
+
+from libs.transformer.get_transcript_2 import get_transcribe_transformers
+
+from libs.header_api_auth import get_api_key
+
+
+router = APIRouter(prefix="/get-transcript-transformer", tags=["transcript"])
+
+@router.get("/")
+def get_transcript(audio_path: str, api_key: str = Depends(get_api_key)):
+    st = time.time()
+
+    try:
+        text, chunks = get_transcribe_transformers(audio_path)
+    except Exception as error:
+        raise HTTPException(status_code=status.HTTP_403_FORBIDDEN, detail=f"error>>>: {error}")
+
+    listSentences = []
+
+    for chunk in chunks:
+        listSentences.append({
+            "start_time": chunk.get("timestamp")[0],
+            "end_time": chunk.get("timestamp")[1],
+            "text": chunk.get("text")
+        })
+
+    et = time.time()
+    elapsed_time = et - st
+
+    return {"text": text,
+            'list_sentence':  listSentences,
+            'elapsed_time': round(elapsed_time, 2)
+            }