Update app/main.py

app/main.py

@@ -13,9 +13,8 @@ from typing import Optional
 # MODEL LOADING, FUNCTIONS, AND TESTING
 
 print("Loading model...")
-SAllm = Llama(model_path="/models/final-gemma2b_SA-Q8_0.gguf", use_mmap=False, use_mlock=True)
-FIllm = Llama(model_path="/models/final-gemma7b_FI-Q8_0.gguf", use_mmap=False, use_mlock=True)
-# WIllm = Llama(model_path="/models/final-GemmaWild7b-Q8_0.gguf", use_mmap=False, use_mlock=True)
+WIllm = Llama(model_path="/models/final-GemmaWild7b-Q8_0.gguf", use_mmap=False, use_mlock=True)
+COllm = Llama(model_path="/models/TunaCodes-Q8_0.gguf", use_mmap=False, use_mlock=True)
       # n_gpu_layers=28, # Uncomment to use GPU acceleration
       # seed=1337, # Uncomment to set a specific seed
       # n_ctx=2048, # Uncomment to increase the context window
@@ -28,31 +27,18 @@ def ask_llm(llm, question, max_new_tokens=200, temperature=0.5):
   prompt = f"""###User: {question}\n###Assistant:"""
   result = extract_restext(llm(prompt, max_tokens=max_new_tokens, temperature=temperature, stop=["###User:", "###Assistant:"], echo=False))
   return result
-    
-def check_sentiment(text):
-  prompt = f'Analyze the sentiment of the tweet enclosed in square brackets, determine if it is positive or negative, and return the answer as the corresponding sentiment label "positive" or  "negative" [{text}] ='
-  response = SAllm(prompt, max_tokens=3, stop=["\n"], echo=False, temperature=0.5)
-  # print(response)
-  result = extract_restext(response)
-  if "positive" in result:
-    return "positive"
-  elif "negative" in result:
-    return "negative"
-  else:
-    return "unknown"
+
 
 # TESTING THE MODEL
 print("Testing model...")
-assert "positive" in check_sentiment("ดอกไม้ร้านนี้สวยจัง")
-assert ask_llm(FIllm, "Hello!, How are you today?", max_new_tokens=1) #Just checking that it can run
-# assert ask_llm(WIllm, "Hello!, How are you today?", max_new_tokens=1) #Just checking that it can run
+assert ask_llm(WIllm, "Hello!, How are you today?", max_new_tokens=1) #Just checking that it can run
 print("Ready.")
 
 
 # START OF FASTAPI APP
 app = FastAPI(
     title = "Gemma Finetuned API",
-    description="Gemma Finetuned API for Sentiment Analysis and Finance Questions.",
+    description="Gemma Finetuned API for Open-ended and Coding questions.",
     version="1.0.0",
 )
 
@@ -67,10 +53,6 @@ app.add_middleware(
 
 
 # API DATA CLASSES
-class SA_Result(str, Enum):
-  positive = "positive"
-  negative = "negative"
-  unknown = "unknown"
 
 class SAResponse(BaseModel):
   code: int = 200
@@ -90,60 +72,44 @@ def docs():
   "Redirects the user from the main page to the docs."
   return responses.RedirectResponse('./docs')
 
-@app.post('/classifications/sentiment')
-async def perform_sentiment_analysis(prompt: str = Body(..., embed=True, example="I like eating fried chicken")) -> SAResponse:
-  """Performs a sentiment analysis using a finetuned version of Gemma-7b"""
+@app.post('/questions/open-ended')
+async def ask_gemmaWild(
+    prompt: str = Body(..., embed=True, example="Why is ice cream so delicious?"),
+    temperature: float = Body(0.5, embed=True), 
+    max_new_tokens: int = Body(200, embed=True)
+) -> QuestionResponse:
+  """
+  Ask a finetuned Gemma an open-ended question..
+  NOTICE: IT MAY PRODUCE RANDOM/INACCURATE ANSWERS. PLEASE SEEK PROFESSIONAL ADVICE BEFORE DOING ANYTHING SERIOUS.
+  """
   if prompt:
     try:
-      print(f"Checking sentiment for {prompt}")
-      result = check_sentiment(prompt)
+      print(f'Asking GemmaWild with the question "{prompt}"')
+      result = ask_llm(WIllm, prompt, max_new_tokens=max_new_tokens, temperature=temperature)
       print(f"Result: {result}")
-      return SAResponse(result=result, text=prompt)
+      return QuestionResponse(answer=result, question=prompt, config={"temperature": temperature, "max_new_tokens": max_new_tokens})
     except Exception as e:
-      return HTTPException(500, SAResponse(code=500, result=str(e), text=prompt))
+      return HTTPException(500, QuestionResponse(code=500, answer=str(e), question=prompt))
   else:
-    return HTTPException(400, SAResponse(code=400, result="Request argument 'prompt' not provided."))
-
+    return HTTPException(400, QuestionResponse(code=400, answer="Request argument 'prompt' not provided."))
 
-@app.post('/questions/finance')
-async def ask_gemmaFinanceTH(
-    prompt: str = Body(..., embed=True, example="What's the best way to invest my money"),
+@app.post('/questions/coding')
+async def ask_gemmaCode(
+    prompt: str = Body(..., embed=True, example="Why is ice cream so delicious?"),
     temperature: float = Body(0.5, embed=True), 
     max_new_tokens: int = Body(200, embed=True)
 ) -> QuestionResponse:
   """
-  Ask a finetuned Gemma a finance-related question, just for fun.
+  Ask a finetuned Gemma an open-ended question..
   NOTICE: IT MAY PRODUCE RANDOM/INACCURATE ANSWERS. PLEASE SEEK PROFESSIONAL ADVICE BEFORE DOING ANYTHING SERIOUS.
   """
   if prompt:
     try:
-      print(f'Asking GemmaFinance with the question "{prompt}"')
-      result = ask_llm(FIllm, prompt, max_new_tokens=max_new_tokens, temperature=temperature)
+      print(f'Asking GemmaCode with the question "{prompt}"')
+      result = ask_llm(COllm, prompt, max_new_tokens=max_new_tokens, temperature=temperature)
       print(f"Result: {result}")
       return QuestionResponse(answer=result, question=prompt, config={"temperature": temperature, "max_new_tokens": max_new_tokens})
     except Exception as e:
       return HTTPException(500, QuestionResponse(code=500, answer=str(e), question=prompt))
   else:
     return HTTPException(400, QuestionResponse(code=400, answer="Request argument 'prompt' not provided."))
-  
-
-# @app.post('/questions/open-ended')
-# async def ask_gemmaWild(
-#     prompt: str = Body(..., embed=True, example="Why is ice cream so delicious?"),
-#     temperature: float = Body(0.5, embed=True), 
-#     max_new_tokens: int = Body(200, embed=True)
-# ) -> QuestionResponse:
-#   """
-#   Ask a finetuned Gemma an open-ended question..
-#   NOTICE: IT MAY PRODUCE RANDOM/INACCURATE ANSWERS. PLEASE SEEK PROFESSIONAL ADVICE BEFORE DOING ANYTHING SERIOUS.
-#   """
-#   if prompt:
-#     try:
-#       print(f'Asking GemmaWild with the question "{prompt}"')
-#       result = ask_llm(WIllm, prompt, max_new_tokens=max_new_tokens, temperature=temperature)
-#       print(f"Result: {result}")
-#       return QuestionResponse(answer=result, question=prompt, config={"temperature": temperature, "max_new_tokens": max_new_tokens})
-#     except Exception as e:
-#       return HTTPException(500, QuestionResponse(code=500, answer=str(e), question=prompt))
-#   else:
-#     return HTTPException(400, QuestionResponse(code=400, answer="Request argument 'prompt' not provided."))