Upload 8 files

.gitattributes

@@ -33,5 +33,3 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
-models/final-gemma2b_SA-Q5_K.gguf filter=lfs diff=lfs merge=lfs -text
-models/final-gemma2b_SA-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text

Dockerfile

@@ -10,9 +10,10 @@ RUN pip install -r requirements.txt
 
 COPY app /app
 
-COPY models /models
+COPY models /models 
+#DO NOT FORGET TO UNCOMMENT THE ABOVE WHEN PUSHING TO HF!!!!
 
-EXPOSE 7860
+# EXPOSE 7860
 
 ENV PYTHONUNBUFFERED=1
 

app/main.py

@@ -5,6 +5,11 @@ from os.path import isdir
 from fastapi import FastAPI, HTTPException, Request, responses
 from fastapi.middleware.cors import CORSMiddleware
 from llama_cpp import Llama
+from fastapi import Body
+
+from pydantic import BaseModel
+from enum import Enum
+from typing import Optional
 
 print("Loading model...")
 llm = Llama(
@@ -28,6 +33,15 @@ def check_sentiment(text):
   result = ask(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}] =', max_new_tokens=3)
   return result['choices'][0]['text'].strip()
 
+def clean_sentiment_response(response_text):
+  result = response_text.strip()
+  if "positive" in result:
+    return "positive"
+  elif "negative" in result:
+    return "negative"
+  else:
+    return "unknown"
+  
 print("Testing model...")
 assert "positive" in check_sentiment("ดอกไม้ร้านนี้สวยจัง")
 print("Ready.")
@@ -47,6 +61,18 @@ app.add_middleware(
     allow_headers=["*"]
 )
 
+class SA_Result(str, Enum):
+  positive = "positive"
+  negative = "negative"
+  unknown = "unknown"
+
+class SA_Response(BaseModel):
+  text: Optional[str] = None
+  code: int = 200
+  result: SA_Result = None
+
+
+
 @app.get('/')
 def docs():
   "Redirects the user from the main page to the docs."
@@ -56,17 +82,16 @@ def docs():
 def add(a: int,b: int):
     return a + b
 
-@app.get('/SA')
-def perform_sentiment_analysis(request: Request):
-    """Performs a sentiment analysis using a finetuned version of Gemma-7b"""
-    prompt = request.query_params.get('prompt')
-    if prompt:
-       try:
-        print(f"Checking sentiment for {prompt}")
-        result = check_sentiment(prompt)
-        print(f"Result: {result}")
-        return {'success': True, 'result': result}
-       except Exception as e:
-          return HTTPException(500, str(e))
-    else:
-       return HTTPException(400, "Request argument 'prompt' not provided.")
+@app.post('/SA')
+def perform_sentiment_analysis(prompt: str = Body(..., embed=True, example="I like eating fried chicken")) -> SA_Response:
+  """Performs a sentiment analysis using a finetuned version of Gemma-7b"""
+  if prompt:
+    try:
+      print(f"Checking sentiment for {prompt}")
+      result = clean_sentiment_response(check_sentiment(prompt))
+      print(f"Result: {result}")
+      return SA_Response(result=result, text=prompt)
+    except Exception as e:
+      return HTTPException(500, SA_Response(code=500, result=str(e), text=prompt))
+  else:
+    return HTTPException(400, SA_Response(code=400, result="Request argument 'prompt' not provided."))

doDockerStuff.sh

@@ -0,0 +1,10 @@
+#!/bin/bash
+
+# Build the Docker image
+docker build -t gemmasa-2bi -f Dockerfile .
+
+# The volumes thing
+docker volume create --opt type=none --opt o=bind --opt device=/home/pawin/Code/GemmaSA_2B/models models
+
+# Run the Docker container
+docker compose up --force-recreate --remove-orphans

docker-compose.yml

@@ -0,0 +1,11 @@
+version: '3.0'
+services:
+  gemmasa-2bc: # Container's name
+    image: "gemmasa-2bi" # Image's name
+    container_name: "gemmasa-2bc" # Container's name
+    ports:
+     - 7860:7860
+    volumes:
+     - /home/pawin/Code/GemmaSA_2B/models:/models/
+# volumes:
+#   - //D/git/oppabank/model_folder:

sa-app.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:672442dbd59d24f744e06e9b5c67f687ce4578729e1f7c1db3619fb660951db5
+size 5848