Switch to smaller model and reduce memory limits

src/config.py

@@ -15,18 +15,17 @@ class Config:
     # API Settings
     API_VERSION = os.getenv("API_VERSION", "v1")
     API_TITLE = os.getenv("API_TITLE", "Code Review Assistant API")
-    API_DESCRIPTION = os.getenv("API_DESCRIPTION", "An automated code review system powered by Gemma-2b-it")
+    API_DESCRIPTION = os.getenv("API_DESCRIPTION", "An automated code review system powered by Gemma")
     HOST = os.getenv("HOST", "0.0.0.0")
     PORT = int(os.getenv("PORT", 8000))
     DEBUG = os.getenv("DEBUG", "false").lower() == "true"
 
     # Model Settings
-    MODEL_NAME = os.getenv("MODEL_NAME", "google/gemma-2-2b-it")
-    MAX_INPUT_LENGTH = int(os.getenv("MAX_INPUT_LENGTH", 2048))
-    MAX_OUTPUT_LENGTH = int(os.getenv("MAX_OUTPUT_LENGTH", 1024))
+    MODEL_NAME = os.getenv("MODEL_NAME", "google/gemma-2b-it")  # Using 2B parameter model instead of 7B
+    MAX_INPUT_LENGTH = int(os.getenv("MAX_INPUT_LENGTH", 512))  # Reduced from 2048
+    MAX_OUTPUT_LENGTH = int(os.getenv("MAX_OUTPUT_LENGTH", 512))  # Reduced from 1024
     TEMPERATURE = float(os.getenv("TEMPERATURE", 0.7))
     TOP_P = float(os.getenv("TOP_P", 0.95))
-    # Token must be set in environment variables
     HUGGING_FACE_TOKEN = os.getenv("HUGGING_FACE_TOKEN")
 
     # Database Settings
@@ -45,8 +44,14 @@ class Config:
     LOG_FILE = os.getenv("LOG_FILE", "logs/code_review.log")
 
     # Review History Settings
-    MAX_HISTORY_ITEMS = int(os.getenv("MAX_HISTORY_ITEMS", 1000))
-    HISTORY_RETENTION_DAYS = int(os.getenv("HISTORY_RETENTION_DAYS", 30))
+    MAX_HISTORY_ITEMS = int(os.getenv("MAX_HISTORY_ITEMS", 100))  # Reduced from 1000
+    HISTORY_RETENTION_DAYS = int(os.getenv("HISTORY_RETENTION_DAYS", 7))  # Reduced from 30
+
+    # Memory Settings
+    MAX_MEMORY_USAGE = "2GB"  # Limit memory usage
+    BATCH_SIZE = 1
+    USE_CACHE = False
+    USE_GRADIENT_CHECKPOINTING = True
 
     @staticmethod
     def validate():
@@ -58,6 +63,8 @@ class Config:
         logger.info(f"HOST: {Config.HOST}")
         logger.info(f"PORT: {Config.PORT}")
         logger.info(f"DEBUG: {Config.DEBUG}")
+        logger.info(f"MAX_INPUT_LENGTH: {Config.MAX_INPUT_LENGTH}")
+        logger.info(f"MAX_OUTPUT_LENGTH: {Config.MAX_OUTPUT_LENGTH}")
 
         if not Config.HUGGING_FACE_TOKEN:
             raise ValueError("HUGGING_FACE_TOKEN must be set in environment variables to access Hugging Face models.")