Update services/model_service.py

services/model_service.py

@@ -1,4 +1,3 @@
-# services/model_service.py
 from transformers import AutoTokenizer, AutoModelForCausalLM, LlamaConfig
 from sentence_transformers import SentenceTransformer
 import torch
@@ -28,15 +27,13 @@ class ModelService:
             # Load tokenizer
             self.tokenizer = AutoTokenizer.from_pretrained(settings.MODEL_NAME)
 
-            # Load model configuration and modify rope_scaling if applicable
+            # Load model configuration without modifying rope_scaling
             config = LlamaConfig.from_pretrained(settings.MODEL_NAME)
 
+            # Remove rope_scaling if present
             if hasattr(config, "rope_scaling"):
-                logger.info("Updating rope_scaling configuration...")
-                config.rope_scaling = {
-                    "type": "linear",  # Ensure the type is valid
-                    "factor": 32.0    # Ensure factor is a valid float
-                }
+                logger.info("Removing rope_scaling from configuration...")
+                config.rope_scaling = None
 
             # Load model with the updated configuration
             self.model = AutoModelForCausalLM.from_pretrained(