Upload main.py

backend/app/main.py

@@ -11,13 +11,12 @@ import torch
 APP_TITLE = "HF Chat (Fathom-R1-14B)"
 APP_VERSION = "0.2.0"
 
-# ---- Config via ENV ----
 MODEL_ID = os.getenv("MODEL_ID", "FractalAIResearch/Fathom-R1-14B")
 PIPELINE_TASK = os.getenv("PIPELINE_TASK", "text-generation")
-MAX_INPUT_TOKENS = int(os.getenv("MAX_INPUT_TOKENS", "8192"))  # keep prompt reasonable
+MAX_INPUT_TOKENS = int(os.getenv("MAX_INPUT_TOKENS", "8192"))
 STATIC_DIR = os.getenv("STATIC_DIR", "/app/static")
 ALLOWED_ORIGINS = os.getenv("ALLOWED_ORIGINS", "")
-QUANTIZE = os.getenv("QUANTIZE", "auto")  # auto|4bit|8bit|none
+QUANTIZE = os.getenv("QUANTIZE", "auto")  
 
 app = FastAPI(title=APP_TITLE, version=APP_VERSION)
 
@@ -109,17 +108,13 @@ def messages_to_prompt(messages: List[Message]) -> str:
         parts = []
         for m in messages:
             if m.role == "system":
-                parts.append(f"System: {m.content}
-")
+                parts.append(f"System: {m.content}")
             elif m.role == "user":
-                parts.append(f"User: {m.content}
-")
+                parts.append(f"User: {m.content}")
             else:
-                parts.append(f"Assistant: {m.content}
-")
+                parts.append(f"Assistant: {m.content}")
         parts.append("Assistant:")
-        return "
-".join(parts)
+        return "".join(parts)
 
 def truncate_prompt(prompt: str, max_tokens: int) -> str:
     ids = tokenizer(prompt, return_tensors="pt", truncation=False)["input_ids"][0]