Fix errors on results

svc/router.py

@@ -2,12 +2,15 @@ from fastapi import APIRouter, HTTPException, Depends
 import logging
 
 from lm_eval import evaluator
-from svc.schemas import LMHarnessTaskRequest, LMHarnessTaskResponse
+from svc.schemas import LMHarnessTaskRequest, LMHarnessTaskResponse, LoadModelRequest
 from fastapi.security import OAuth2PasswordRequestForm
 from auth.authentication import get_current_user, create_access_token
 from dotenv import load_dotenv
 import os
 import json
+
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
 router = APIRouter()
 
 logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
@@ -47,7 +50,9 @@ def inference_model(request: LMHarnessTaskRequest, username: str = Depends(get_c
         raise HTTPException(status_code=500, detail=f"lm-harness task execution failed for model: {request.model_args}")
 
     results["config"]["model_dtype"] = request.precision
-    results["config"]["model_name"] = request.model
+
+    model_name = request.model_args.split(",")[0].split("=")[1]
+    results["config"]["model_name"] = model_name
     results["config"]["model_sha"] = request.model_sha
 
     dumped = json.dumps(results, indent=2)
@@ -56,3 +61,13 @@ def inference_model(request: LMHarnessTaskRequest, username: str = Depends(get_c
     logger.info("-------------------results end------------------\n")
     return LMHarnessTaskResponse(results=dumped)
 
+
+# @router.post("/load_model")
+# async def load_model(request: LoadModelRequest):
+#     model = AutoModelForCausalLM.from_pretrained(
+#         request.model_name,
+#         torch_dtype="auto",
+#         device_map="auto"
+#     )
+#     tokenizer = AutoTokenizer.from_pretrained(request.model_name)
+#     return model, tokenizer