fix lm_eval import [3x](#1708)

Signed-off-by: Kaihui-intel <[email protected]>

Author: Kaihui-intel

examples/3.x_api/pytorch/nlp/huggingface_models/language-modeling/quantization/llm/run_clm_no_trainer.py

@@ -337,7 +337,7 @@ def run_fn_for_gptq(model, dataloader_for_calibration, *args):
 
 if args.accuracy:
     user_model.eval()
-    from intel_extension_for_transformers.llm.evaluation.lm_eval import evaluate
+    from intel_extension_for_transformers.transformers.llm.evaluation.lm_eval import evaluate
 
     results = evaluate(
         model="hf-causal",
@@ -360,7 +360,7 @@ def run_fn_for_gptq(model, dataloader_for_calibration, *args):
 
 if args.performance:
     user_model.eval()
-    from intel_extension_for_transformers.llm.evaluation.lm_eval import evaluate
+    from intel_extension_for_transformers.transformers.llm.evaluation.lm_eval import evaluate
     import time
 
     samples = args.iters * args.batch_size