Merge 7b47afb757 into 7a7a4517ab

docs/en/advanced_guides/llm_judge.md

@@ -194,6 +194,8 @@ eval_cfg = dict(
         dict_postprocessor=dict(type=generic_llmjudge_postprocess),
     ),
     pred_role='BOT',
+    #num_gpus=1
+    #Add this in case of use local model as judge, adjust to the number of GPUs you wish to use for evaluation.
 )
 
 # Dataset configuration

docs/zh_cn/advanced_guides/llm_judge.md

@@ -194,6 +194,8 @@ eval_cfg = dict(
         dict_postprocessor=dict(type=generic_llmjudge_postprocess),
     ),
     pred_role='BOT',
+    #num_gpus=1
+    #使用本地模型评估时，设置num_gpus为希望在评估阶段使用的GPU数量
 )
 
 # 数据集配置

examples/eval_llm_judge.py

@@ -92,6 +92,7 @@ math_eval_cfg = dict(
         judge_cfg=lmdeploy_qwen2_5_14b_instruct_model[0],
         dict_postprocessor=dict(type=generic_llmjudge_postprocess),
     ),
+    num_gpus=1 # The amount of GPU used for LLM as judge evaluation
 )
 
 # Dataset configuration

opencompass/evaluator/generic_llm_evaluator.py

@@ -10,11 +10,11 @@ from opencompass.openicl.icl_evaluator import BaseEvaluator
 from opencompass.openicl.icl_inferencer import GenInferencer
 from opencompass.openicl.icl_retriever import ZeroRetriever
 from opencompass.registry import (DICT_POSTPROCESSORS, ICL_PROMPT_TEMPLATES,
-                                  TEXT_POSTPROCESSORS)
+                                  TEXT_POSTPROCESSORS, ICL_EVALUATORS)
 from opencompass.utils import build_dataset_from_cfg, build_model_from_cfg
 from opencompass.utils.logging import get_logger
 
-
+@ICL_EVALUATORS.register_module()
 class GenericLLMEvaluator(BaseEvaluator):
     """Generic LLM evaluator.