Revert "fix: llm_judge resume from breakpoint when judging" (lm-sys#2334

)

fastchat/llm_judge/gen_judgment.py

@@ -288,38 +288,14 @@ def make_judge_single(judge_model, judge_prompts):
         multi_turn=True,
     )
 
-    # Filter out existed matches
-    total_num_matches = len(matches)
-    filtered_matches = []
-    try:
-        with open(output_file, "r") as f:
-            existed_matches = [json.loads(line) for line in f]
-    except FileNotFoundError:
-        existed_matches = []
-    uniq_ids = set(
-        [
-            f"{e['question_id']}_{e['model']}_{e['judge'][0]}_{e['judge'][1]}_{e['turn']}"
-            for e in existed_matches
-        ]
-    )
-    for match in matches:
-        turn = 2 if match.judge.multi_turn else 1
-        uniq_id = f"{match.question['question_id']}_{match.answer['model_id']}_{match.judge.model_name}_{match.judge.prompt_template['name']}_{turn}"
-        if uniq_id in uniq_ids:
-            print(f"Skip {uniq_id}")
-        else:
-            filtered_matches.append(match)
-    matches = filtered_matches
-
     match_stat = {}
     match_stat["bench_name"] = args.bench_name
     match_stat["mode"] = args.mode
     match_stat["judge"] = args.judge_model
     match_stat["baseline"] = baseline_model
     match_stat["model_list"] = models
     match_stat["total_num_questions"] = len(questions)
-    match_stat["total_num_matches"] = total_num_matches
-    match_stat["current_num_matches"] = len(matches)
+    match_stat["total_num_matches"] = len(matches)
     match_stat["output_path"] = output_file
 
     # Show match stats and prompt enter to continue