diff options
| author | YurenHao0426 <blackhao0426@gmail.com> | 2026-02-10 20:16:36 +0000 |
|---|---|---|
| committer | YurenHao0426 <blackhao0426@gmail.com> | 2026-02-10 20:16:36 +0000 |
| commit | 5626080ca4c4219aec4888d6b9406d0d3349fb55 (patch) | |
| tree | 86287d9fd5833e11ccd78566992540f2664fd195 /src/personalization/feedback/local_llm_reward.py | |
| parent | a2036838807428424bbbaff507a6563749a83145 (diff) | |
Add RAG rewrite, 60-session experiment scripts, and analysis tools
- RAG rewrite adapter and vector preference pipeline in personalized_llm
- 60-session experiment queue scripts (reflection, rag, rag_vector, rag_rewrite)
- Vector-preference correlation analysis and visualization scripts
- Local reward model batch processing improvements
- Updated CLAUDE.md with full experiment documentation and notes
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Diffstat (limited to 'src/personalization/feedback/local_llm_reward.py')
| -rw-r--r-- | src/personalization/feedback/local_llm_reward.py | 36 |
1 files changed, 32 insertions, 4 deletions
diff --git a/src/personalization/feedback/local_llm_reward.py b/src/personalization/feedback/local_llm_reward.py index 9837ff0..70bbeb8 100644 --- a/src/personalization/feedback/local_llm_reward.py +++ b/src/personalization/feedback/local_llm_reward.py @@ -307,11 +307,39 @@ class LocalLLMRewardClient: This is the main entry point for batch reward estimation. """ - return asyncio.run(self.judge_batch_async(samples)) - - def judge(self, sample: TurnSample) -> RewardResult: + try: + loop = asyncio.get_running_loop() + except RuntimeError: + loop = None + + if loop is not None: + # Already in an event loop - create a new thread to run the coroutine + import concurrent.futures + with concurrent.futures.ThreadPoolExecutor() as executor: + future = executor.submit(asyncio.run, self.judge_batch_async(samples)) + return future.result() + else: + return asyncio.run(self.judge_batch_async(samples)) + + async def judge(self, sample: TurnSample) -> RewardResult: + """Judge a single turn (async interface for compatibility with LLMRewardClient).""" + return await self.judge_async(sample) + + def judge_sync(self, sample: TurnSample) -> RewardResult: """Judge a single turn (sync wrapper).""" - return asyncio.run(self.judge_async(sample)) + try: + loop = asyncio.get_running_loop() + except RuntimeError: + loop = None + + if loop is not None: + # Already in an event loop - create a new thread to run the coroutine + import concurrent.futures + with concurrent.futures.ThreadPoolExecutor() as executor: + future = executor.submit(asyncio.run, self.judge_async(sample)) + return future.result() + else: + return asyncio.run(self.judge_async(sample)) # --- Convenience Functions --- |
