From 5626080ca4c4219aec4888d6b9406d0d3349fb55 Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Tue, 10 Feb 2026 20:16:36 +0000 Subject: 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 --- collaborativeagents/adapters/reflection_grpo_adapter.py | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'collaborativeagents/adapters/reflection_grpo_adapter.py') diff --git a/collaborativeagents/adapters/reflection_grpo_adapter.py b/collaborativeagents/adapters/reflection_grpo_adapter.py index 09c5b26..3c10942 100644 --- a/collaborativeagents/adapters/reflection_grpo_adapter.py +++ b/collaborativeagents/adapters/reflection_grpo_adapter.py @@ -18,10 +18,11 @@ import torch from transformers import AutoModelForCausalLM, AutoTokenizer from json_repair import repair_json -# Model paths - Use GRPO-trained model if available, fallback to base -GRPO_MODEL_PATH = "/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/collaborativeagents/training/outputs/grpo_reflection/final" -SFT_MODEL_PATH = "/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/collaborativeagents/training/outputs/sft_reflection" -DEFAULT_MODEL_PATH = "/projects/bfqt/users/yurenh2/ml-projects/personalization-user-model/models/llama-3.1-8b-instruct" +# Model paths - computed relative to project root +_PROJECT_ROOT = Path(__file__).parent.parent.parent +GRPO_MODEL_PATH = str(_PROJECT_ROOT / "collaborativeagents/training/outputs/grpo_reflection/final") +SFT_MODEL_PATH = str(_PROJECT_ROOT / "collaborativeagents/training/outputs/sft_reflection") +DEFAULT_MODEL_PATH = str(_PROJECT_ROOT / "models/llama-3.1-8b-instruct") def get_best_available_model(): """Get the best available model path (GRPO > SFT > base).""" -- cgit v1.2.3