From d956fc0291cef0bd7abcf25f326603338501640d Mon Sep 17 00:00:00 2001 From: N!no Date: Sat, 4 Jul 2026 12:47:21 -0400 Subject: [PATCH 1/2] add cen explore --- configs/madpo_iter_che_config.yaml | 3 +++ configs/marlhf_iter_che_config.yaml | 3 +++ 2 files changed, 6 insertions(+) diff --git a/configs/madpo_iter_che_config.yaml b/configs/madpo_iter_che_config.yaml index 04aad19..4f3b9f4 100644 --- a/configs/madpo_iter_che_config.yaml +++ b/configs/madpo_iter_che_config.yaml @@ -38,6 +38,7 @@ madpo_iter: preference_replay_lambda: 0.8 preference_replay_sample_size: 264 log_reward_distribution: true + log_comparator_generations: false policy_checkpoint_dir: null num_train_epochs: 1 agent_learning_rate: 2.0e-05 @@ -45,6 +46,8 @@ madpo_iter: num_generations: 4 preference_num_candidates: 20 comparator_policy: current + comparator_generation_mode: decentralized + comparator_centralized_agent_index: 0 comparator_model_name: null comparator_agents: null comparator_devices: null diff --git a/configs/marlhf_iter_che_config.yaml b/configs/marlhf_iter_che_config.yaml index da69184..25af07c 100644 --- a/configs/marlhf_iter_che_config.yaml +++ b/configs/marlhf_iter_che_config.yaml @@ -38,6 +38,7 @@ marlhf_iter: preference_replay_lambda: 0.8 preference_replay_sample_size: 264 log_reward_distribution: true + log_comparator_generations: false preference_scoring_reward: task policy_checkpoint_dir: null num_train_epochs: 2 @@ -46,6 +47,8 @@ marlhf_iter: num_generations: 4 preference_num_candidates: 20 comparator_policy: current + comparator_generation_mode: decentralized + comparator_centralized_agent_index: 0 comparator_model_name: null comparator_agents: null comparator_devices: null From 9a2e4dc9e56b3b01d6ee02e556225452d9027ea0 Mon Sep 17 00:00:00 2001 From: N!no Date: Sat, 4 Jul 2026 15:46:30 -0400 Subject: [PATCH 2/2] rm log of raw comparator api input --- configs/madpo_iter_che_config.yaml | 1 - configs/marlhf_iter_che_config.yaml | 1 - 2 files changed, 2 deletions(-) diff --git a/configs/madpo_iter_che_config.yaml b/configs/madpo_iter_che_config.yaml index 4f3b9f4..d7e3243 100644 --- a/configs/madpo_iter_che_config.yaml +++ b/configs/madpo_iter_che_config.yaml @@ -38,7 +38,6 @@ madpo_iter: preference_replay_lambda: 0.8 preference_replay_sample_size: 264 log_reward_distribution: true - log_comparator_generations: false policy_checkpoint_dir: null num_train_epochs: 1 agent_learning_rate: 2.0e-05 diff --git a/configs/marlhf_iter_che_config.yaml b/configs/marlhf_iter_che_config.yaml index 25af07c..163ec3a 100644 --- a/configs/marlhf_iter_che_config.yaml +++ b/configs/marlhf_iter_che_config.yaml @@ -38,7 +38,6 @@ marlhf_iter: preference_replay_lambda: 0.8 preference_replay_sample_size: 264 log_reward_distribution: true - log_comparator_generations: false preference_scoring_reward: task policy_checkpoint_dir: null num_train_epochs: 2