Oct 13 23 days ago 12:51 Cite the paper with all five authors and link OpenReview Aug 27 13:48 Upload Craftax ReMDM planner checkpoints and results (part 2) 13:48 Upload Craftax ReMDM planner checkpoints and results 13:25 Upload Craftax ReMDM planner checkpoints and results (part 2) 13:25 Upload Craftax ReMDM planner checkpoints and results 05:01 Upload Craftax ReMDM planner checkpoints and results 01:41 Upload Craftax ReMDM planner checkpoints and results 01:40 Upload Craftax ReMDM planner checkpoints and results Aug 22 19:57 Upload Craftax ReMDM planner checkpoints and results 19:56 remove superseded seed-0 inference results 19:56 remove superseded seed-0 inference results 19:56 remove superseded seed-0 release: experiments/rl_finetuning/outputs/craftax_classic_ablations 19:56 remove superseded seed-0 release: checkpoints/online/Craftax-Classic-Symbolic-v1-Online-Diffusion-DAgger-100M 19:56 remove superseded seed-0 release: checkpoints/offline/Craftax-Classic-Symbolic-v1-Offline-Diffusion-BC-100M Aug 17 18:34 Rename the offline step dir to the frame-denominated unit The released Classic offline BC checkpoint's step subdirectory read 1000000000, from a convention that predates the current one. The canonical unit is the resolved env-frame budget, which this run's own resume_metadata.json records as OFFLINE_TOTAL_TIMESTEPS = 99942400 (1525 updates x 512 envs x 128 steps).
Contents are unchanged. Restoring from the renamed directory returns the identical parameter tree: 113 leaves, 9,334,289 parameters, checksum 3324.268310546875, verified before and after. Orbax resolves the latest step by directory name and the step number appears in no file inside the checkpoint, so loading is unaffected apart from the reported step. Aug 13 21:21 Delete experiments/rl_finetuning/outputs/retrain_fix1_classic 21:21 Delete experiments/rl_finetuning/outputs/calibration 21:20 Delete checkpoints/offline/Minihack-Offline-Diffusion-BC-100M 21:20 Delete checkpoints/online/Minihack-Online-Diffusion-DAgger-100M Aug 12 22:28 Upload Craftax ReMDM planner checkpoints and results (part 2) 22:28 Upload Craftax ReMDM planner checkpoints and results 21:46 Upload Craftax ReMDM planner checkpoints and results 21:37 Upload Craftax ReMDM planner checkpoints and results (part 2) 21:37 Upload Craftax ReMDM planner checkpoints and results 18:00 Upload MiniHack ReMDM planner checkpoints and results 16:55 Upload Craftax ReMDM planner checkpoints and results (part 2) 16:55 Upload Craftax ReMDM planner checkpoints and results Aug 11 02:04 Delete checkpoints/offline/Craftax-Symbolic-v1-OfflineDiffusion-BC-100M 02:04 Delete checkpoints/offline/Craftax-Classic-Symbolic-v1-OfflineDiffusion-BC-100M 02:04 Delete checkpoints/online/Craftax-Symbolic-v1-OnlineDiffusion-DAgger-100M 02:04 Delete checkpoints/online/Craftax-Classic-Symbolic-v1-OnlineDiffusion-DAgger-100M Aug 10 15:27 Upload Craftax ReMDM planner and PPO expert checkpoints 15:19 Upload Craftax ReMDM planner and PPO expert checkpoints Aug 7 11:40 Upload Craftax ReMDM planner and PPO expert checkpoints Mathis Weil
11:40 initial commit Mathis Weil
