Sep 20

1. 08:06

   Remove stray __pycache__ artifact

   Ronaldo-GOAT
   Verified signature

   7b41ce6
2. 08:06

   Fix meta/info.json: chunks_size 1000 -> 1921 (total_episodes+1) so every episode maps to chunk-000

   Ronaldo-GOAT
   Verified signature

   07cf02c
3. 08:06

   subset_960.json: update provenance fields

   Ronaldo-GOAT
   Verified signature

   2ccf6b3
4. 08:06

   SUBSET_960.md: self-contained subset instructions

   Ronaldo-GOAT
   Verified signature

   d90022e
5. 08:06

   Merged README: dataset card + method + reproduce + resume guide

   Ronaldo-GOAT
   Verified signature

   3e29d69
6. 08:06

   Add training configs

   Ronaldo-GOAT
   Verified signature

   c69de0f
7. 08:06

   Add resume bundle: RESUME.md, env spec, launch scripts, subset builder + subset meta

   Ronaldo-GOAT
   Verified signature

   7ccc1a6
8. 08:01

   Add checkpoint-10000 (960-subset run) with full trainer state for exact resume

   Ronaldo-GOAT
   Verified signature

   ee9f047
9. 08:01

   Add checkpoint-30000-base (base policy: GR00T-N1.5 + 120 real episodes, 30k steps)

   Ronaldo-GOAT
   Verified signature

   7f76f73
10. 07:57

    960-episode subset index into the 1920-episode jh_data_aug

    Ronaldo-GOAT
    Verified signature

    39c4616
11. 04:13

    jh_data_aug 1920 buf10 abs-joint

    Ronaldo-GOAT
    Verified signature

    7ac8cd8
12. 04:12

    initial commit

    Ronaldo-GOAT
    Verified signature

    a7a8296