coffee TopK policy ep500 sr0.60 + tokenizer ep1940
Browse files
README.md
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
tags:
|
| 4 |
+
- robotics
|
| 5 |
+
- oat
|
| 6 |
+
- robocasa
|
| 7 |
+
- coffee_press_button
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
# RoboCasa coffee — OAT TopK policy + tokenizer
|
| 11 |
+
|
| 12 |
+
Handoff artifacts for mentee BoN16/32 + AWR on `coffee_press_button` (branch `robocasa`).
|
| 13 |
+
|
| 14 |
+
## Files
|
| 15 |
+
|
| 16 |
+
| Path | Role |
|
| 17 |
+
|------|------|
|
| 18 |
+
| `policies/robocasa_coffee_press_button_topk_ep0500_sr0.600.ckpt` | BASE_CKPT (TopK @ selection seed 2000, ep 500, SR 0.60) |
|
| 19 |
+
| `policies/robocasa_coffee_press_button_topk_lock.txt` | lock metadata |
|
| 20 |
+
| `tokenizers/coffee_ep-1940_mse-0.003.ckpt` | frozen task tokenizer (loaded via policy) |
|
| 21 |
+
|
| 22 |
+
## Protocol
|
| 23 |
+
|
| 24 |
+
See repo guide `oat/AGENT_GUIDE_ROBOCASA_COFFEE_BON16_AWR.md` in `Mirageinvo/OAT_BLT_research` (`robocasa` branch).
|
| 25 |
+
|
| 26 |
+
- Report seeds: literal 5 × `10000…10004` (`-n 1`, `n_test=50`)
|
| 27 |
+
- AWR: collect `--bon_n 16`, train `--epochs 100`
|
| 28 |
+
- Env: `.venv_robocasa` (robosuite 1.5), `MUJOCO_GL=egl`
|
| 29 |
+
|
| 30 |
+
Wave1 baseline + BoN8 eval logs / zarr: separate dataset repo (not this model repo).
|
policies/robocasa_coffee_press_button_topk_ep0500_sr0.600.ckpt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a634fe9b925d3d074cbea1dcc394f5b6258ecefd7f7d54ee5f52571a9ad145cb
|
| 3 |
+
size 936052104
|
policies/robocasa_coffee_press_button_topk_lock.txt
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
BASE_CKPT=my_models/robocasa_coffee_press_button_topk_ep0500_sr0.600.ckpt
|
| 2 |
+
SOURCE=output/20260724/220823_train_oatpolicy_coffee_press_button_N200/checkpoints/ep-0500_sr-0.600.ckpt
|
| 3 |
+
SR=0.600
|
| 4 |
+
EPOCH=500
|
| 5 |
+
SELECTION_SEED=2000
|
| 6 |
+
LOCKED_AT=2026-07-25T23:06:01+00:00
|
| 7 |
+
NOTE=user stop-fit; Wave1 literal5
|
tokenizers/coffee_ep-1940_mse-0.003.ckpt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c0ace615f465737369f658cf934923ecb6a1c875d080074310eb4906d20404b5
|
| 3 |
+
size 93074890
|