Jeongeun commited on
Commit
65ba06a
·
verified ·
1 Parent(s): 02c1d39

README: eval assets

Browse files
Files changed (1) hide show
  1. README.md +21 -0
README.md CHANGED
@@ -70,3 +70,24 @@ Episodes were generated in RoboCasa kitchens: a floating-hand pool per task and
70
  GR1, then selected into the final per-task sets. The selection provenance is kept per demo
71
  (`v6_origin`) in the source HDF5s, which are **not** part of this upload — this repository is the
72
  training-ready half (manifest + zarr).
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
70
  GR1, then selected into the final per-task sets. The selection provenance is kept per demo
71
  (`v6_origin`) in the source HDF5s, which are **not** part of this upload — this repository is the
72
  training-ready half (manifest + zarr).
73
+
74
+ ## Eval assets (`eval/`)
75
+
76
+ These are the files the GR1 one-shot closed-loop eval (`gr1_local_eval.sh` in
77
+ [jeongeun980906/watch](https://github.com/jeongeun980906/watch)) needs in addition to the training data:
78
+
79
+ ```
80
+ eval/taskdemo_l2l3.tar # 25 one-shot taskdemo zarr episodes for the L2/L3 tasks (unpack into egoverse/)
81
+ eval/pools/taskdemo_pool_v6.npz, eval/pools/dth_pool_1x_v6.npz # eval_gr1_server --pool / --play-pool
82
+ eval/sha256.txt
83
+ ```
84
+
85
+ The 12 seen-task `taskdemo_*` episodes are already in `data/`. The L2/L3 ones were built from the TaskDemos
86
+ hdf5 of HF `Rutav/MimicDroidDataset`, with the same converter as the seen ones. The rebuilt seen-task
87
+ episodes match the ones in `data/`: the arrays are identical, the pixels differ by at most 2, and the
88
+ attributes are identical. The pools are built from these episodes. With `--oneshot-pool3` the server loads
89
+ them at startup but does not use them.
90
+
91
+ ```bash
92
+ tar -xf eval/taskdemo_l2l3.tar -C egoverse # next to the episodes unpacked from data/
93
+ ```