-
Provenance and deployment notes for step 50000
-
Cosmos-Policy-ALOHA finetuned on DisAss all-7-components, step 50000 (model weights, DCP sharded)
-
Provenance and deployment notes for step 20000
-
Cosmos-Policy-ALOHA finetuned on DisAss all-7-components, step 20000 (model weights, DCP sharded)
-
Remove ram-only checkpoint: trained from scratch (load_path .pt was silently ignored) on 1/7 of the data; README was inaccurate
-
Add checkpoint provenance and deployment notes
-
Cosmos-Policy-ALOHA finetuned on DisAss ram, step 20000 (model weights, DCP sharded)
-
initial commit
