[Feat] Add Cosmos3-Nano LIBERO full finetune checkpoint (bs2048)
#4
by RJK - opened
- .gitattributes +1 -0
- cosmos3_nano_libero_full_finetune_bs2048/checkpoints/step-005000-epoch-22-loss=0.3622.safetensors +3 -0
- cosmos3_nano_libero_full_finetune_bs2048/config.json +1740 -0
- cosmos3_nano_libero_full_finetune_bs2048/config.yaml +1392 -0
- cosmos3_nano_libero_full_finetune_bs2048/cosmos3nano_libero_full_finetune_2026_08_14_18_30_47.jsonl +0 -0
- cosmos3_nano_libero_full_finetune_bs2048/dataset_statistics.json +138 -0
- cosmos3_nano_libero_full_finetune_bs2048/run-metrics.jsonl +1 -0
- cosmos3_nano_libero_full_finetune_bs2048/tokenizer/added_tokens.json +28 -0
- cosmos3_nano_libero_full_finetune_bs2048/tokenizer/chat_template.jinja +120 -0
- cosmos3_nano_libero_full_finetune_bs2048/tokenizer/merges.txt +0 -0
- cosmos3_nano_libero_full_finetune_bs2048/tokenizer/special_tokens_map.json +31 -0
- cosmos3_nano_libero_full_finetune_bs2048/tokenizer/tokenizer.json +3 -0
- cosmos3_nano_libero_full_finetune_bs2048/tokenizer/tokenizer_config.json +239 -0
- cosmos3_nano_libero_full_finetune_bs2048/tokenizer/vocab.json +0 -0
- cosmos3_nano_libero_full_finetune_bs2048/vlm_backbone_config.json +100 -0
.gitattributes
CHANGED
|
@@ -51,3 +51,4 @@ pi05_paligemma_robocasa_full_data_full_finetune_bs256/tokenizer/tokenizer.json f
|
|
| 51 |
fastwam_base/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 52 |
dit4dit_libero_all_full_finetune_bc071270ff2_bs256/dit4dit_libero_all_full_finetune_2026_08_12_20_51_03.jsonl filter=lfs diff=lfs merge=lfs -text
|
| 53 |
dit4dit_libero_all_full_finetune_bc071270ff2_bs256/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 51 |
fastwam_base/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 52 |
dit4dit_libero_all_full_finetune_bc071270ff2_bs256/dit4dit_libero_all_full_finetune_2026_08_12_20_51_03.jsonl filter=lfs diff=lfs merge=lfs -text
|
| 53 |
dit4dit_libero_all_full_finetune_bc071270ff2_bs256/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 54 |
+
cosmos3_nano_libero_full_finetune_bs2048/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
cosmos3_nano_libero_full_finetune_bs2048/checkpoints/step-005000-epoch-22-loss=0.3622.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:418c6bb0215ad0598f6a16c6633a91a0381c905ada647d6d93dc7bb87f915277
|
| 3 |
+
size 63511447112
|
cosmos3_nano_libero_full_finetune_bs2048/config.json
ADDED
|
@@ -0,0 +1,1740 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_action_dim": 10,
|
| 3 |
+
"_action_horizon": 16,
|
| 4 |
+
"_action_lr": 0.00025,
|
| 5 |
+
"_base_lr": 5e-05,
|
| 6 |
+
"_cfg_dropout_rate": 0.1,
|
| 7 |
+
"_ckpt_root": "./checkpoints",
|
| 8 |
+
"_conditioning_fps": 20.0,
|
| 9 |
+
"_cosmos3_nano_ckpt": "./checkpoints/Cosmos3-Nano",
|
| 10 |
+
"_cosmos3_nano_name_mapping": {
|
| 11 |
+
".self_attn.k_norm.": ".self_attn.norm_k.",
|
| 12 |
+
".self_attn.k_norm_moe_gen.": ".self_attn.norm_added_k.",
|
| 13 |
+
".self_attn.k_proj.": ".self_attn.to_k.",
|
| 14 |
+
".self_attn.k_proj_moe_gen.": ".self_attn.add_k_proj.",
|
| 15 |
+
".self_attn.o_proj.": ".self_attn.to_out.",
|
| 16 |
+
".self_attn.o_proj_moe_gen.": ".self_attn.to_add_out.",
|
| 17 |
+
".self_attn.q_norm.": ".self_attn.norm_q.",
|
| 18 |
+
".self_attn.q_norm_moe_gen.": ".self_attn.norm_added_q.",
|
| 19 |
+
".self_attn.q_proj.": ".self_attn.to_q.",
|
| 20 |
+
".self_attn.q_proj_moe_gen.": ".self_attn.add_q_proj.",
|
| 21 |
+
".self_attn.v_proj.": ".self_attn.to_v.",
|
| 22 |
+
".self_attn.v_proj_moe_gen.": ".self_attn.add_v_proj.",
|
| 23 |
+
"action_in_proj.": "action_proj_in.",
|
| 24 |
+
"action_modality_embed": "action_modality_embed",
|
| 25 |
+
"action_out_proj.": "action_proj_out.",
|
| 26 |
+
"time_embedder.mlp.0.": "time_embedder.linear_1.",
|
| 27 |
+
"time_embedder.mlp.2.": "time_embedder.linear_2.",
|
| 28 |
+
"vision_in_proj.projector.": "proj_in.",
|
| 29 |
+
"vision_out_proj.projector.": "proj_out.",
|
| 30 |
+
"vlm_backbone.lm_head.weight": "lm_head.weight",
|
| 31 |
+
"vlm_backbone.model.language_model.embed_tokens.weight": "embed_tokens.weight",
|
| 32 |
+
"vlm_backbone.model.language_model.layers.": "layers.",
|
| 33 |
+
"vlm_backbone.model.language_model.norm.weight": "norm.weight",
|
| 34 |
+
"vlm_backbone.model.language_model.norm_moe_gen.weight": "norm_moe_gen.weight",
|
| 35 |
+
"vlm_backbone.model.visual.blocks.": "blocks.",
|
| 36 |
+
"vlm_backbone.model.visual.deepstack_merger_list.": "deepstack_merger_list.",
|
| 37 |
+
"vlm_backbone.model.visual.merger.": "merger.",
|
| 38 |
+
"vlm_backbone.model.visual.patch_embed.": "patch_embed.",
|
| 39 |
+
"vlm_backbone.model.visual.pos_embed.": "pos_embed."
|
| 40 |
+
},
|
| 41 |
+
"_cosmos3_nano_special_tokens": {
|
| 42 |
+
"end_of_generation": 151653,
|
| 43 |
+
"eos_token_id": 151645,
|
| 44 |
+
"start_of_generation": 151652
|
| 45 |
+
},
|
| 46 |
+
"_cosmos3_nano_tokenizer": {
|
| 47 |
+
"model_max_length": 4096,
|
| 48 |
+
"model_path": "./checkpoints/Cosmos3-Nano/text_tokenizer",
|
| 49 |
+
"padding_side": "right",
|
| 50 |
+
"trust_remote_code": true,
|
| 51 |
+
"type": "PretrainedTokenizer"
|
| 52 |
+
},
|
| 53 |
+
"_cosmos3_nano_transformer": "./checkpoints/Cosmos3-Nano/transformer",
|
| 54 |
+
"_cosmos3_nano_vision_encoder": "./checkpoints/Cosmos3-Nano/vision_encoder",
|
| 55 |
+
"_cosmos3_nano_vlm_config": {
|
| 56 |
+
"image_token_id": 151655,
|
| 57 |
+
"model_type": "qwen3_vl",
|
| 58 |
+
"text_config": {
|
| 59 |
+
"attention_bias": false,
|
| 60 |
+
"attention_dropout": 0.0,
|
| 61 |
+
"bos_token_id": 151643,
|
| 62 |
+
"eos_token_id": 151645,
|
| 63 |
+
"head_dim": 128,
|
| 64 |
+
"hidden_act": "silu",
|
| 65 |
+
"hidden_size": 4096,
|
| 66 |
+
"intermediate_size": 12288,
|
| 67 |
+
"layer_types": [
|
| 68 |
+
"full_attention",
|
| 69 |
+
"full_attention",
|
| 70 |
+
"full_attention",
|
| 71 |
+
"full_attention",
|
| 72 |
+
"full_attention",
|
| 73 |
+
"full_attention",
|
| 74 |
+
"full_attention",
|
| 75 |
+
"full_attention",
|
| 76 |
+
"full_attention",
|
| 77 |
+
"full_attention",
|
| 78 |
+
"full_attention",
|
| 79 |
+
"full_attention",
|
| 80 |
+
"full_attention",
|
| 81 |
+
"full_attention",
|
| 82 |
+
"full_attention",
|
| 83 |
+
"full_attention",
|
| 84 |
+
"full_attention",
|
| 85 |
+
"full_attention",
|
| 86 |
+
"full_attention",
|
| 87 |
+
"full_attention",
|
| 88 |
+
"full_attention",
|
| 89 |
+
"full_attention",
|
| 90 |
+
"full_attention",
|
| 91 |
+
"full_attention",
|
| 92 |
+
"full_attention",
|
| 93 |
+
"full_attention",
|
| 94 |
+
"full_attention",
|
| 95 |
+
"full_attention",
|
| 96 |
+
"full_attention",
|
| 97 |
+
"full_attention",
|
| 98 |
+
"full_attention",
|
| 99 |
+
"full_attention",
|
| 100 |
+
"full_attention",
|
| 101 |
+
"full_attention",
|
| 102 |
+
"full_attention",
|
| 103 |
+
"full_attention"
|
| 104 |
+
],
|
| 105 |
+
"max_position_embeddings": 262144,
|
| 106 |
+
"model_type": "qwen3_vl_text",
|
| 107 |
+
"num_attention_heads": 32,
|
| 108 |
+
"num_hidden_layers": 36,
|
| 109 |
+
"num_key_value_heads": 8,
|
| 110 |
+
"pad_token_id": 0,
|
| 111 |
+
"rms_norm_eps": 1e-06,
|
| 112 |
+
"rope_scaling": {
|
| 113 |
+
"mrope_interleaved": true,
|
| 114 |
+
"mrope_section": [
|
| 115 |
+
24,
|
| 116 |
+
20,
|
| 117 |
+
20
|
| 118 |
+
],
|
| 119 |
+
"rope_type": "default"
|
| 120 |
+
},
|
| 121 |
+
"rope_theta": 5000000,
|
| 122 |
+
"tie_word_embeddings": false,
|
| 123 |
+
"vocab_size": 151936
|
| 124 |
+
},
|
| 125 |
+
"tie_word_embeddings": false,
|
| 126 |
+
"video_token_id": 151656,
|
| 127 |
+
"vision_config": {
|
| 128 |
+
"deepstack_visual_indexes": [
|
| 129 |
+
8,
|
| 130 |
+
16,
|
| 131 |
+
24
|
| 132 |
+
],
|
| 133 |
+
"depth": 27,
|
| 134 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 135 |
+
"hidden_size": 1152,
|
| 136 |
+
"in_channels": 3,
|
| 137 |
+
"initializer_range": 0.02,
|
| 138 |
+
"intermediate_size": 4304,
|
| 139 |
+
"model_type": "qwen3_vl",
|
| 140 |
+
"num_heads": 16,
|
| 141 |
+
"num_position_embeddings": 2304,
|
| 142 |
+
"out_hidden_size": 4096,
|
| 143 |
+
"patch_size": 16,
|
| 144 |
+
"spatial_merge_size": 2,
|
| 145 |
+
"temporal_patch_size": 2
|
| 146 |
+
},
|
| 147 |
+
"vision_end_token_id": 151653,
|
| 148 |
+
"vision_start_token_id": 151652,
|
| 149 |
+
"vocab_size": 151936
|
| 150 |
+
},
|
| 151 |
+
"_data_root_path": [
|
| 152 |
+
"./datasets/libero_spatial_no_noops_lerobotv2.1",
|
| 153 |
+
"./datasets/libero_spatial_no_noops_lerobotv2.1",
|
| 154 |
+
"./datasets/libero_object_no_noops_lerobotv2.1",
|
| 155 |
+
"./datasets/libero_object_no_noops_lerobotv2.1",
|
| 156 |
+
"./datasets/libero_goal_no_noops_lerobotv2.1",
|
| 157 |
+
"./datasets/libero_goal_no_noops_lerobotv2.1",
|
| 158 |
+
"./datasets/libero_10_no_noops_lerobotv2.1"
|
| 159 |
+
],
|
| 160 |
+
"_frame_window_size": 17,
|
| 161 |
+
"_grad_accumulation_steps": 8,
|
| 162 |
+
"_image_height": 256,
|
| 163 |
+
"_image_width": 256,
|
| 164 |
+
"_libero_action_stats": {
|
| 165 |
+
"max": [
|
| 166 |
+
0.9375,
|
| 167 |
+
0.9375,
|
| 168 |
+
0.9375,
|
| 169 |
+
1.0,
|
| 170 |
+
0.368853,
|
| 171 |
+
0.341214,
|
| 172 |
+
0.356395,
|
| 173 |
+
1.0,
|
| 174 |
+
0.348251,
|
| 175 |
+
1.0
|
| 176 |
+
],
|
| 177 |
+
"mean": [
|
| 178 |
+
0.050704,
|
| 179 |
+
0.097407,
|
| 180 |
+
-0.094833,
|
| 181 |
+
0.994873,
|
| 182 |
+
-0.004579,
|
| 183 |
+
-0.004288,
|
| 184 |
+
0.004389,
|
| 185 |
+
0.996104,
|
| 186 |
+
0.001109,
|
| 187 |
+
0.476725
|
| 188 |
+
],
|
| 189 |
+
"min": [
|
| 190 |
+
-0.9375,
|
| 191 |
+
-0.9375,
|
| 192 |
+
-0.9375,
|
| 193 |
+
0.902028,
|
| 194 |
+
-0.356085,
|
| 195 |
+
-0.367416,
|
| 196 |
+
-0.370434,
|
| 197 |
+
0.921907,
|
| 198 |
+
-0.255,
|
| 199 |
+
0.0
|
| 200 |
+
],
|
| 201 |
+
"q01": [
|
| 202 |
+
-0.723214,
|
| 203 |
+
-0.808929,
|
| 204 |
+
-0.9375,
|
| 205 |
+
0.934955,
|
| 206 |
+
-0.223431,
|
| 207 |
+
-0.189878,
|
| 208 |
+
-0.334735,
|
| 209 |
+
0.938516,
|
| 210 |
+
-0.107736,
|
| 211 |
+
0.0
|
| 212 |
+
],
|
| 213 |
+
"q99": [
|
| 214 |
+
0.9375,
|
| 215 |
+
0.870536,
|
| 216 |
+
0.9375,
|
| 217 |
+
1.0,
|
| 218 |
+
0.331,
|
| 219 |
+
0.163153,
|
| 220 |
+
0.226216,
|
| 221 |
+
1.0,
|
| 222 |
+
0.127158,
|
| 223 |
+
1.0
|
| 224 |
+
],
|
| 225 |
+
"std": [
|
| 226 |
+
0.333621,
|
| 227 |
+
0.387175,
|
| 228 |
+
0.45714,
|
| 229 |
+
0.010807,
|
| 230 |
+
0.077802,
|
| 231 |
+
0.063386,
|
| 232 |
+
0.078571,
|
| 233 |
+
0.009994,
|
| 234 |
+
0.038504,
|
| 235 |
+
0.49946
|
| 236 |
+
]
|
| 237 |
+
},
|
| 238 |
+
"_max_action_dim": 64,
|
| 239 |
+
"_max_state_dim": 64,
|
| 240 |
+
"_max_steps": 5000,
|
| 241 |
+
"_per_device_batch_size": 16,
|
| 242 |
+
"_prepend_state_to_action": false,
|
| 243 |
+
"_save_iter_interval": 500,
|
| 244 |
+
"_statistic_name": "all_libero_no_noops",
|
| 245 |
+
"_transforms": [
|
| 246 |
+
{
|
| 247 |
+
"embodiment_id": 5,
|
| 248 |
+
"name_mappings": {
|
| 249 |
+
"actions": [
|
| 250 |
+
"actions"
|
| 251 |
+
],
|
| 252 |
+
"observation.state": [
|
| 253 |
+
"states"
|
| 254 |
+
]
|
| 255 |
+
},
|
| 256 |
+
"parquet_keys": [
|
| 257 |
+
"observation.state",
|
| 258 |
+
"timestamp",
|
| 259 |
+
"actions",
|
| 260 |
+
"info",
|
| 261 |
+
"stats",
|
| 262 |
+
"action_masks"
|
| 263 |
+
],
|
| 264 |
+
"type": "ProcessParquetInputs",
|
| 265 |
+
"video_keys": [
|
| 266 |
+
"observation.images.image",
|
| 267 |
+
"observation.images.wrist_image"
|
| 268 |
+
]
|
| 269 |
+
},
|
| 270 |
+
{
|
| 271 |
+
"type": "LiberoFramewiseActionToRot6D"
|
| 272 |
+
},
|
| 273 |
+
{
|
| 274 |
+
"action_metadata": {
|
| 275 |
+
"append_viewpoint": false,
|
| 276 |
+
"conditioning_fps": 20.0,
|
| 277 |
+
"frame_window_size": 17,
|
| 278 |
+
"video_height": 192,
|
| 279 |
+
"video_width": 320
|
| 280 |
+
},
|
| 281 |
+
"cfg_dropout_rate": 0.1,
|
| 282 |
+
"format_prompt_as_json": true,
|
| 283 |
+
"max_len": 512,
|
| 284 |
+
"tokenizer": {
|
| 285 |
+
"model_max_length": 4096,
|
| 286 |
+
"model_path": "./checkpoints/Cosmos3-Nano/text_tokenizer",
|
| 287 |
+
"padding_side": "right",
|
| 288 |
+
"trust_remote_code": true,
|
| 289 |
+
"type": "PretrainedTokenizer"
|
| 290 |
+
},
|
| 291 |
+
"type": "ProcessCosmos3Prompt"
|
| 292 |
+
},
|
| 293 |
+
{
|
| 294 |
+
"height": 256,
|
| 295 |
+
"type": "ResizeImages",
|
| 296 |
+
"width": 256
|
| 297 |
+
},
|
| 298 |
+
{
|
| 299 |
+
"type": "SimpleNormalizeImages"
|
| 300 |
+
},
|
| 301 |
+
{
|
| 302 |
+
"action_dim": 64,
|
| 303 |
+
"action_key": "action",
|
| 304 |
+
"action_norm_type": "quantile",
|
| 305 |
+
"state_dim": 64,
|
| 306 |
+
"state_key": "proprio",
|
| 307 |
+
"state_norm_type": "none",
|
| 308 |
+
"type": "NormalizeStatesAndActions"
|
| 309 |
+
},
|
| 310 |
+
{
|
| 311 |
+
"conditioning_fps": 20.0,
|
| 312 |
+
"frame_window_size": 17,
|
| 313 |
+
"mode": "wam",
|
| 314 |
+
"prepend_state_to_action": false,
|
| 315 |
+
"raw_action_dim": 10,
|
| 316 |
+
"type": "BuildCosmos3Sequence"
|
| 317 |
+
},
|
| 318 |
+
{
|
| 319 |
+
"frame_window_size": 17,
|
| 320 |
+
"num_views": 2,
|
| 321 |
+
"tile_direction": "horizontal",
|
| 322 |
+
"type": "PrepareVideo"
|
| 323 |
+
},
|
| 324 |
+
{
|
| 325 |
+
"height": 192,
|
| 326 |
+
"type": "ResizeAndReflectPad",
|
| 327 |
+
"width": 320
|
| 328 |
+
}
|
| 329 |
+
],
|
| 330 |
+
"_video_height": 192,
|
| 331 |
+
"_video_width": 320,
|
| 332 |
+
"_vision_vae": {
|
| 333 |
+
"encode_exact_durations": [
|
| 334 |
+
17
|
| 335 |
+
],
|
| 336 |
+
"pretrained_name_or_path": "./checkpoints/Wan2.2-TI2V-5B/Wan2.2_VAE.pth",
|
| 337 |
+
"type": "Cosmos3Wan22VAE"
|
| 338 |
+
},
|
| 339 |
+
"_wan22_vae_path": "./checkpoints/Wan2.2-TI2V-5B/Wan2.2_VAE.pth",
|
| 340 |
+
"eval": {
|
| 341 |
+
"dataset": {
|
| 342 |
+
"extra_tensor_keys": [
|
| 343 |
+
"conditioning_fps",
|
| 344 |
+
"prepend_state_to_action"
|
| 345 |
+
],
|
| 346 |
+
"img_buffer_len": 1,
|
| 347 |
+
"transforms": [
|
| 348 |
+
{
|
| 349 |
+
"embodiment_id": 5,
|
| 350 |
+
"img_keys": [
|
| 351 |
+
"agentview_image",
|
| 352 |
+
"robot0_eye_in_hand_image"
|
| 353 |
+
],
|
| 354 |
+
"resize_size": 256,
|
| 355 |
+
"type": "ProcessLiberoEvalInputs"
|
| 356 |
+
},
|
| 357 |
+
{
|
| 358 |
+
"conditioning_fps": 20.0,
|
| 359 |
+
"prepend_state_to_action": false,
|
| 360 |
+
"type": "SetCosmos3ActionMetadata"
|
| 361 |
+
},
|
| 362 |
+
{
|
| 363 |
+
"image_resize_strategy": "resize-crop",
|
| 364 |
+
"input_sizes": [
|
| 365 |
+
[
|
| 366 |
+
3,
|
| 367 |
+
256,
|
| 368 |
+
256
|
| 369 |
+
],
|
| 370 |
+
[
|
| 371 |
+
3,
|
| 372 |
+
256,
|
| 373 |
+
256
|
| 374 |
+
]
|
| 375 |
+
],
|
| 376 |
+
"means": [
|
| 377 |
+
[
|
| 378 |
+
127.5,
|
| 379 |
+
127.5,
|
| 380 |
+
127.5
|
| 381 |
+
],
|
| 382 |
+
[
|
| 383 |
+
127.5,
|
| 384 |
+
127.5,
|
| 385 |
+
127.5
|
| 386 |
+
]
|
| 387 |
+
],
|
| 388 |
+
"stds": [
|
| 389 |
+
[
|
| 390 |
+
127.5,
|
| 391 |
+
127.5,
|
| 392 |
+
127.5
|
| 393 |
+
],
|
| 394 |
+
[
|
| 395 |
+
127.5,
|
| 396 |
+
127.5,
|
| 397 |
+
127.5
|
| 398 |
+
]
|
| 399 |
+
],
|
| 400 |
+
"type": "TransformImage"
|
| 401 |
+
},
|
| 402 |
+
{
|
| 403 |
+
"action_metadata": {
|
| 404 |
+
"append_viewpoint": false,
|
| 405 |
+
"conditioning_fps": 20.0,
|
| 406 |
+
"frame_window_size": 17,
|
| 407 |
+
"video_height": 192,
|
| 408 |
+
"video_width": 320
|
| 409 |
+
},
|
| 410 |
+
"cfg_dropout_rate": 0.0,
|
| 411 |
+
"format_prompt_as_json": true,
|
| 412 |
+
"max_len": 512,
|
| 413 |
+
"output_attention_mask_key": "lang_masks",
|
| 414 |
+
"output_key": "lang_tokens",
|
| 415 |
+
"tokenizer": {
|
| 416 |
+
"model_max_length": 4096,
|
| 417 |
+
"model_path": "./checkpoints/Cosmos3-Nano/text_tokenizer",
|
| 418 |
+
"padding_side": "right",
|
| 419 |
+
"trust_remote_code": true,
|
| 420 |
+
"type": "PretrainedTokenizer"
|
| 421 |
+
},
|
| 422 |
+
"type": "ProcessCosmos3Prompt"
|
| 423 |
+
},
|
| 424 |
+
{
|
| 425 |
+
"frame_window_size": 1,
|
| 426 |
+
"num_views": 2,
|
| 427 |
+
"tile_direction": "horizontal",
|
| 428 |
+
"type": "PrepareVideo"
|
| 429 |
+
},
|
| 430 |
+
{
|
| 431 |
+
"height": 192,
|
| 432 |
+
"type": "ResizeAndReflectPad",
|
| 433 |
+
"width": 320
|
| 434 |
+
}
|
| 435 |
+
],
|
| 436 |
+
"type": "LiberoParquetEvalDataset"
|
| 437 |
+
},
|
| 438 |
+
"denormalize_action": {
|
| 439 |
+
"action_dim": 10,
|
| 440 |
+
"norm_type": "quantile_rot",
|
| 441 |
+
"type": "DenormalizeLiberoFramewiseRot6DAction"
|
| 442 |
+
},
|
| 443 |
+
"enable_mixed_precision_training": true,
|
| 444 |
+
"eval_chunk_size": 16,
|
| 445 |
+
"mixed_precision_dtype": "bf16",
|
| 446 |
+
"model_family": "cosmos3",
|
| 447 |
+
"norm_stats_key": "all_libero_no_noops",
|
| 448 |
+
"num_steps_wait": 10,
|
| 449 |
+
"num_trials_per_task": 50,
|
| 450 |
+
"seed": 7,
|
| 451 |
+
"task_suite_name": "libero_10",
|
| 452 |
+
"type": "LiberoEvalRunner"
|
| 453 |
+
},
|
| 454 |
+
"inference_model": {
|
| 455 |
+
"action_horizon": 16,
|
| 456 |
+
"action_in_proj": {
|
| 457 |
+
"input_size": 64,
|
| 458 |
+
"num_domains": 32,
|
| 459 |
+
"output_size": 4096,
|
| 460 |
+
"type": "DomainAwareLinear"
|
| 461 |
+
},
|
| 462 |
+
"action_out_proj": {
|
| 463 |
+
"input_size": 4096,
|
| 464 |
+
"num_domains": 32,
|
| 465 |
+
"output_size": 64,
|
| 466 |
+
"type": "DomainAwareLinear"
|
| 467 |
+
},
|
| 468 |
+
"base_fps": 24.0,
|
| 469 |
+
"enable_fps_modulation": true,
|
| 470 |
+
"enable_vision_loss": true,
|
| 471 |
+
"freeze_non_moe_vlm_backbone": true,
|
| 472 |
+
"freeze_vlm_backbone": false,
|
| 473 |
+
"latent_patch_size": 2,
|
| 474 |
+
"max_action_dim": 64,
|
| 475 |
+
"name_mapping": {
|
| 476 |
+
".self_attn.k_norm.": ".self_attn.norm_k.",
|
| 477 |
+
".self_attn.k_norm_moe_gen.": ".self_attn.norm_added_k.",
|
| 478 |
+
".self_attn.k_proj.": ".self_attn.to_k.",
|
| 479 |
+
".self_attn.k_proj_moe_gen.": ".self_attn.add_k_proj.",
|
| 480 |
+
".self_attn.o_proj.": ".self_attn.to_out.",
|
| 481 |
+
".self_attn.o_proj_moe_gen.": ".self_attn.to_add_out.",
|
| 482 |
+
".self_attn.q_norm.": ".self_attn.norm_q.",
|
| 483 |
+
".self_attn.q_norm_moe_gen.": ".self_attn.norm_added_q.",
|
| 484 |
+
".self_attn.q_proj.": ".self_attn.to_q.",
|
| 485 |
+
".self_attn.q_proj_moe_gen.": ".self_attn.add_q_proj.",
|
| 486 |
+
".self_attn.v_proj.": ".self_attn.to_v.",
|
| 487 |
+
".self_attn.v_proj_moe_gen.": ".self_attn.add_v_proj.",
|
| 488 |
+
"action_in_proj.": "action_proj_in.",
|
| 489 |
+
"action_modality_embed": "action_modality_embed",
|
| 490 |
+
"action_out_proj.": "action_proj_out.",
|
| 491 |
+
"time_embedder.mlp.0.": "time_embedder.linear_1.",
|
| 492 |
+
"time_embedder.mlp.2.": "time_embedder.linear_2.",
|
| 493 |
+
"vision_in_proj.projector.": "proj_in.",
|
| 494 |
+
"vision_out_proj.projector.": "proj_out.",
|
| 495 |
+
"vlm_backbone.lm_head.weight": "lm_head.weight",
|
| 496 |
+
"vlm_backbone.model.language_model.embed_tokens.weight": "embed_tokens.weight",
|
| 497 |
+
"vlm_backbone.model.language_model.layers.": "layers.",
|
| 498 |
+
"vlm_backbone.model.language_model.norm.weight": "norm.weight",
|
| 499 |
+
"vlm_backbone.model.language_model.norm_moe_gen.weight": "norm_moe_gen.weight",
|
| 500 |
+
"vlm_backbone.model.visual.blocks.": "blocks.",
|
| 501 |
+
"vlm_backbone.model.visual.deepstack_merger_list.": "deepstack_merger_list.",
|
| 502 |
+
"vlm_backbone.model.visual.merger.": "merger.",
|
| 503 |
+
"vlm_backbone.model.visual.patch_embed.": "patch_embed.",
|
| 504 |
+
"vlm_backbone.model.visual.pos_embed.": "pos_embed."
|
| 505 |
+
},
|
| 506 |
+
"num_embodiment_domains": 32,
|
| 507 |
+
"ori_action_dim": 10,
|
| 508 |
+
"packed_attention_backend": "flash2",
|
| 509 |
+
"position_embedding_type": "unified_3d_mrope",
|
| 510 |
+
"pretrained_name_or_path": "./checkpoints/Cosmos3-Nano/transformer",
|
| 511 |
+
"rectified_flow_inference_config": {
|
| 512 |
+
"num_steps": 30,
|
| 513 |
+
"num_train_timesteps": 1000,
|
| 514 |
+
"scheduler_type": "unipc",
|
| 515 |
+
"shift": 10.0,
|
| 516 |
+
"use_dynamic_shifting": false,
|
| 517 |
+
"use_karras_sigmas": false
|
| 518 |
+
},
|
| 519 |
+
"rectified_flow_training_config": {
|
| 520 |
+
"action_loss_weight": 10.0,
|
| 521 |
+
"high_sigma_ratio": 0.05,
|
| 522 |
+
"high_sigma_timesteps_max": 1000,
|
| 523 |
+
"high_sigma_timesteps_min": 995,
|
| 524 |
+
"independent_action_schedule": false,
|
| 525 |
+
"normalize_loss_by_active": false,
|
| 526 |
+
"shift": {
|
| 527 |
+
"256": 3,
|
| 528 |
+
"480": 5,
|
| 529 |
+
"720": 10
|
| 530 |
+
},
|
| 531 |
+
"shift_action": null,
|
| 532 |
+
"train_time_action_distribution": "logitnormal",
|
| 533 |
+
"train_time_image_distribution": "logitnormal",
|
| 534 |
+
"train_time_video_distribution": "waver",
|
| 535 |
+
"train_time_weight": "uniform",
|
| 536 |
+
"use_discrete_rf": false,
|
| 537 |
+
"use_dynamic_shift": false,
|
| 538 |
+
"use_high_sigma_strategy": false,
|
| 539 |
+
"use_high_sigma_strategy_action": false,
|
| 540 |
+
"vision_loss_weight": 10.0
|
| 541 |
+
},
|
| 542 |
+
"special_tokens": {
|
| 543 |
+
"end_of_generation": 151653,
|
| 544 |
+
"eos_token_id": 151645,
|
| 545 |
+
"start_of_generation": 151652
|
| 546 |
+
},
|
| 547 |
+
"timestep_scale": 0.001,
|
| 548 |
+
"type": "Cosmos3FlowMatching",
|
| 549 |
+
"unified_3d_mrope_reset_spatial_ids": true,
|
| 550 |
+
"unified_3d_mrope_temporal_modality_margin": 15000,
|
| 551 |
+
"vision_in_proj": {
|
| 552 |
+
"in_dim": 192,
|
| 553 |
+
"out_dim": 4096,
|
| 554 |
+
"type": "LinearProjector"
|
| 555 |
+
},
|
| 556 |
+
"vision_latent_dim": 48,
|
| 557 |
+
"vision_out_proj": {
|
| 558 |
+
"in_dim": 4096,
|
| 559 |
+
"out_dim": 192,
|
| 560 |
+
"type": "LinearProjector"
|
| 561 |
+
},
|
| 562 |
+
"vision_vae": {
|
| 563 |
+
"encode_exact_durations": [
|
| 564 |
+
17
|
| 565 |
+
],
|
| 566 |
+
"pretrained_name_or_path": null,
|
| 567 |
+
"type": "Cosmos3Wan22VAE"
|
| 568 |
+
},
|
| 569 |
+
"vlm_backbone": {
|
| 570 |
+
"include_visual": false,
|
| 571 |
+
"skip_init_weights": true,
|
| 572 |
+
"type": "Cosmos3MoTBackbone",
|
| 573 |
+
"vision_encoder_path": "./checkpoints/Cosmos3-Nano/vision_encoder",
|
| 574 |
+
"vlm_config": {
|
| 575 |
+
"image_token_id": 151655,
|
| 576 |
+
"model_type": "qwen3_vl",
|
| 577 |
+
"text_config": {
|
| 578 |
+
"attention_bias": false,
|
| 579 |
+
"attention_dropout": 0.0,
|
| 580 |
+
"bos_token_id": 151643,
|
| 581 |
+
"eos_token_id": 151645,
|
| 582 |
+
"head_dim": 128,
|
| 583 |
+
"hidden_act": "silu",
|
| 584 |
+
"hidden_size": 4096,
|
| 585 |
+
"intermediate_size": 12288,
|
| 586 |
+
"layer_types": [
|
| 587 |
+
"full_attention",
|
| 588 |
+
"full_attention",
|
| 589 |
+
"full_attention",
|
| 590 |
+
"full_attention",
|
| 591 |
+
"full_attention",
|
| 592 |
+
"full_attention",
|
| 593 |
+
"full_attention",
|
| 594 |
+
"full_attention",
|
| 595 |
+
"full_attention",
|
| 596 |
+
"full_attention",
|
| 597 |
+
"full_attention",
|
| 598 |
+
"full_attention",
|
| 599 |
+
"full_attention",
|
| 600 |
+
"full_attention",
|
| 601 |
+
"full_attention",
|
| 602 |
+
"full_attention",
|
| 603 |
+
"full_attention",
|
| 604 |
+
"full_attention",
|
| 605 |
+
"full_attention",
|
| 606 |
+
"full_attention",
|
| 607 |
+
"full_attention",
|
| 608 |
+
"full_attention",
|
| 609 |
+
"full_attention",
|
| 610 |
+
"full_attention",
|
| 611 |
+
"full_attention",
|
| 612 |
+
"full_attention",
|
| 613 |
+
"full_attention",
|
| 614 |
+
"full_attention",
|
| 615 |
+
"full_attention",
|
| 616 |
+
"full_attention",
|
| 617 |
+
"full_attention",
|
| 618 |
+
"full_attention",
|
| 619 |
+
"full_attention",
|
| 620 |
+
"full_attention",
|
| 621 |
+
"full_attention",
|
| 622 |
+
"full_attention"
|
| 623 |
+
],
|
| 624 |
+
"max_position_embeddings": 262144,
|
| 625 |
+
"model_type": "qwen3_vl_text",
|
| 626 |
+
"num_attention_heads": 32,
|
| 627 |
+
"num_hidden_layers": 36,
|
| 628 |
+
"num_key_value_heads": 8,
|
| 629 |
+
"pad_token_id": 0,
|
| 630 |
+
"rms_norm_eps": 1e-06,
|
| 631 |
+
"rope_scaling": {
|
| 632 |
+
"mrope_interleaved": true,
|
| 633 |
+
"mrope_section": [
|
| 634 |
+
24,
|
| 635 |
+
20,
|
| 636 |
+
20
|
| 637 |
+
],
|
| 638 |
+
"rope_type": "default"
|
| 639 |
+
},
|
| 640 |
+
"rope_theta": 5000000,
|
| 641 |
+
"tie_word_embeddings": false,
|
| 642 |
+
"vocab_size": 151936
|
| 643 |
+
},
|
| 644 |
+
"tie_word_embeddings": false,
|
| 645 |
+
"video_token_id": 151656,
|
| 646 |
+
"vision_config": {
|
| 647 |
+
"deepstack_visual_indexes": [
|
| 648 |
+
8,
|
| 649 |
+
16,
|
| 650 |
+
24
|
| 651 |
+
],
|
| 652 |
+
"depth": 27,
|
| 653 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 654 |
+
"hidden_size": 1152,
|
| 655 |
+
"in_channels": 3,
|
| 656 |
+
"initializer_range": 0.02,
|
| 657 |
+
"intermediate_size": 4304,
|
| 658 |
+
"model_type": "qwen3_vl",
|
| 659 |
+
"num_heads": 16,
|
| 660 |
+
"num_position_embeddings": 2304,
|
| 661 |
+
"out_hidden_size": 4096,
|
| 662 |
+
"patch_size": 16,
|
| 663 |
+
"spatial_merge_size": 2,
|
| 664 |
+
"temporal_patch_size": 2
|
| 665 |
+
},
|
| 666 |
+
"vision_end_token_id": 151653,
|
| 667 |
+
"vision_start_token_id": 151652,
|
| 668 |
+
"vocab_size": 151936
|
| 669 |
+
}
|
| 670 |
+
}
|
| 671 |
+
},
|
| 672 |
+
"model": {
|
| 673 |
+
"action_horizon": 16,
|
| 674 |
+
"action_in_proj": {
|
| 675 |
+
"input_size": 64,
|
| 676 |
+
"num_domains": 32,
|
| 677 |
+
"output_size": 4096,
|
| 678 |
+
"type": "DomainAwareLinear"
|
| 679 |
+
},
|
| 680 |
+
"action_out_proj": {
|
| 681 |
+
"input_size": 4096,
|
| 682 |
+
"num_domains": 32,
|
| 683 |
+
"output_size": 64,
|
| 684 |
+
"type": "DomainAwareLinear"
|
| 685 |
+
},
|
| 686 |
+
"base_fps": 24.0,
|
| 687 |
+
"enable_fps_modulation": true,
|
| 688 |
+
"enable_vision_loss": true,
|
| 689 |
+
"freeze_non_moe_vlm_backbone": true,
|
| 690 |
+
"freeze_vlm_backbone": false,
|
| 691 |
+
"latent_patch_size": 2,
|
| 692 |
+
"max_action_dim": 64,
|
| 693 |
+
"name_mapping": {
|
| 694 |
+
".self_attn.k_norm.": ".self_attn.norm_k.",
|
| 695 |
+
".self_attn.k_norm_moe_gen.": ".self_attn.norm_added_k.",
|
| 696 |
+
".self_attn.k_proj.": ".self_attn.to_k.",
|
| 697 |
+
".self_attn.k_proj_moe_gen.": ".self_attn.add_k_proj.",
|
| 698 |
+
".self_attn.o_proj.": ".self_attn.to_out.",
|
| 699 |
+
".self_attn.o_proj_moe_gen.": ".self_attn.to_add_out.",
|
| 700 |
+
".self_attn.q_norm.": ".self_attn.norm_q.",
|
| 701 |
+
".self_attn.q_norm_moe_gen.": ".self_attn.norm_added_q.",
|
| 702 |
+
".self_attn.q_proj.": ".self_attn.to_q.",
|
| 703 |
+
".self_attn.q_proj_moe_gen.": ".self_attn.add_q_proj.",
|
| 704 |
+
".self_attn.v_proj.": ".self_attn.to_v.",
|
| 705 |
+
".self_attn.v_proj_moe_gen.": ".self_attn.add_v_proj.",
|
| 706 |
+
"action_in_proj.": "action_proj_in.",
|
| 707 |
+
"action_modality_embed": "action_modality_embed",
|
| 708 |
+
"action_out_proj.": "action_proj_out.",
|
| 709 |
+
"time_embedder.mlp.0.": "time_embedder.linear_1.",
|
| 710 |
+
"time_embedder.mlp.2.": "time_embedder.linear_2.",
|
| 711 |
+
"vision_in_proj.projector.": "proj_in.",
|
| 712 |
+
"vision_out_proj.projector.": "proj_out.",
|
| 713 |
+
"vlm_backbone.lm_head.weight": "lm_head.weight",
|
| 714 |
+
"vlm_backbone.model.language_model.embed_tokens.weight": "embed_tokens.weight",
|
| 715 |
+
"vlm_backbone.model.language_model.layers.": "layers.",
|
| 716 |
+
"vlm_backbone.model.language_model.norm.weight": "norm.weight",
|
| 717 |
+
"vlm_backbone.model.language_model.norm_moe_gen.weight": "norm_moe_gen.weight",
|
| 718 |
+
"vlm_backbone.model.visual.blocks.": "blocks.",
|
| 719 |
+
"vlm_backbone.model.visual.deepstack_merger_list.": "deepstack_merger_list.",
|
| 720 |
+
"vlm_backbone.model.visual.merger.": "merger.",
|
| 721 |
+
"vlm_backbone.model.visual.patch_embed.": "patch_embed.",
|
| 722 |
+
"vlm_backbone.model.visual.pos_embed.": "pos_embed."
|
| 723 |
+
},
|
| 724 |
+
"num_embodiment_domains": 32,
|
| 725 |
+
"ori_action_dim": 10,
|
| 726 |
+
"packed_attention_backend": "flash2",
|
| 727 |
+
"position_embedding_type": "unified_3d_mrope",
|
| 728 |
+
"pretrained_name_or_path": "./checkpoints/Cosmos3-Nano/transformer",
|
| 729 |
+
"rectified_flow_inference_config": {
|
| 730 |
+
"num_steps": 30,
|
| 731 |
+
"num_train_timesteps": 1000,
|
| 732 |
+
"scheduler_type": "unipc",
|
| 733 |
+
"shift": 10.0,
|
| 734 |
+
"use_dynamic_shifting": false,
|
| 735 |
+
"use_karras_sigmas": false
|
| 736 |
+
},
|
| 737 |
+
"rectified_flow_training_config": {
|
| 738 |
+
"action_loss_weight": 10.0,
|
| 739 |
+
"high_sigma_ratio": 0.05,
|
| 740 |
+
"high_sigma_timesteps_max": 1000,
|
| 741 |
+
"high_sigma_timesteps_min": 995,
|
| 742 |
+
"independent_action_schedule": false,
|
| 743 |
+
"normalize_loss_by_active": false,
|
| 744 |
+
"shift": {
|
| 745 |
+
"256": 3,
|
| 746 |
+
"480": 5,
|
| 747 |
+
"720": 10
|
| 748 |
+
},
|
| 749 |
+
"shift_action": null,
|
| 750 |
+
"train_time_action_distribution": "logitnormal",
|
| 751 |
+
"train_time_image_distribution": "logitnormal",
|
| 752 |
+
"train_time_video_distribution": "waver",
|
| 753 |
+
"train_time_weight": "uniform",
|
| 754 |
+
"use_discrete_rf": false,
|
| 755 |
+
"use_dynamic_shift": false,
|
| 756 |
+
"use_high_sigma_strategy": false,
|
| 757 |
+
"use_high_sigma_strategy_action": false,
|
| 758 |
+
"vision_loss_weight": 10.0
|
| 759 |
+
},
|
| 760 |
+
"special_tokens": {
|
| 761 |
+
"end_of_generation": 151653,
|
| 762 |
+
"eos_token_id": 151645,
|
| 763 |
+
"start_of_generation": 151652
|
| 764 |
+
},
|
| 765 |
+
"timestep_scale": 0.001,
|
| 766 |
+
"type": "Cosmos3FlowMatching",
|
| 767 |
+
"unified_3d_mrope_reset_spatial_ids": true,
|
| 768 |
+
"unified_3d_mrope_temporal_modality_margin": 15000,
|
| 769 |
+
"vision_in_proj": {
|
| 770 |
+
"in_dim": 192,
|
| 771 |
+
"out_dim": 4096,
|
| 772 |
+
"type": "LinearProjector"
|
| 773 |
+
},
|
| 774 |
+
"vision_latent_dim": 48,
|
| 775 |
+
"vision_out_proj": {
|
| 776 |
+
"in_dim": 4096,
|
| 777 |
+
"out_dim": 192,
|
| 778 |
+
"type": "LinearProjector"
|
| 779 |
+
},
|
| 780 |
+
"vision_vae": {
|
| 781 |
+
"encode_exact_durations": [
|
| 782 |
+
17
|
| 783 |
+
],
|
| 784 |
+
"pretrained_name_or_path": "./checkpoints/Wan2.2-TI2V-5B/Wan2.2_VAE.pth",
|
| 785 |
+
"type": "Cosmos3Wan22VAE"
|
| 786 |
+
},
|
| 787 |
+
"vlm_backbone": {
|
| 788 |
+
"include_visual": false,
|
| 789 |
+
"skip_init_weights": true,
|
| 790 |
+
"type": "Cosmos3MoTBackbone",
|
| 791 |
+
"vision_encoder_path": "./checkpoints/Cosmos3-Nano/vision_encoder",
|
| 792 |
+
"vlm_config": {
|
| 793 |
+
"image_token_id": 151655,
|
| 794 |
+
"model_type": "qwen3_vl",
|
| 795 |
+
"text_config": {
|
| 796 |
+
"attention_bias": false,
|
| 797 |
+
"attention_dropout": 0.0,
|
| 798 |
+
"bos_token_id": 151643,
|
| 799 |
+
"eos_token_id": 151645,
|
| 800 |
+
"head_dim": 128,
|
| 801 |
+
"hidden_act": "silu",
|
| 802 |
+
"hidden_size": 4096,
|
| 803 |
+
"intermediate_size": 12288,
|
| 804 |
+
"layer_types": [
|
| 805 |
+
"full_attention",
|
| 806 |
+
"full_attention",
|
| 807 |
+
"full_attention",
|
| 808 |
+
"full_attention",
|
| 809 |
+
"full_attention",
|
| 810 |
+
"full_attention",
|
| 811 |
+
"full_attention",
|
| 812 |
+
"full_attention",
|
| 813 |
+
"full_attention",
|
| 814 |
+
"full_attention",
|
| 815 |
+
"full_attention",
|
| 816 |
+
"full_attention",
|
| 817 |
+
"full_attention",
|
| 818 |
+
"full_attention",
|
| 819 |
+
"full_attention",
|
| 820 |
+
"full_attention",
|
| 821 |
+
"full_attention",
|
| 822 |
+
"full_attention",
|
| 823 |
+
"full_attention",
|
| 824 |
+
"full_attention",
|
| 825 |
+
"full_attention",
|
| 826 |
+
"full_attention",
|
| 827 |
+
"full_attention",
|
| 828 |
+
"full_attention",
|
| 829 |
+
"full_attention",
|
| 830 |
+
"full_attention",
|
| 831 |
+
"full_attention",
|
| 832 |
+
"full_attention",
|
| 833 |
+
"full_attention",
|
| 834 |
+
"full_attention",
|
| 835 |
+
"full_attention",
|
| 836 |
+
"full_attention",
|
| 837 |
+
"full_attention",
|
| 838 |
+
"full_attention",
|
| 839 |
+
"full_attention",
|
| 840 |
+
"full_attention"
|
| 841 |
+
],
|
| 842 |
+
"max_position_embeddings": 262144,
|
| 843 |
+
"model_type": "qwen3_vl_text",
|
| 844 |
+
"num_attention_heads": 32,
|
| 845 |
+
"num_hidden_layers": 36,
|
| 846 |
+
"num_key_value_heads": 8,
|
| 847 |
+
"pad_token_id": 0,
|
| 848 |
+
"rms_norm_eps": 1e-06,
|
| 849 |
+
"rope_scaling": {
|
| 850 |
+
"mrope_interleaved": true,
|
| 851 |
+
"mrope_section": [
|
| 852 |
+
24,
|
| 853 |
+
20,
|
| 854 |
+
20
|
| 855 |
+
],
|
| 856 |
+
"rope_type": "default"
|
| 857 |
+
},
|
| 858 |
+
"rope_theta": 5000000,
|
| 859 |
+
"tie_word_embeddings": false,
|
| 860 |
+
"vocab_size": 151936
|
| 861 |
+
},
|
| 862 |
+
"tie_word_embeddings": false,
|
| 863 |
+
"video_token_id": 151656,
|
| 864 |
+
"vision_config": {
|
| 865 |
+
"deepstack_visual_indexes": [
|
| 866 |
+
8,
|
| 867 |
+
16,
|
| 868 |
+
24
|
| 869 |
+
],
|
| 870 |
+
"depth": 27,
|
| 871 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 872 |
+
"hidden_size": 1152,
|
| 873 |
+
"in_channels": 3,
|
| 874 |
+
"initializer_range": 0.02,
|
| 875 |
+
"intermediate_size": 4304,
|
| 876 |
+
"model_type": "qwen3_vl",
|
| 877 |
+
"num_heads": 16,
|
| 878 |
+
"num_position_embeddings": 2304,
|
| 879 |
+
"out_hidden_size": 4096,
|
| 880 |
+
"patch_size": 16,
|
| 881 |
+
"spatial_merge_size": 2,
|
| 882 |
+
"temporal_patch_size": 2
|
| 883 |
+
},
|
| 884 |
+
"vision_end_token_id": 151653,
|
| 885 |
+
"vision_start_token_id": 151652,
|
| 886 |
+
"vocab_size": 151936
|
| 887 |
+
}
|
| 888 |
+
}
|
| 889 |
+
},
|
| 890 |
+
"runner": {
|
| 891 |
+
"change_key_name": false,
|
| 892 |
+
"collator": {
|
| 893 |
+
"list_keys": [
|
| 894 |
+
"sequence_plan"
|
| 895 |
+
],
|
| 896 |
+
"meta_keys": [
|
| 897 |
+
"task_description",
|
| 898 |
+
"stats",
|
| 899 |
+
"info",
|
| 900 |
+
"timestamp"
|
| 901 |
+
],
|
| 902 |
+
"pad_id": 0,
|
| 903 |
+
"sequence_keys": [
|
| 904 |
+
"text_token_ids"
|
| 905 |
+
],
|
| 906 |
+
"tensor_keys": [
|
| 907 |
+
"images",
|
| 908 |
+
"actions",
|
| 909 |
+
"embodiment_ids",
|
| 910 |
+
"raw_action_dim",
|
| 911 |
+
"conditioning_fps",
|
| 912 |
+
"action_fps"
|
| 913 |
+
],
|
| 914 |
+
"type": "Cosmos3Collator"
|
| 915 |
+
},
|
| 916 |
+
"enable_gradient_checkpointing": true,
|
| 917 |
+
"enable_mixed_precision_training": true,
|
| 918 |
+
"grad_accumulation_steps": 8,
|
| 919 |
+
"lr_scheduler": {
|
| 920 |
+
"cycle_length": 16000,
|
| 921 |
+
"type": "linear-warmup+linear-decay",
|
| 922 |
+
"warmup_steps": 500
|
| 923 |
+
},
|
| 924 |
+
"max_grad_norm": 1.0,
|
| 925 |
+
"max_keep_ckpts": 2,
|
| 926 |
+
"max_steps": 5000,
|
| 927 |
+
"metric": {
|
| 928 |
+
"active_trackers": [
|
| 929 |
+
"jsonl",
|
| 930 |
+
"wandb"
|
| 931 |
+
],
|
| 932 |
+
"grad_accumulation_steps": 8,
|
| 933 |
+
"run_dir": "work_dirs",
|
| 934 |
+
"type": "VLAMetric",
|
| 935 |
+
"window_size": 1
|
| 936 |
+
},
|
| 937 |
+
"mixed_precision_dtype": "bf16",
|
| 938 |
+
"optimizer": {
|
| 939 |
+
"betas": [
|
| 940 |
+
0.9,
|
| 941 |
+
0.99
|
| 942 |
+
],
|
| 943 |
+
"eps": 1e-08,
|
| 944 |
+
"exclude_1d_from_weight_decay": false,
|
| 945 |
+
"fused": true,
|
| 946 |
+
"lr": 5e-05,
|
| 947 |
+
"paramwise_learning_rate": {
|
| 948 |
+
"action_in_proj.": 0.00025,
|
| 949 |
+
"action_modality_embed": 0.00025,
|
| 950 |
+
"action_out_proj.": 0.00025
|
| 951 |
+
},
|
| 952 |
+
"type": "AdamW",
|
| 953 |
+
"weight_decay": 0.05
|
| 954 |
+
},
|
| 955 |
+
"sampler": null,
|
| 956 |
+
"save_iter_interval": 500,
|
| 957 |
+
"sharding_strategy": "full-shard",
|
| 958 |
+
"tokenizer": {
|
| 959 |
+
"model_max_length": 4096,
|
| 960 |
+
"model_path": "./checkpoints/Cosmos3-Nano/text_tokenizer",
|
| 961 |
+
"padding_side": "right",
|
| 962 |
+
"trust_remote_code": true,
|
| 963 |
+
"type": "PretrainedTokenizer"
|
| 964 |
+
},
|
| 965 |
+
"type": "FSDPTrainRunner"
|
| 966 |
+
},
|
| 967 |
+
"train_dataloader": {
|
| 968 |
+
"dataset": {
|
| 969 |
+
"datasets": [
|
| 970 |
+
{
|
| 971 |
+
"action_key": "action",
|
| 972 |
+
"action_window_size": 16,
|
| 973 |
+
"data_root_path": "./datasets/libero_spatial_no_noops_lerobotv2.1",
|
| 974 |
+
"frame_window_size": 17,
|
| 975 |
+
"require_full_window": true,
|
| 976 |
+
"statistic_name": "all_libero_no_noops",
|
| 977 |
+
"transforms": [
|
| 978 |
+
{
|
| 979 |
+
"embodiment_id": 5,
|
| 980 |
+
"name_mappings": {
|
| 981 |
+
"actions": [
|
| 982 |
+
"actions"
|
| 983 |
+
],
|
| 984 |
+
"observation.state": [
|
| 985 |
+
"states"
|
| 986 |
+
]
|
| 987 |
+
},
|
| 988 |
+
"parquet_keys": [
|
| 989 |
+
"observation.state",
|
| 990 |
+
"timestamp",
|
| 991 |
+
"actions",
|
| 992 |
+
"info",
|
| 993 |
+
"stats",
|
| 994 |
+
"action_masks"
|
| 995 |
+
],
|
| 996 |
+
"type": "ProcessParquetInputs",
|
| 997 |
+
"video_keys": [
|
| 998 |
+
"observation.images.image",
|
| 999 |
+
"observation.images.wrist_image"
|
| 1000 |
+
]
|
| 1001 |
+
},
|
| 1002 |
+
{
|
| 1003 |
+
"type": "LiberoFramewiseActionToRot6D"
|
| 1004 |
+
},
|
| 1005 |
+
{
|
| 1006 |
+
"action_metadata": {
|
| 1007 |
+
"append_viewpoint": false,
|
| 1008 |
+
"conditioning_fps": 20.0,
|
| 1009 |
+
"frame_window_size": 17,
|
| 1010 |
+
"video_height": 192,
|
| 1011 |
+
"video_width": 320
|
| 1012 |
+
},
|
| 1013 |
+
"cfg_dropout_rate": 0.1,
|
| 1014 |
+
"format_prompt_as_json": true,
|
| 1015 |
+
"max_len": 512,
|
| 1016 |
+
"tokenizer": {
|
| 1017 |
+
"model_max_length": 4096,
|
| 1018 |
+
"model_path": "./checkpoints/Cosmos3-Nano/text_tokenizer",
|
| 1019 |
+
"padding_side": "right",
|
| 1020 |
+
"trust_remote_code": true,
|
| 1021 |
+
"type": "PretrainedTokenizer"
|
| 1022 |
+
},
|
| 1023 |
+
"type": "ProcessCosmos3Prompt"
|
| 1024 |
+
},
|
| 1025 |
+
{
|
| 1026 |
+
"height": 256,
|
| 1027 |
+
"type": "ResizeImages",
|
| 1028 |
+
"width": 256
|
| 1029 |
+
},
|
| 1030 |
+
{
|
| 1031 |
+
"type": "SimpleNormalizeImages"
|
| 1032 |
+
},
|
| 1033 |
+
{
|
| 1034 |
+
"action_dim": 64,
|
| 1035 |
+
"action_key": "action",
|
| 1036 |
+
"action_norm_type": "quantile",
|
| 1037 |
+
"state_dim": 64,
|
| 1038 |
+
"state_key": "proprio",
|
| 1039 |
+
"state_norm_type": "none",
|
| 1040 |
+
"type": "NormalizeStatesAndActions"
|
| 1041 |
+
},
|
| 1042 |
+
{
|
| 1043 |
+
"conditioning_fps": 20.0,
|
| 1044 |
+
"frame_window_size": 17,
|
| 1045 |
+
"mode": "wam",
|
| 1046 |
+
"prepend_state_to_action": false,
|
| 1047 |
+
"raw_action_dim": 10,
|
| 1048 |
+
"type": "BuildCosmos3Sequence"
|
| 1049 |
+
},
|
| 1050 |
+
{
|
| 1051 |
+
"frame_window_size": 17,
|
| 1052 |
+
"num_views": 2,
|
| 1053 |
+
"tile_direction": "horizontal",
|
| 1054 |
+
"type": "PrepareVideo"
|
| 1055 |
+
},
|
| 1056 |
+
{
|
| 1057 |
+
"height": 192,
|
| 1058 |
+
"type": "ResizeAndReflectPad",
|
| 1059 |
+
"width": 320
|
| 1060 |
+
}
|
| 1061 |
+
],
|
| 1062 |
+
"type": "ParquetDataset",
|
| 1063 |
+
"use_delta": false,
|
| 1064 |
+
"window_start_idx": 0
|
| 1065 |
+
},
|
| 1066 |
+
{
|
| 1067 |
+
"action_key": "action",
|
| 1068 |
+
"action_window_size": 16,
|
| 1069 |
+
"data_root_path": "./datasets/libero_spatial_no_noops_lerobotv2.1",
|
| 1070 |
+
"frame_window_size": 17,
|
| 1071 |
+
"require_full_window": true,
|
| 1072 |
+
"statistic_name": "all_libero_no_noops",
|
| 1073 |
+
"transforms": [
|
| 1074 |
+
{
|
| 1075 |
+
"embodiment_id": 5,
|
| 1076 |
+
"name_mappings": {
|
| 1077 |
+
"actions": [
|
| 1078 |
+
"actions"
|
| 1079 |
+
],
|
| 1080 |
+
"observation.state": [
|
| 1081 |
+
"states"
|
| 1082 |
+
]
|
| 1083 |
+
},
|
| 1084 |
+
"parquet_keys": [
|
| 1085 |
+
"observation.state",
|
| 1086 |
+
"timestamp",
|
| 1087 |
+
"actions",
|
| 1088 |
+
"info",
|
| 1089 |
+
"stats",
|
| 1090 |
+
"action_masks"
|
| 1091 |
+
],
|
| 1092 |
+
"type": "ProcessParquetInputs",
|
| 1093 |
+
"video_keys": [
|
| 1094 |
+
"observation.images.image",
|
| 1095 |
+
"observation.images.wrist_image"
|
| 1096 |
+
]
|
| 1097 |
+
},
|
| 1098 |
+
{
|
| 1099 |
+
"type": "LiberoFramewiseActionToRot6D"
|
| 1100 |
+
},
|
| 1101 |
+
{
|
| 1102 |
+
"action_metadata": {
|
| 1103 |
+
"append_viewpoint": false,
|
| 1104 |
+
"conditioning_fps": 20.0,
|
| 1105 |
+
"frame_window_size": 17,
|
| 1106 |
+
"video_height": 192,
|
| 1107 |
+
"video_width": 320
|
| 1108 |
+
},
|
| 1109 |
+
"cfg_dropout_rate": 0.1,
|
| 1110 |
+
"format_prompt_as_json": true,
|
| 1111 |
+
"max_len": 512,
|
| 1112 |
+
"tokenizer": {
|
| 1113 |
+
"model_max_length": 4096,
|
| 1114 |
+
"model_path": "./checkpoints/Cosmos3-Nano/text_tokenizer",
|
| 1115 |
+
"padding_side": "right",
|
| 1116 |
+
"trust_remote_code": true,
|
| 1117 |
+
"type": "PretrainedTokenizer"
|
| 1118 |
+
},
|
| 1119 |
+
"type": "ProcessCosmos3Prompt"
|
| 1120 |
+
},
|
| 1121 |
+
{
|
| 1122 |
+
"height": 256,
|
| 1123 |
+
"type": "ResizeImages",
|
| 1124 |
+
"width": 256
|
| 1125 |
+
},
|
| 1126 |
+
{
|
| 1127 |
+
"type": "SimpleNormalizeImages"
|
| 1128 |
+
},
|
| 1129 |
+
{
|
| 1130 |
+
"action_dim": 64,
|
| 1131 |
+
"action_key": "action",
|
| 1132 |
+
"action_norm_type": "quantile",
|
| 1133 |
+
"state_dim": 64,
|
| 1134 |
+
"state_key": "proprio",
|
| 1135 |
+
"state_norm_type": "none",
|
| 1136 |
+
"type": "NormalizeStatesAndActions"
|
| 1137 |
+
},
|
| 1138 |
+
{
|
| 1139 |
+
"conditioning_fps": 20.0,
|
| 1140 |
+
"frame_window_size": 17,
|
| 1141 |
+
"mode": "wam",
|
| 1142 |
+
"prepend_state_to_action": false,
|
| 1143 |
+
"raw_action_dim": 10,
|
| 1144 |
+
"type": "BuildCosmos3Sequence"
|
| 1145 |
+
},
|
| 1146 |
+
{
|
| 1147 |
+
"frame_window_size": 17,
|
| 1148 |
+
"num_views": 2,
|
| 1149 |
+
"tile_direction": "horizontal",
|
| 1150 |
+
"type": "PrepareVideo"
|
| 1151 |
+
},
|
| 1152 |
+
{
|
| 1153 |
+
"height": 192,
|
| 1154 |
+
"type": "ResizeAndReflectPad",
|
| 1155 |
+
"width": 320
|
| 1156 |
+
}
|
| 1157 |
+
],
|
| 1158 |
+
"type": "ParquetDataset",
|
| 1159 |
+
"use_delta": false,
|
| 1160 |
+
"window_start_idx": 0
|
| 1161 |
+
},
|
| 1162 |
+
{
|
| 1163 |
+
"action_key": "action",
|
| 1164 |
+
"action_window_size": 16,
|
| 1165 |
+
"data_root_path": "./datasets/libero_object_no_noops_lerobotv2.1",
|
| 1166 |
+
"frame_window_size": 17,
|
| 1167 |
+
"require_full_window": true,
|
| 1168 |
+
"statistic_name": "all_libero_no_noops",
|
| 1169 |
+
"transforms": [
|
| 1170 |
+
{
|
| 1171 |
+
"embodiment_id": 5,
|
| 1172 |
+
"name_mappings": {
|
| 1173 |
+
"actions": [
|
| 1174 |
+
"actions"
|
| 1175 |
+
],
|
| 1176 |
+
"observation.state": [
|
| 1177 |
+
"states"
|
| 1178 |
+
]
|
| 1179 |
+
},
|
| 1180 |
+
"parquet_keys": [
|
| 1181 |
+
"observation.state",
|
| 1182 |
+
"timestamp",
|
| 1183 |
+
"actions",
|
| 1184 |
+
"info",
|
| 1185 |
+
"stats",
|
| 1186 |
+
"action_masks"
|
| 1187 |
+
],
|
| 1188 |
+
"type": "ProcessParquetInputs",
|
| 1189 |
+
"video_keys": [
|
| 1190 |
+
"observation.images.image",
|
| 1191 |
+
"observation.images.wrist_image"
|
| 1192 |
+
]
|
| 1193 |
+
},
|
| 1194 |
+
{
|
| 1195 |
+
"type": "LiberoFramewiseActionToRot6D"
|
| 1196 |
+
},
|
| 1197 |
+
{
|
| 1198 |
+
"action_metadata": {
|
| 1199 |
+
"append_viewpoint": false,
|
| 1200 |
+
"conditioning_fps": 20.0,
|
| 1201 |
+
"frame_window_size": 17,
|
| 1202 |
+
"video_height": 192,
|
| 1203 |
+
"video_width": 320
|
| 1204 |
+
},
|
| 1205 |
+
"cfg_dropout_rate": 0.1,
|
| 1206 |
+
"format_prompt_as_json": true,
|
| 1207 |
+
"max_len": 512,
|
| 1208 |
+
"tokenizer": {
|
| 1209 |
+
"model_max_length": 4096,
|
| 1210 |
+
"model_path": "./checkpoints/Cosmos3-Nano/text_tokenizer",
|
| 1211 |
+
"padding_side": "right",
|
| 1212 |
+
"trust_remote_code": true,
|
| 1213 |
+
"type": "PretrainedTokenizer"
|
| 1214 |
+
},
|
| 1215 |
+
"type": "ProcessCosmos3Prompt"
|
| 1216 |
+
},
|
| 1217 |
+
{
|
| 1218 |
+
"height": 256,
|
| 1219 |
+
"type": "ResizeImages",
|
| 1220 |
+
"width": 256
|
| 1221 |
+
},
|
| 1222 |
+
{
|
| 1223 |
+
"type": "SimpleNormalizeImages"
|
| 1224 |
+
},
|
| 1225 |
+
{
|
| 1226 |
+
"action_dim": 64,
|
| 1227 |
+
"action_key": "action",
|
| 1228 |
+
"action_norm_type": "quantile",
|
| 1229 |
+
"state_dim": 64,
|
| 1230 |
+
"state_key": "proprio",
|
| 1231 |
+
"state_norm_type": "none",
|
| 1232 |
+
"type": "NormalizeStatesAndActions"
|
| 1233 |
+
},
|
| 1234 |
+
{
|
| 1235 |
+
"conditioning_fps": 20.0,
|
| 1236 |
+
"frame_window_size": 17,
|
| 1237 |
+
"mode": "wam",
|
| 1238 |
+
"prepend_state_to_action": false,
|
| 1239 |
+
"raw_action_dim": 10,
|
| 1240 |
+
"type": "BuildCosmos3Sequence"
|
| 1241 |
+
},
|
| 1242 |
+
{
|
| 1243 |
+
"frame_window_size": 17,
|
| 1244 |
+
"num_views": 2,
|
| 1245 |
+
"tile_direction": "horizontal",
|
| 1246 |
+
"type": "PrepareVideo"
|
| 1247 |
+
},
|
| 1248 |
+
{
|
| 1249 |
+
"height": 192,
|
| 1250 |
+
"type": "ResizeAndReflectPad",
|
| 1251 |
+
"width": 320
|
| 1252 |
+
}
|
| 1253 |
+
],
|
| 1254 |
+
"type": "ParquetDataset",
|
| 1255 |
+
"use_delta": false,
|
| 1256 |
+
"window_start_idx": 0
|
| 1257 |
+
},
|
| 1258 |
+
{
|
| 1259 |
+
"action_key": "action",
|
| 1260 |
+
"action_window_size": 16,
|
| 1261 |
+
"data_root_path": "./datasets/libero_object_no_noops_lerobotv2.1",
|
| 1262 |
+
"frame_window_size": 17,
|
| 1263 |
+
"require_full_window": true,
|
| 1264 |
+
"statistic_name": "all_libero_no_noops",
|
| 1265 |
+
"transforms": [
|
| 1266 |
+
{
|
| 1267 |
+
"embodiment_id": 5,
|
| 1268 |
+
"name_mappings": {
|
| 1269 |
+
"actions": [
|
| 1270 |
+
"actions"
|
| 1271 |
+
],
|
| 1272 |
+
"observation.state": [
|
| 1273 |
+
"states"
|
| 1274 |
+
]
|
| 1275 |
+
},
|
| 1276 |
+
"parquet_keys": [
|
| 1277 |
+
"observation.state",
|
| 1278 |
+
"timestamp",
|
| 1279 |
+
"actions",
|
| 1280 |
+
"info",
|
| 1281 |
+
"stats",
|
| 1282 |
+
"action_masks"
|
| 1283 |
+
],
|
| 1284 |
+
"type": "ProcessParquetInputs",
|
| 1285 |
+
"video_keys": [
|
| 1286 |
+
"observation.images.image",
|
| 1287 |
+
"observation.images.wrist_image"
|
| 1288 |
+
]
|
| 1289 |
+
},
|
| 1290 |
+
{
|
| 1291 |
+
"type": "LiberoFramewiseActionToRot6D"
|
| 1292 |
+
},
|
| 1293 |
+
{
|
| 1294 |
+
"action_metadata": {
|
| 1295 |
+
"append_viewpoint": false,
|
| 1296 |
+
"conditioning_fps": 20.0,
|
| 1297 |
+
"frame_window_size": 17,
|
| 1298 |
+
"video_height": 192,
|
| 1299 |
+
"video_width": 320
|
| 1300 |
+
},
|
| 1301 |
+
"cfg_dropout_rate": 0.1,
|
| 1302 |
+
"format_prompt_as_json": true,
|
| 1303 |
+
"max_len": 512,
|
| 1304 |
+
"tokenizer": {
|
| 1305 |
+
"model_max_length": 4096,
|
| 1306 |
+
"model_path": "./checkpoints/Cosmos3-Nano/text_tokenizer",
|
| 1307 |
+
"padding_side": "right",
|
| 1308 |
+
"trust_remote_code": true,
|
| 1309 |
+
"type": "PretrainedTokenizer"
|
| 1310 |
+
},
|
| 1311 |
+
"type": "ProcessCosmos3Prompt"
|
| 1312 |
+
},
|
| 1313 |
+
{
|
| 1314 |
+
"height": 256,
|
| 1315 |
+
"type": "ResizeImages",
|
| 1316 |
+
"width": 256
|
| 1317 |
+
},
|
| 1318 |
+
{
|
| 1319 |
+
"type": "SimpleNormalizeImages"
|
| 1320 |
+
},
|
| 1321 |
+
{
|
| 1322 |
+
"action_dim": 64,
|
| 1323 |
+
"action_key": "action",
|
| 1324 |
+
"action_norm_type": "quantile",
|
| 1325 |
+
"state_dim": 64,
|
| 1326 |
+
"state_key": "proprio",
|
| 1327 |
+
"state_norm_type": "none",
|
| 1328 |
+
"type": "NormalizeStatesAndActions"
|
| 1329 |
+
},
|
| 1330 |
+
{
|
| 1331 |
+
"conditioning_fps": 20.0,
|
| 1332 |
+
"frame_window_size": 17,
|
| 1333 |
+
"mode": "wam",
|
| 1334 |
+
"prepend_state_to_action": false,
|
| 1335 |
+
"raw_action_dim": 10,
|
| 1336 |
+
"type": "BuildCosmos3Sequence"
|
| 1337 |
+
},
|
| 1338 |
+
{
|
| 1339 |
+
"frame_window_size": 17,
|
| 1340 |
+
"num_views": 2,
|
| 1341 |
+
"tile_direction": "horizontal",
|
| 1342 |
+
"type": "PrepareVideo"
|
| 1343 |
+
},
|
| 1344 |
+
{
|
| 1345 |
+
"height": 192,
|
| 1346 |
+
"type": "ResizeAndReflectPad",
|
| 1347 |
+
"width": 320
|
| 1348 |
+
}
|
| 1349 |
+
],
|
| 1350 |
+
"type": "ParquetDataset",
|
| 1351 |
+
"use_delta": false,
|
| 1352 |
+
"window_start_idx": 0
|
| 1353 |
+
},
|
| 1354 |
+
{
|
| 1355 |
+
"action_key": "action",
|
| 1356 |
+
"action_window_size": 16,
|
| 1357 |
+
"data_root_path": "./datasets/libero_goal_no_noops_lerobotv2.1",
|
| 1358 |
+
"frame_window_size": 17,
|
| 1359 |
+
"require_full_window": true,
|
| 1360 |
+
"statistic_name": "all_libero_no_noops",
|
| 1361 |
+
"transforms": [
|
| 1362 |
+
{
|
| 1363 |
+
"embodiment_id": 5,
|
| 1364 |
+
"name_mappings": {
|
| 1365 |
+
"actions": [
|
| 1366 |
+
"actions"
|
| 1367 |
+
],
|
| 1368 |
+
"observation.state": [
|
| 1369 |
+
"states"
|
| 1370 |
+
]
|
| 1371 |
+
},
|
| 1372 |
+
"parquet_keys": [
|
| 1373 |
+
"observation.state",
|
| 1374 |
+
"timestamp",
|
| 1375 |
+
"actions",
|
| 1376 |
+
"info",
|
| 1377 |
+
"stats",
|
| 1378 |
+
"action_masks"
|
| 1379 |
+
],
|
| 1380 |
+
"type": "ProcessParquetInputs",
|
| 1381 |
+
"video_keys": [
|
| 1382 |
+
"observation.images.image",
|
| 1383 |
+
"observation.images.wrist_image"
|
| 1384 |
+
]
|
| 1385 |
+
},
|
| 1386 |
+
{
|
| 1387 |
+
"type": "LiberoFramewiseActionToRot6D"
|
| 1388 |
+
},
|
| 1389 |
+
{
|
| 1390 |
+
"action_metadata": {
|
| 1391 |
+
"append_viewpoint": false,
|
| 1392 |
+
"conditioning_fps": 20.0,
|
| 1393 |
+
"frame_window_size": 17,
|
| 1394 |
+
"video_height": 192,
|
| 1395 |
+
"video_width": 320
|
| 1396 |
+
},
|
| 1397 |
+
"cfg_dropout_rate": 0.1,
|
| 1398 |
+
"format_prompt_as_json": true,
|
| 1399 |
+
"max_len": 512,
|
| 1400 |
+
"tokenizer": {
|
| 1401 |
+
"model_max_length": 4096,
|
| 1402 |
+
"model_path": "./checkpoints/Cosmos3-Nano/text_tokenizer",
|
| 1403 |
+
"padding_side": "right",
|
| 1404 |
+
"trust_remote_code": true,
|
| 1405 |
+
"type": "PretrainedTokenizer"
|
| 1406 |
+
},
|
| 1407 |
+
"type": "ProcessCosmos3Prompt"
|
| 1408 |
+
},
|
| 1409 |
+
{
|
| 1410 |
+
"height": 256,
|
| 1411 |
+
"type": "ResizeImages",
|
| 1412 |
+
"width": 256
|
| 1413 |
+
},
|
| 1414 |
+
{
|
| 1415 |
+
"type": "SimpleNormalizeImages"
|
| 1416 |
+
},
|
| 1417 |
+
{
|
| 1418 |
+
"action_dim": 64,
|
| 1419 |
+
"action_key": "action",
|
| 1420 |
+
"action_norm_type": "quantile",
|
| 1421 |
+
"state_dim": 64,
|
| 1422 |
+
"state_key": "proprio",
|
| 1423 |
+
"state_norm_type": "none",
|
| 1424 |
+
"type": "NormalizeStatesAndActions"
|
| 1425 |
+
},
|
| 1426 |
+
{
|
| 1427 |
+
"conditioning_fps": 20.0,
|
| 1428 |
+
"frame_window_size": 17,
|
| 1429 |
+
"mode": "wam",
|
| 1430 |
+
"prepend_state_to_action": false,
|
| 1431 |
+
"raw_action_dim": 10,
|
| 1432 |
+
"type": "BuildCosmos3Sequence"
|
| 1433 |
+
},
|
| 1434 |
+
{
|
| 1435 |
+
"frame_window_size": 17,
|
| 1436 |
+
"num_views": 2,
|
| 1437 |
+
"tile_direction": "horizontal",
|
| 1438 |
+
"type": "PrepareVideo"
|
| 1439 |
+
},
|
| 1440 |
+
{
|
| 1441 |
+
"height": 192,
|
| 1442 |
+
"type": "ResizeAndReflectPad",
|
| 1443 |
+
"width": 320
|
| 1444 |
+
}
|
| 1445 |
+
],
|
| 1446 |
+
"type": "ParquetDataset",
|
| 1447 |
+
"use_delta": false,
|
| 1448 |
+
"window_start_idx": 0
|
| 1449 |
+
},
|
| 1450 |
+
{
|
| 1451 |
+
"action_key": "action",
|
| 1452 |
+
"action_window_size": 16,
|
| 1453 |
+
"data_root_path": "./datasets/libero_goal_no_noops_lerobotv2.1",
|
| 1454 |
+
"frame_window_size": 17,
|
| 1455 |
+
"require_full_window": true,
|
| 1456 |
+
"statistic_name": "all_libero_no_noops",
|
| 1457 |
+
"transforms": [
|
| 1458 |
+
{
|
| 1459 |
+
"embodiment_id": 5,
|
| 1460 |
+
"name_mappings": {
|
| 1461 |
+
"actions": [
|
| 1462 |
+
"actions"
|
| 1463 |
+
],
|
| 1464 |
+
"observation.state": [
|
| 1465 |
+
"states"
|
| 1466 |
+
]
|
| 1467 |
+
},
|
| 1468 |
+
"parquet_keys": [
|
| 1469 |
+
"observation.state",
|
| 1470 |
+
"timestamp",
|
| 1471 |
+
"actions",
|
| 1472 |
+
"info",
|
| 1473 |
+
"stats",
|
| 1474 |
+
"action_masks"
|
| 1475 |
+
],
|
| 1476 |
+
"type": "ProcessParquetInputs",
|
| 1477 |
+
"video_keys": [
|
| 1478 |
+
"observation.images.image",
|
| 1479 |
+
"observation.images.wrist_image"
|
| 1480 |
+
]
|
| 1481 |
+
},
|
| 1482 |
+
{
|
| 1483 |
+
"type": "LiberoFramewiseActionToRot6D"
|
| 1484 |
+
},
|
| 1485 |
+
{
|
| 1486 |
+
"action_metadata": {
|
| 1487 |
+
"append_viewpoint": false,
|
| 1488 |
+
"conditioning_fps": 20.0,
|
| 1489 |
+
"frame_window_size": 17,
|
| 1490 |
+
"video_height": 192,
|
| 1491 |
+
"video_width": 320
|
| 1492 |
+
},
|
| 1493 |
+
"cfg_dropout_rate": 0.1,
|
| 1494 |
+
"format_prompt_as_json": true,
|
| 1495 |
+
"max_len": 512,
|
| 1496 |
+
"tokenizer": {
|
| 1497 |
+
"model_max_length": 4096,
|
| 1498 |
+
"model_path": "./checkpoints/Cosmos3-Nano/text_tokenizer",
|
| 1499 |
+
"padding_side": "right",
|
| 1500 |
+
"trust_remote_code": true,
|
| 1501 |
+
"type": "PretrainedTokenizer"
|
| 1502 |
+
},
|
| 1503 |
+
"type": "ProcessCosmos3Prompt"
|
| 1504 |
+
},
|
| 1505 |
+
{
|
| 1506 |
+
"height": 256,
|
| 1507 |
+
"type": "ResizeImages",
|
| 1508 |
+
"width": 256
|
| 1509 |
+
},
|
| 1510 |
+
{
|
| 1511 |
+
"type": "SimpleNormalizeImages"
|
| 1512 |
+
},
|
| 1513 |
+
{
|
| 1514 |
+
"action_dim": 64,
|
| 1515 |
+
"action_key": "action",
|
| 1516 |
+
"action_norm_type": "quantile",
|
| 1517 |
+
"state_dim": 64,
|
| 1518 |
+
"state_key": "proprio",
|
| 1519 |
+
"state_norm_type": "none",
|
| 1520 |
+
"type": "NormalizeStatesAndActions"
|
| 1521 |
+
},
|
| 1522 |
+
{
|
| 1523 |
+
"conditioning_fps": 20.0,
|
| 1524 |
+
"frame_window_size": 17,
|
| 1525 |
+
"mode": "wam",
|
| 1526 |
+
"prepend_state_to_action": false,
|
| 1527 |
+
"raw_action_dim": 10,
|
| 1528 |
+
"type": "BuildCosmos3Sequence"
|
| 1529 |
+
},
|
| 1530 |
+
{
|
| 1531 |
+
"frame_window_size": 17,
|
| 1532 |
+
"num_views": 2,
|
| 1533 |
+
"tile_direction": "horizontal",
|
| 1534 |
+
"type": "PrepareVideo"
|
| 1535 |
+
},
|
| 1536 |
+
{
|
| 1537 |
+
"height": 192,
|
| 1538 |
+
"type": "ResizeAndReflectPad",
|
| 1539 |
+
"width": 320
|
| 1540 |
+
}
|
| 1541 |
+
],
|
| 1542 |
+
"type": "ParquetDataset",
|
| 1543 |
+
"use_delta": false,
|
| 1544 |
+
"window_start_idx": 0
|
| 1545 |
+
},
|
| 1546 |
+
{
|
| 1547 |
+
"action_key": "action",
|
| 1548 |
+
"action_window_size": 16,
|
| 1549 |
+
"data_root_path": "./datasets/libero_10_no_noops_lerobotv2.1",
|
| 1550 |
+
"frame_window_size": 17,
|
| 1551 |
+
"require_full_window": true,
|
| 1552 |
+
"statistic_name": "all_libero_no_noops",
|
| 1553 |
+
"transforms": [
|
| 1554 |
+
{
|
| 1555 |
+
"embodiment_id": 5,
|
| 1556 |
+
"name_mappings": {
|
| 1557 |
+
"actions": [
|
| 1558 |
+
"actions"
|
| 1559 |
+
],
|
| 1560 |
+
"observation.state": [
|
| 1561 |
+
"states"
|
| 1562 |
+
]
|
| 1563 |
+
},
|
| 1564 |
+
"parquet_keys": [
|
| 1565 |
+
"observation.state",
|
| 1566 |
+
"timestamp",
|
| 1567 |
+
"actions",
|
| 1568 |
+
"info",
|
| 1569 |
+
"stats",
|
| 1570 |
+
"action_masks"
|
| 1571 |
+
],
|
| 1572 |
+
"type": "ProcessParquetInputs",
|
| 1573 |
+
"video_keys": [
|
| 1574 |
+
"observation.images.image",
|
| 1575 |
+
"observation.images.wrist_image"
|
| 1576 |
+
]
|
| 1577 |
+
},
|
| 1578 |
+
{
|
| 1579 |
+
"type": "LiberoFramewiseActionToRot6D"
|
| 1580 |
+
},
|
| 1581 |
+
{
|
| 1582 |
+
"action_metadata": {
|
| 1583 |
+
"append_viewpoint": false,
|
| 1584 |
+
"conditioning_fps": 20.0,
|
| 1585 |
+
"frame_window_size": 17,
|
| 1586 |
+
"video_height": 192,
|
| 1587 |
+
"video_width": 320
|
| 1588 |
+
},
|
| 1589 |
+
"cfg_dropout_rate": 0.1,
|
| 1590 |
+
"format_prompt_as_json": true,
|
| 1591 |
+
"max_len": 512,
|
| 1592 |
+
"tokenizer": {
|
| 1593 |
+
"model_max_length": 4096,
|
| 1594 |
+
"model_path": "./checkpoints/Cosmos3-Nano/text_tokenizer",
|
| 1595 |
+
"padding_side": "right",
|
| 1596 |
+
"trust_remote_code": true,
|
| 1597 |
+
"type": "PretrainedTokenizer"
|
| 1598 |
+
},
|
| 1599 |
+
"type": "ProcessCosmos3Prompt"
|
| 1600 |
+
},
|
| 1601 |
+
{
|
| 1602 |
+
"height": 256,
|
| 1603 |
+
"type": "ResizeImages",
|
| 1604 |
+
"width": 256
|
| 1605 |
+
},
|
| 1606 |
+
{
|
| 1607 |
+
"type": "SimpleNormalizeImages"
|
| 1608 |
+
},
|
| 1609 |
+
{
|
| 1610 |
+
"action_dim": 64,
|
| 1611 |
+
"action_key": "action",
|
| 1612 |
+
"action_norm_type": "quantile",
|
| 1613 |
+
"state_dim": 64,
|
| 1614 |
+
"state_key": "proprio",
|
| 1615 |
+
"state_norm_type": "none",
|
| 1616 |
+
"type": "NormalizeStatesAndActions"
|
| 1617 |
+
},
|
| 1618 |
+
{
|
| 1619 |
+
"conditioning_fps": 20.0,
|
| 1620 |
+
"frame_window_size": 17,
|
| 1621 |
+
"mode": "wam",
|
| 1622 |
+
"prepend_state_to_action": false,
|
| 1623 |
+
"raw_action_dim": 10,
|
| 1624 |
+
"type": "BuildCosmos3Sequence"
|
| 1625 |
+
},
|
| 1626 |
+
{
|
| 1627 |
+
"frame_window_size": 17,
|
| 1628 |
+
"num_views": 2,
|
| 1629 |
+
"tile_direction": "horizontal",
|
| 1630 |
+
"type": "PrepareVideo"
|
| 1631 |
+
},
|
| 1632 |
+
{
|
| 1633 |
+
"height": 192,
|
| 1634 |
+
"type": "ResizeAndReflectPad",
|
| 1635 |
+
"width": 320
|
| 1636 |
+
}
|
| 1637 |
+
],
|
| 1638 |
+
"type": "ParquetDataset",
|
| 1639 |
+
"use_delta": false,
|
| 1640 |
+
"window_start_idx": 0
|
| 1641 |
+
}
|
| 1642 |
+
],
|
| 1643 |
+
"name_mappings": {
|
| 1644 |
+
"action": [
|
| 1645 |
+
"action"
|
| 1646 |
+
],
|
| 1647 |
+
"observation.state": [
|
| 1648 |
+
"proprio"
|
| 1649 |
+
]
|
| 1650 |
+
},
|
| 1651 |
+
"statistic_keys": [
|
| 1652 |
+
"observation.state",
|
| 1653 |
+
"timestamp",
|
| 1654 |
+
"action"
|
| 1655 |
+
],
|
| 1656 |
+
"statistic_name": "all_libero_no_noops",
|
| 1657 |
+
"statistics_overrides": {
|
| 1658 |
+
"all_libero_no_noops": {
|
| 1659 |
+
"action": {
|
| 1660 |
+
"max": [
|
| 1661 |
+
0.9375,
|
| 1662 |
+
0.9375,
|
| 1663 |
+
0.9375,
|
| 1664 |
+
1.0,
|
| 1665 |
+
0.368853,
|
| 1666 |
+
0.341214,
|
| 1667 |
+
0.356395,
|
| 1668 |
+
1.0,
|
| 1669 |
+
0.348251,
|
| 1670 |
+
1.0
|
| 1671 |
+
],
|
| 1672 |
+
"mean": [
|
| 1673 |
+
0.050704,
|
| 1674 |
+
0.097407,
|
| 1675 |
+
-0.094833,
|
| 1676 |
+
0.994873,
|
| 1677 |
+
-0.004579,
|
| 1678 |
+
-0.004288,
|
| 1679 |
+
0.004389,
|
| 1680 |
+
0.996104,
|
| 1681 |
+
0.001109,
|
| 1682 |
+
0.476725
|
| 1683 |
+
],
|
| 1684 |
+
"min": [
|
| 1685 |
+
-0.9375,
|
| 1686 |
+
-0.9375,
|
| 1687 |
+
-0.9375,
|
| 1688 |
+
0.902028,
|
| 1689 |
+
-0.356085,
|
| 1690 |
+
-0.367416,
|
| 1691 |
+
-0.370434,
|
| 1692 |
+
0.921907,
|
| 1693 |
+
-0.255,
|
| 1694 |
+
0.0
|
| 1695 |
+
],
|
| 1696 |
+
"q01": [
|
| 1697 |
+
-0.723214,
|
| 1698 |
+
-0.808929,
|
| 1699 |
+
-0.9375,
|
| 1700 |
+
0.934955,
|
| 1701 |
+
-0.223431,
|
| 1702 |
+
-0.189878,
|
| 1703 |
+
-0.334735,
|
| 1704 |
+
0.938516,
|
| 1705 |
+
-0.107736,
|
| 1706 |
+
0.0
|
| 1707 |
+
],
|
| 1708 |
+
"q99": [
|
| 1709 |
+
0.9375,
|
| 1710 |
+
0.870536,
|
| 1711 |
+
0.9375,
|
| 1712 |
+
1.0,
|
| 1713 |
+
0.331,
|
| 1714 |
+
0.163153,
|
| 1715 |
+
0.226216,
|
| 1716 |
+
1.0,
|
| 1717 |
+
0.127158,
|
| 1718 |
+
1.0
|
| 1719 |
+
],
|
| 1720 |
+
"std": [
|
| 1721 |
+
0.333621,
|
| 1722 |
+
0.387175,
|
| 1723 |
+
0.45714,
|
| 1724 |
+
0.010807,
|
| 1725 |
+
0.077802,
|
| 1726 |
+
0.063386,
|
| 1727 |
+
0.078571,
|
| 1728 |
+
0.009994,
|
| 1729 |
+
0.038504,
|
| 1730 |
+
0.49946
|
| 1731 |
+
]
|
| 1732 |
+
}
|
| 1733 |
+
}
|
| 1734 |
+
},
|
| 1735 |
+
"type": "DistributedRepeatingDataset"
|
| 1736 |
+
},
|
| 1737 |
+
"per_device_batch_size": 16,
|
| 1738 |
+
"per_device_num_workers": 4
|
| 1739 |
+
}
|
| 1740 |
+
}
|
cosmos3_nano_libero_full_finetune_bs2048/config.yaml
ADDED
|
@@ -0,0 +1,1392 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
_action_dim: 10
|
| 2 |
+
_action_horizon: 16
|
| 3 |
+
_action_lr: 0.00025
|
| 4 |
+
_base_lr: 5.0e-05
|
| 5 |
+
_cfg_dropout_rate: 0.1
|
| 6 |
+
_ckpt_root: ./checkpoints
|
| 7 |
+
_conditioning_fps: 20.0
|
| 8 |
+
_cosmos3_nano_ckpt: ./checkpoints/Cosmos3-Nano
|
| 9 |
+
_cosmos3_nano_name_mapping:
|
| 10 |
+
.self_attn.k_norm.: .self_attn.norm_k.
|
| 11 |
+
.self_attn.k_norm_moe_gen.: .self_attn.norm_added_k.
|
| 12 |
+
.self_attn.k_proj.: .self_attn.to_k.
|
| 13 |
+
.self_attn.k_proj_moe_gen.: .self_attn.add_k_proj.
|
| 14 |
+
.self_attn.o_proj.: .self_attn.to_out.
|
| 15 |
+
.self_attn.o_proj_moe_gen.: .self_attn.to_add_out.
|
| 16 |
+
.self_attn.q_norm.: .self_attn.norm_q.
|
| 17 |
+
.self_attn.q_norm_moe_gen.: .self_attn.norm_added_q.
|
| 18 |
+
.self_attn.q_proj.: .self_attn.to_q.
|
| 19 |
+
.self_attn.q_proj_moe_gen.: .self_attn.add_q_proj.
|
| 20 |
+
.self_attn.v_proj.: .self_attn.to_v.
|
| 21 |
+
.self_attn.v_proj_moe_gen.: .self_attn.add_v_proj.
|
| 22 |
+
action_in_proj.: action_proj_in.
|
| 23 |
+
action_modality_embed: action_modality_embed
|
| 24 |
+
action_out_proj.: action_proj_out.
|
| 25 |
+
time_embedder.mlp.0.: time_embedder.linear_1.
|
| 26 |
+
time_embedder.mlp.2.: time_embedder.linear_2.
|
| 27 |
+
vision_in_proj.projector.: proj_in.
|
| 28 |
+
vision_out_proj.projector.: proj_out.
|
| 29 |
+
vlm_backbone.lm_head.weight: lm_head.weight
|
| 30 |
+
vlm_backbone.model.language_model.embed_tokens.weight: embed_tokens.weight
|
| 31 |
+
vlm_backbone.model.language_model.layers.: layers.
|
| 32 |
+
vlm_backbone.model.language_model.norm.weight: norm.weight
|
| 33 |
+
vlm_backbone.model.language_model.norm_moe_gen.weight: norm_moe_gen.weight
|
| 34 |
+
vlm_backbone.model.visual.blocks.: blocks.
|
| 35 |
+
vlm_backbone.model.visual.deepstack_merger_list.: deepstack_merger_list.
|
| 36 |
+
vlm_backbone.model.visual.merger.: merger.
|
| 37 |
+
vlm_backbone.model.visual.patch_embed.: patch_embed.
|
| 38 |
+
vlm_backbone.model.visual.pos_embed.: pos_embed.
|
| 39 |
+
_cosmos3_nano_special_tokens:
|
| 40 |
+
end_of_generation: 151653
|
| 41 |
+
eos_token_id: 151645
|
| 42 |
+
start_of_generation: 151652
|
| 43 |
+
_cosmos3_nano_tokenizer:
|
| 44 |
+
model_max_length: 4096
|
| 45 |
+
model_path: ./checkpoints/Cosmos3-Nano/text_tokenizer
|
| 46 |
+
padding_side: right
|
| 47 |
+
trust_remote_code: true
|
| 48 |
+
type: PretrainedTokenizer
|
| 49 |
+
_cosmos3_nano_transformer: ./checkpoints/Cosmos3-Nano/transformer
|
| 50 |
+
_cosmos3_nano_vision_encoder: ./checkpoints/Cosmos3-Nano/vision_encoder
|
| 51 |
+
_cosmos3_nano_vlm_config:
|
| 52 |
+
image_token_id: 151655
|
| 53 |
+
model_type: qwen3_vl
|
| 54 |
+
text_config:
|
| 55 |
+
attention_bias: false
|
| 56 |
+
attention_dropout: 0.0
|
| 57 |
+
bos_token_id: 151643
|
| 58 |
+
eos_token_id: 151645
|
| 59 |
+
head_dim: 128
|
| 60 |
+
hidden_act: silu
|
| 61 |
+
hidden_size: 4096
|
| 62 |
+
intermediate_size: 12288
|
| 63 |
+
layer_types:
|
| 64 |
+
- full_attention
|
| 65 |
+
- full_attention
|
| 66 |
+
- full_attention
|
| 67 |
+
- full_attention
|
| 68 |
+
- full_attention
|
| 69 |
+
- full_attention
|
| 70 |
+
- full_attention
|
| 71 |
+
- full_attention
|
| 72 |
+
- full_attention
|
| 73 |
+
- full_attention
|
| 74 |
+
- full_attention
|
| 75 |
+
- full_attention
|
| 76 |
+
- full_attention
|
| 77 |
+
- full_attention
|
| 78 |
+
- full_attention
|
| 79 |
+
- full_attention
|
| 80 |
+
- full_attention
|
| 81 |
+
- full_attention
|
| 82 |
+
- full_attention
|
| 83 |
+
- full_attention
|
| 84 |
+
- full_attention
|
| 85 |
+
- full_attention
|
| 86 |
+
- full_attention
|
| 87 |
+
- full_attention
|
| 88 |
+
- full_attention
|
| 89 |
+
- full_attention
|
| 90 |
+
- full_attention
|
| 91 |
+
- full_attention
|
| 92 |
+
- full_attention
|
| 93 |
+
- full_attention
|
| 94 |
+
- full_attention
|
| 95 |
+
- full_attention
|
| 96 |
+
- full_attention
|
| 97 |
+
- full_attention
|
| 98 |
+
- full_attention
|
| 99 |
+
- full_attention
|
| 100 |
+
max_position_embeddings: 262144
|
| 101 |
+
model_type: qwen3_vl_text
|
| 102 |
+
num_attention_heads: 32
|
| 103 |
+
num_hidden_layers: 36
|
| 104 |
+
num_key_value_heads: 8
|
| 105 |
+
pad_token_id: 0
|
| 106 |
+
rms_norm_eps: 1.0e-06
|
| 107 |
+
rope_scaling:
|
| 108 |
+
mrope_interleaved: true
|
| 109 |
+
mrope_section:
|
| 110 |
+
- 24
|
| 111 |
+
- 20
|
| 112 |
+
- 20
|
| 113 |
+
rope_type: default
|
| 114 |
+
rope_theta: 5000000
|
| 115 |
+
tie_word_embeddings: false
|
| 116 |
+
vocab_size: 151936
|
| 117 |
+
tie_word_embeddings: false
|
| 118 |
+
video_token_id: 151656
|
| 119 |
+
vision_config:
|
| 120 |
+
deepstack_visual_indexes:
|
| 121 |
+
- 8
|
| 122 |
+
- 16
|
| 123 |
+
- 24
|
| 124 |
+
depth: 27
|
| 125 |
+
hidden_act: gelu_pytorch_tanh
|
| 126 |
+
hidden_size: 1152
|
| 127 |
+
in_channels: 3
|
| 128 |
+
initializer_range: 0.02
|
| 129 |
+
intermediate_size: 4304
|
| 130 |
+
model_type: qwen3_vl
|
| 131 |
+
num_heads: 16
|
| 132 |
+
num_position_embeddings: 2304
|
| 133 |
+
out_hidden_size: 4096
|
| 134 |
+
patch_size: 16
|
| 135 |
+
spatial_merge_size: 2
|
| 136 |
+
temporal_patch_size: 2
|
| 137 |
+
vision_end_token_id: 151653
|
| 138 |
+
vision_start_token_id: 151652
|
| 139 |
+
vocab_size: 151936
|
| 140 |
+
_data_root_path:
|
| 141 |
+
- ./datasets/libero_spatial_no_noops_lerobotv2.1
|
| 142 |
+
- ./datasets/libero_spatial_no_noops_lerobotv2.1
|
| 143 |
+
- ./datasets/libero_object_no_noops_lerobotv2.1
|
| 144 |
+
- ./datasets/libero_object_no_noops_lerobotv2.1
|
| 145 |
+
- ./datasets/libero_goal_no_noops_lerobotv2.1
|
| 146 |
+
- ./datasets/libero_goal_no_noops_lerobotv2.1
|
| 147 |
+
- ./datasets/libero_10_no_noops_lerobotv2.1
|
| 148 |
+
_frame_window_size: 17
|
| 149 |
+
_grad_accumulation_steps: 8
|
| 150 |
+
_image_height: 256
|
| 151 |
+
_image_width: 256
|
| 152 |
+
_libero_action_stats:
|
| 153 |
+
max:
|
| 154 |
+
- 0.9375
|
| 155 |
+
- 0.9375
|
| 156 |
+
- 0.9375
|
| 157 |
+
- 1.0
|
| 158 |
+
- 0.368853
|
| 159 |
+
- 0.341214
|
| 160 |
+
- 0.356395
|
| 161 |
+
- 1.0
|
| 162 |
+
- 0.348251
|
| 163 |
+
- 1.0
|
| 164 |
+
mean:
|
| 165 |
+
- 0.050704
|
| 166 |
+
- 0.097407
|
| 167 |
+
- -0.094833
|
| 168 |
+
- 0.994873
|
| 169 |
+
- -0.004579
|
| 170 |
+
- -0.004288
|
| 171 |
+
- 0.004389
|
| 172 |
+
- 0.996104
|
| 173 |
+
- 0.001109
|
| 174 |
+
- 0.476725
|
| 175 |
+
min:
|
| 176 |
+
- -0.9375
|
| 177 |
+
- -0.9375
|
| 178 |
+
- -0.9375
|
| 179 |
+
- 0.902028
|
| 180 |
+
- -0.356085
|
| 181 |
+
- -0.367416
|
| 182 |
+
- -0.370434
|
| 183 |
+
- 0.921907
|
| 184 |
+
- -0.255
|
| 185 |
+
- 0.0
|
| 186 |
+
q01:
|
| 187 |
+
- -0.723214
|
| 188 |
+
- -0.808929
|
| 189 |
+
- -0.9375
|
| 190 |
+
- 0.934955
|
| 191 |
+
- -0.223431
|
| 192 |
+
- -0.189878
|
| 193 |
+
- -0.334735
|
| 194 |
+
- 0.938516
|
| 195 |
+
- -0.107736
|
| 196 |
+
- 0.0
|
| 197 |
+
q99:
|
| 198 |
+
- 0.9375
|
| 199 |
+
- 0.870536
|
| 200 |
+
- 0.9375
|
| 201 |
+
- 1.0
|
| 202 |
+
- 0.331
|
| 203 |
+
- 0.163153
|
| 204 |
+
- 0.226216
|
| 205 |
+
- 1.0
|
| 206 |
+
- 0.127158
|
| 207 |
+
- 1.0
|
| 208 |
+
std:
|
| 209 |
+
- 0.333621
|
| 210 |
+
- 0.387175
|
| 211 |
+
- 0.45714
|
| 212 |
+
- 0.010807
|
| 213 |
+
- 0.077802
|
| 214 |
+
- 0.063386
|
| 215 |
+
- 0.078571
|
| 216 |
+
- 0.009994
|
| 217 |
+
- 0.038504
|
| 218 |
+
- 0.49946
|
| 219 |
+
_max_action_dim: 64
|
| 220 |
+
_max_state_dim: 64
|
| 221 |
+
_max_steps: 5000
|
| 222 |
+
_per_device_batch_size: 16
|
| 223 |
+
_prepend_state_to_action: false
|
| 224 |
+
_save_iter_interval: 500
|
| 225 |
+
_statistic_name: all_libero_no_noops
|
| 226 |
+
_transforms:
|
| 227 |
+
- embodiment_id: 5
|
| 228 |
+
name_mappings:
|
| 229 |
+
actions:
|
| 230 |
+
- actions
|
| 231 |
+
observation.state:
|
| 232 |
+
- states
|
| 233 |
+
parquet_keys:
|
| 234 |
+
- observation.state
|
| 235 |
+
- timestamp
|
| 236 |
+
- actions
|
| 237 |
+
- info
|
| 238 |
+
- stats
|
| 239 |
+
- action_masks
|
| 240 |
+
type: ProcessParquetInputs
|
| 241 |
+
video_keys:
|
| 242 |
+
- observation.images.image
|
| 243 |
+
- observation.images.wrist_image
|
| 244 |
+
- type: LiberoFramewiseActionToRot6D
|
| 245 |
+
- action_metadata:
|
| 246 |
+
append_viewpoint: false
|
| 247 |
+
conditioning_fps: 20.0
|
| 248 |
+
frame_window_size: 17
|
| 249 |
+
video_height: 192
|
| 250 |
+
video_width: 320
|
| 251 |
+
cfg_dropout_rate: 0.1
|
| 252 |
+
format_prompt_as_json: true
|
| 253 |
+
max_len: 512
|
| 254 |
+
tokenizer:
|
| 255 |
+
model_max_length: 4096
|
| 256 |
+
model_path: ./checkpoints/Cosmos3-Nano/text_tokenizer
|
| 257 |
+
padding_side: right
|
| 258 |
+
trust_remote_code: true
|
| 259 |
+
type: PretrainedTokenizer
|
| 260 |
+
type: ProcessCosmos3Prompt
|
| 261 |
+
- height: 256
|
| 262 |
+
type: ResizeImages
|
| 263 |
+
width: 256
|
| 264 |
+
- type: SimpleNormalizeImages
|
| 265 |
+
- action_dim: 64
|
| 266 |
+
action_key: action
|
| 267 |
+
action_norm_type: quantile
|
| 268 |
+
state_dim: 64
|
| 269 |
+
state_key: proprio
|
| 270 |
+
state_norm_type: none
|
| 271 |
+
type: NormalizeStatesAndActions
|
| 272 |
+
- conditioning_fps: 20.0
|
| 273 |
+
frame_window_size: 17
|
| 274 |
+
mode: wam
|
| 275 |
+
prepend_state_to_action: false
|
| 276 |
+
raw_action_dim: 10
|
| 277 |
+
type: BuildCosmos3Sequence
|
| 278 |
+
- frame_window_size: 17
|
| 279 |
+
num_views: 2
|
| 280 |
+
tile_direction: horizontal
|
| 281 |
+
type: PrepareVideo
|
| 282 |
+
- height: 192
|
| 283 |
+
type: ResizeAndReflectPad
|
| 284 |
+
width: 320
|
| 285 |
+
_video_height: 192
|
| 286 |
+
_video_width: 320
|
| 287 |
+
_vision_vae:
|
| 288 |
+
encode_exact_durations:
|
| 289 |
+
- 17
|
| 290 |
+
pretrained_name_or_path: ./checkpoints/Wan2.2-TI2V-5B/Wan2.2_VAE.pth
|
| 291 |
+
type: Cosmos3Wan22VAE
|
| 292 |
+
_wan22_vae_path: ./checkpoints/Wan2.2-TI2V-5B/Wan2.2_VAE.pth
|
| 293 |
+
eval:
|
| 294 |
+
dataset:
|
| 295 |
+
extra_tensor_keys:
|
| 296 |
+
- conditioning_fps
|
| 297 |
+
- prepend_state_to_action
|
| 298 |
+
img_buffer_len: 1
|
| 299 |
+
transforms:
|
| 300 |
+
- embodiment_id: 5
|
| 301 |
+
img_keys:
|
| 302 |
+
- agentview_image
|
| 303 |
+
- robot0_eye_in_hand_image
|
| 304 |
+
resize_size: 256
|
| 305 |
+
type: ProcessLiberoEvalInputs
|
| 306 |
+
- conditioning_fps: 20.0
|
| 307 |
+
prepend_state_to_action: false
|
| 308 |
+
type: SetCosmos3ActionMetadata
|
| 309 |
+
- image_resize_strategy: resize-crop
|
| 310 |
+
input_sizes:
|
| 311 |
+
- - 3
|
| 312 |
+
- 256
|
| 313 |
+
- 256
|
| 314 |
+
- - 3
|
| 315 |
+
- 256
|
| 316 |
+
- 256
|
| 317 |
+
means:
|
| 318 |
+
- - 127.5
|
| 319 |
+
- 127.5
|
| 320 |
+
- 127.5
|
| 321 |
+
- - 127.5
|
| 322 |
+
- 127.5
|
| 323 |
+
- 127.5
|
| 324 |
+
stds:
|
| 325 |
+
- - 127.5
|
| 326 |
+
- 127.5
|
| 327 |
+
- 127.5
|
| 328 |
+
- - 127.5
|
| 329 |
+
- 127.5
|
| 330 |
+
- 127.5
|
| 331 |
+
type: TransformImage
|
| 332 |
+
- action_metadata:
|
| 333 |
+
append_viewpoint: false
|
| 334 |
+
conditioning_fps: 20.0
|
| 335 |
+
frame_window_size: 17
|
| 336 |
+
video_height: 192
|
| 337 |
+
video_width: 320
|
| 338 |
+
cfg_dropout_rate: 0.0
|
| 339 |
+
format_prompt_as_json: true
|
| 340 |
+
max_len: 512
|
| 341 |
+
output_attention_mask_key: lang_masks
|
| 342 |
+
output_key: lang_tokens
|
| 343 |
+
tokenizer:
|
| 344 |
+
model_max_length: 4096
|
| 345 |
+
model_path: ./checkpoints/Cosmos3-Nano/text_tokenizer
|
| 346 |
+
padding_side: right
|
| 347 |
+
trust_remote_code: true
|
| 348 |
+
type: PretrainedTokenizer
|
| 349 |
+
type: ProcessCosmos3Prompt
|
| 350 |
+
- frame_window_size: 1
|
| 351 |
+
num_views: 2
|
| 352 |
+
tile_direction: horizontal
|
| 353 |
+
type: PrepareVideo
|
| 354 |
+
- height: 192
|
| 355 |
+
type: ResizeAndReflectPad
|
| 356 |
+
width: 320
|
| 357 |
+
type: LiberoParquetEvalDataset
|
| 358 |
+
denormalize_action:
|
| 359 |
+
action_dim: 10
|
| 360 |
+
norm_type: quantile_rot
|
| 361 |
+
type: DenormalizeLiberoFramewiseRot6DAction
|
| 362 |
+
enable_mixed_precision_training: true
|
| 363 |
+
eval_chunk_size: 16
|
| 364 |
+
mixed_precision_dtype: bf16
|
| 365 |
+
model_family: cosmos3
|
| 366 |
+
norm_stats_key: all_libero_no_noops
|
| 367 |
+
num_steps_wait: 10
|
| 368 |
+
num_trials_per_task: 50
|
| 369 |
+
seed: 7
|
| 370 |
+
task_suite_name: libero_10
|
| 371 |
+
type: LiberoEvalRunner
|
| 372 |
+
inference_model:
|
| 373 |
+
action_horizon: 16
|
| 374 |
+
action_in_proj:
|
| 375 |
+
input_size: 64
|
| 376 |
+
num_domains: 32
|
| 377 |
+
output_size: 4096
|
| 378 |
+
type: DomainAwareLinear
|
| 379 |
+
action_out_proj:
|
| 380 |
+
input_size: 4096
|
| 381 |
+
num_domains: 32
|
| 382 |
+
output_size: 64
|
| 383 |
+
type: DomainAwareLinear
|
| 384 |
+
base_fps: 24.0
|
| 385 |
+
enable_fps_modulation: true
|
| 386 |
+
enable_vision_loss: true
|
| 387 |
+
freeze_non_moe_vlm_backbone: true
|
| 388 |
+
freeze_vlm_backbone: false
|
| 389 |
+
latent_patch_size: 2
|
| 390 |
+
max_action_dim: 64
|
| 391 |
+
name_mapping:
|
| 392 |
+
.self_attn.k_norm.: .self_attn.norm_k.
|
| 393 |
+
.self_attn.k_norm_moe_gen.: .self_attn.norm_added_k.
|
| 394 |
+
.self_attn.k_proj.: .self_attn.to_k.
|
| 395 |
+
.self_attn.k_proj_moe_gen.: .self_attn.add_k_proj.
|
| 396 |
+
.self_attn.o_proj.: .self_attn.to_out.
|
| 397 |
+
.self_attn.o_proj_moe_gen.: .self_attn.to_add_out.
|
| 398 |
+
.self_attn.q_norm.: .self_attn.norm_q.
|
| 399 |
+
.self_attn.q_norm_moe_gen.: .self_attn.norm_added_q.
|
| 400 |
+
.self_attn.q_proj.: .self_attn.to_q.
|
| 401 |
+
.self_attn.q_proj_moe_gen.: .self_attn.add_q_proj.
|
| 402 |
+
.self_attn.v_proj.: .self_attn.to_v.
|
| 403 |
+
.self_attn.v_proj_moe_gen.: .self_attn.add_v_proj.
|
| 404 |
+
action_in_proj.: action_proj_in.
|
| 405 |
+
action_modality_embed: action_modality_embed
|
| 406 |
+
action_out_proj.: action_proj_out.
|
| 407 |
+
time_embedder.mlp.0.: time_embedder.linear_1.
|
| 408 |
+
time_embedder.mlp.2.: time_embedder.linear_2.
|
| 409 |
+
vision_in_proj.projector.: proj_in.
|
| 410 |
+
vision_out_proj.projector.: proj_out.
|
| 411 |
+
vlm_backbone.lm_head.weight: lm_head.weight
|
| 412 |
+
vlm_backbone.model.language_model.embed_tokens.weight: embed_tokens.weight
|
| 413 |
+
vlm_backbone.model.language_model.layers.: layers.
|
| 414 |
+
vlm_backbone.model.language_model.norm.weight: norm.weight
|
| 415 |
+
vlm_backbone.model.language_model.norm_moe_gen.weight: norm_moe_gen.weight
|
| 416 |
+
vlm_backbone.model.visual.blocks.: blocks.
|
| 417 |
+
vlm_backbone.model.visual.deepstack_merger_list.: deepstack_merger_list.
|
| 418 |
+
vlm_backbone.model.visual.merger.: merger.
|
| 419 |
+
vlm_backbone.model.visual.patch_embed.: patch_embed.
|
| 420 |
+
vlm_backbone.model.visual.pos_embed.: pos_embed.
|
| 421 |
+
num_embodiment_domains: 32
|
| 422 |
+
ori_action_dim: 10
|
| 423 |
+
packed_attention_backend: flash2
|
| 424 |
+
position_embedding_type: unified_3d_mrope
|
| 425 |
+
pretrained_name_or_path: ./checkpoints/Cosmos3-Nano/transformer
|
| 426 |
+
rectified_flow_inference_config:
|
| 427 |
+
num_steps: 30
|
| 428 |
+
num_train_timesteps: 1000
|
| 429 |
+
scheduler_type: unipc
|
| 430 |
+
shift: 10.0
|
| 431 |
+
use_dynamic_shifting: false
|
| 432 |
+
use_karras_sigmas: false
|
| 433 |
+
rectified_flow_training_config:
|
| 434 |
+
action_loss_weight: 10.0
|
| 435 |
+
high_sigma_ratio: 0.05
|
| 436 |
+
high_sigma_timesteps_max: 1000
|
| 437 |
+
high_sigma_timesteps_min: 995
|
| 438 |
+
independent_action_schedule: false
|
| 439 |
+
normalize_loss_by_active: false
|
| 440 |
+
shift:
|
| 441 |
+
'256': 3
|
| 442 |
+
'480': 5
|
| 443 |
+
'720': 10
|
| 444 |
+
shift_action: null
|
| 445 |
+
train_time_action_distribution: logitnormal
|
| 446 |
+
train_time_image_distribution: logitnormal
|
| 447 |
+
train_time_video_distribution: waver
|
| 448 |
+
train_time_weight: uniform
|
| 449 |
+
use_discrete_rf: false
|
| 450 |
+
use_dynamic_shift: false
|
| 451 |
+
use_high_sigma_strategy: false
|
| 452 |
+
use_high_sigma_strategy_action: false
|
| 453 |
+
vision_loss_weight: 10.0
|
| 454 |
+
special_tokens:
|
| 455 |
+
end_of_generation: 151653
|
| 456 |
+
eos_token_id: 151645
|
| 457 |
+
start_of_generation: 151652
|
| 458 |
+
timestep_scale: 0.001
|
| 459 |
+
type: Cosmos3FlowMatching
|
| 460 |
+
unified_3d_mrope_reset_spatial_ids: true
|
| 461 |
+
unified_3d_mrope_temporal_modality_margin: 15000
|
| 462 |
+
vision_in_proj:
|
| 463 |
+
in_dim: 192
|
| 464 |
+
out_dim: 4096
|
| 465 |
+
type: LinearProjector
|
| 466 |
+
vision_latent_dim: 48
|
| 467 |
+
vision_out_proj:
|
| 468 |
+
in_dim: 4096
|
| 469 |
+
out_dim: 192
|
| 470 |
+
type: LinearProjector
|
| 471 |
+
vision_vae:
|
| 472 |
+
encode_exact_durations:
|
| 473 |
+
- 17
|
| 474 |
+
pretrained_name_or_path: null
|
| 475 |
+
type: Cosmos3Wan22VAE
|
| 476 |
+
vlm_backbone:
|
| 477 |
+
include_visual: false
|
| 478 |
+
skip_init_weights: true
|
| 479 |
+
type: Cosmos3MoTBackbone
|
| 480 |
+
vision_encoder_path: ./checkpoints/Cosmos3-Nano/vision_encoder
|
| 481 |
+
vlm_config:
|
| 482 |
+
image_token_id: 151655
|
| 483 |
+
model_type: qwen3_vl
|
| 484 |
+
text_config:
|
| 485 |
+
attention_bias: false
|
| 486 |
+
attention_dropout: 0.0
|
| 487 |
+
bos_token_id: 151643
|
| 488 |
+
eos_token_id: 151645
|
| 489 |
+
head_dim: 128
|
| 490 |
+
hidden_act: silu
|
| 491 |
+
hidden_size: 4096
|
| 492 |
+
intermediate_size: 12288
|
| 493 |
+
layer_types:
|
| 494 |
+
- full_attention
|
| 495 |
+
- full_attention
|
| 496 |
+
- full_attention
|
| 497 |
+
- full_attention
|
| 498 |
+
- full_attention
|
| 499 |
+
- full_attention
|
| 500 |
+
- full_attention
|
| 501 |
+
- full_attention
|
| 502 |
+
- full_attention
|
| 503 |
+
- full_attention
|
| 504 |
+
- full_attention
|
| 505 |
+
- full_attention
|
| 506 |
+
- full_attention
|
| 507 |
+
- full_attention
|
| 508 |
+
- full_attention
|
| 509 |
+
- full_attention
|
| 510 |
+
- full_attention
|
| 511 |
+
- full_attention
|
| 512 |
+
- full_attention
|
| 513 |
+
- full_attention
|
| 514 |
+
- full_attention
|
| 515 |
+
- full_attention
|
| 516 |
+
- full_attention
|
| 517 |
+
- full_attention
|
| 518 |
+
- full_attention
|
| 519 |
+
- full_attention
|
| 520 |
+
- full_attention
|
| 521 |
+
- full_attention
|
| 522 |
+
- full_attention
|
| 523 |
+
- full_attention
|
| 524 |
+
- full_attention
|
| 525 |
+
- full_attention
|
| 526 |
+
- full_attention
|
| 527 |
+
- full_attention
|
| 528 |
+
- full_attention
|
| 529 |
+
- full_attention
|
| 530 |
+
max_position_embeddings: 262144
|
| 531 |
+
model_type: qwen3_vl_text
|
| 532 |
+
num_attention_heads: 32
|
| 533 |
+
num_hidden_layers: 36
|
| 534 |
+
num_key_value_heads: 8
|
| 535 |
+
pad_token_id: 0
|
| 536 |
+
rms_norm_eps: 1.0e-06
|
| 537 |
+
rope_scaling:
|
| 538 |
+
mrope_interleaved: true
|
| 539 |
+
mrope_section:
|
| 540 |
+
- 24
|
| 541 |
+
- 20
|
| 542 |
+
- 20
|
| 543 |
+
rope_type: default
|
| 544 |
+
rope_theta: 5000000
|
| 545 |
+
tie_word_embeddings: false
|
| 546 |
+
vocab_size: 151936
|
| 547 |
+
tie_word_embeddings: false
|
| 548 |
+
video_token_id: 151656
|
| 549 |
+
vision_config:
|
| 550 |
+
deepstack_visual_indexes:
|
| 551 |
+
- 8
|
| 552 |
+
- 16
|
| 553 |
+
- 24
|
| 554 |
+
depth: 27
|
| 555 |
+
hidden_act: gelu_pytorch_tanh
|
| 556 |
+
hidden_size: 1152
|
| 557 |
+
in_channels: 3
|
| 558 |
+
initializer_range: 0.02
|
| 559 |
+
intermediate_size: 4304
|
| 560 |
+
model_type: qwen3_vl
|
| 561 |
+
num_heads: 16
|
| 562 |
+
num_position_embeddings: 2304
|
| 563 |
+
out_hidden_size: 4096
|
| 564 |
+
patch_size: 16
|
| 565 |
+
spatial_merge_size: 2
|
| 566 |
+
temporal_patch_size: 2
|
| 567 |
+
vision_end_token_id: 151653
|
| 568 |
+
vision_start_token_id: 151652
|
| 569 |
+
vocab_size: 151936
|
| 570 |
+
model:
|
| 571 |
+
action_horizon: 16
|
| 572 |
+
action_in_proj:
|
| 573 |
+
input_size: 64
|
| 574 |
+
num_domains: 32
|
| 575 |
+
output_size: 4096
|
| 576 |
+
type: DomainAwareLinear
|
| 577 |
+
action_out_proj:
|
| 578 |
+
input_size: 4096
|
| 579 |
+
num_domains: 32
|
| 580 |
+
output_size: 64
|
| 581 |
+
type: DomainAwareLinear
|
| 582 |
+
base_fps: 24.0
|
| 583 |
+
enable_fps_modulation: true
|
| 584 |
+
enable_vision_loss: true
|
| 585 |
+
freeze_non_moe_vlm_backbone: true
|
| 586 |
+
freeze_vlm_backbone: false
|
| 587 |
+
latent_patch_size: 2
|
| 588 |
+
max_action_dim: 64
|
| 589 |
+
name_mapping:
|
| 590 |
+
.self_attn.k_norm.: .self_attn.norm_k.
|
| 591 |
+
.self_attn.k_norm_moe_gen.: .self_attn.norm_added_k.
|
| 592 |
+
.self_attn.k_proj.: .self_attn.to_k.
|
| 593 |
+
.self_attn.k_proj_moe_gen.: .self_attn.add_k_proj.
|
| 594 |
+
.self_attn.o_proj.: .self_attn.to_out.
|
| 595 |
+
.self_attn.o_proj_moe_gen.: .self_attn.to_add_out.
|
| 596 |
+
.self_attn.q_norm.: .self_attn.norm_q.
|
| 597 |
+
.self_attn.q_norm_moe_gen.: .self_attn.norm_added_q.
|
| 598 |
+
.self_attn.q_proj.: .self_attn.to_q.
|
| 599 |
+
.self_attn.q_proj_moe_gen.: .self_attn.add_q_proj.
|
| 600 |
+
.self_attn.v_proj.: .self_attn.to_v.
|
| 601 |
+
.self_attn.v_proj_moe_gen.: .self_attn.add_v_proj.
|
| 602 |
+
action_in_proj.: action_proj_in.
|
| 603 |
+
action_modality_embed: action_modality_embed
|
| 604 |
+
action_out_proj.: action_proj_out.
|
| 605 |
+
time_embedder.mlp.0.: time_embedder.linear_1.
|
| 606 |
+
time_embedder.mlp.2.: time_embedder.linear_2.
|
| 607 |
+
vision_in_proj.projector.: proj_in.
|
| 608 |
+
vision_out_proj.projector.: proj_out.
|
| 609 |
+
vlm_backbone.lm_head.weight: lm_head.weight
|
| 610 |
+
vlm_backbone.model.language_model.embed_tokens.weight: embed_tokens.weight
|
| 611 |
+
vlm_backbone.model.language_model.layers.: layers.
|
| 612 |
+
vlm_backbone.model.language_model.norm.weight: norm.weight
|
| 613 |
+
vlm_backbone.model.language_model.norm_moe_gen.weight: norm_moe_gen.weight
|
| 614 |
+
vlm_backbone.model.visual.blocks.: blocks.
|
| 615 |
+
vlm_backbone.model.visual.deepstack_merger_list.: deepstack_merger_list.
|
| 616 |
+
vlm_backbone.model.visual.merger.: merger.
|
| 617 |
+
vlm_backbone.model.visual.patch_embed.: patch_embed.
|
| 618 |
+
vlm_backbone.model.visual.pos_embed.: pos_embed.
|
| 619 |
+
num_embodiment_domains: 32
|
| 620 |
+
ori_action_dim: 10
|
| 621 |
+
packed_attention_backend: flash2
|
| 622 |
+
position_embedding_type: unified_3d_mrope
|
| 623 |
+
pretrained_name_or_path: ./checkpoints/Cosmos3-Nano/transformer
|
| 624 |
+
rectified_flow_inference_config:
|
| 625 |
+
num_steps: 30
|
| 626 |
+
num_train_timesteps: 1000
|
| 627 |
+
scheduler_type: unipc
|
| 628 |
+
shift: 10.0
|
| 629 |
+
use_dynamic_shifting: false
|
| 630 |
+
use_karras_sigmas: false
|
| 631 |
+
rectified_flow_training_config:
|
| 632 |
+
action_loss_weight: 10.0
|
| 633 |
+
high_sigma_ratio: 0.05
|
| 634 |
+
high_sigma_timesteps_max: 1000
|
| 635 |
+
high_sigma_timesteps_min: 995
|
| 636 |
+
independent_action_schedule: false
|
| 637 |
+
normalize_loss_by_active: false
|
| 638 |
+
shift:
|
| 639 |
+
'256': 3
|
| 640 |
+
'480': 5
|
| 641 |
+
'720': 10
|
| 642 |
+
shift_action: null
|
| 643 |
+
train_time_action_distribution: logitnormal
|
| 644 |
+
train_time_image_distribution: logitnormal
|
| 645 |
+
train_time_video_distribution: waver
|
| 646 |
+
train_time_weight: uniform
|
| 647 |
+
use_discrete_rf: false
|
| 648 |
+
use_dynamic_shift: false
|
| 649 |
+
use_high_sigma_strategy: false
|
| 650 |
+
use_high_sigma_strategy_action: false
|
| 651 |
+
vision_loss_weight: 10.0
|
| 652 |
+
special_tokens:
|
| 653 |
+
end_of_generation: 151653
|
| 654 |
+
eos_token_id: 151645
|
| 655 |
+
start_of_generation: 151652
|
| 656 |
+
timestep_scale: 0.001
|
| 657 |
+
type: Cosmos3FlowMatching
|
| 658 |
+
unified_3d_mrope_reset_spatial_ids: true
|
| 659 |
+
unified_3d_mrope_temporal_modality_margin: 15000
|
| 660 |
+
vision_in_proj:
|
| 661 |
+
in_dim: 192
|
| 662 |
+
out_dim: 4096
|
| 663 |
+
type: LinearProjector
|
| 664 |
+
vision_latent_dim: 48
|
| 665 |
+
vision_out_proj:
|
| 666 |
+
in_dim: 4096
|
| 667 |
+
out_dim: 192
|
| 668 |
+
type: LinearProjector
|
| 669 |
+
vision_vae:
|
| 670 |
+
encode_exact_durations:
|
| 671 |
+
- 17
|
| 672 |
+
pretrained_name_or_path: ./checkpoints/Wan2.2-TI2V-5B/Wan2.2_VAE.pth
|
| 673 |
+
type: Cosmos3Wan22VAE
|
| 674 |
+
vlm_backbone:
|
| 675 |
+
include_visual: false
|
| 676 |
+
skip_init_weights: true
|
| 677 |
+
type: Cosmos3MoTBackbone
|
| 678 |
+
vision_encoder_path: ./checkpoints/Cosmos3-Nano/vision_encoder
|
| 679 |
+
vlm_config:
|
| 680 |
+
image_token_id: 151655
|
| 681 |
+
model_type: qwen3_vl
|
| 682 |
+
text_config:
|
| 683 |
+
attention_bias: false
|
| 684 |
+
attention_dropout: 0.0
|
| 685 |
+
bos_token_id: 151643
|
| 686 |
+
eos_token_id: 151645
|
| 687 |
+
head_dim: 128
|
| 688 |
+
hidden_act: silu
|
| 689 |
+
hidden_size: 4096
|
| 690 |
+
intermediate_size: 12288
|
| 691 |
+
layer_types:
|
| 692 |
+
- full_attention
|
| 693 |
+
- full_attention
|
| 694 |
+
- full_attention
|
| 695 |
+
- full_attention
|
| 696 |
+
- full_attention
|
| 697 |
+
- full_attention
|
| 698 |
+
- full_attention
|
| 699 |
+
- full_attention
|
| 700 |
+
- full_attention
|
| 701 |
+
- full_attention
|
| 702 |
+
- full_attention
|
| 703 |
+
- full_attention
|
| 704 |
+
- full_attention
|
| 705 |
+
- full_attention
|
| 706 |
+
- full_attention
|
| 707 |
+
- full_attention
|
| 708 |
+
- full_attention
|
| 709 |
+
- full_attention
|
| 710 |
+
- full_attention
|
| 711 |
+
- full_attention
|
| 712 |
+
- full_attention
|
| 713 |
+
- full_attention
|
| 714 |
+
- full_attention
|
| 715 |
+
- full_attention
|
| 716 |
+
- full_attention
|
| 717 |
+
- full_attention
|
| 718 |
+
- full_attention
|
| 719 |
+
- full_attention
|
| 720 |
+
- full_attention
|
| 721 |
+
- full_attention
|
| 722 |
+
- full_attention
|
| 723 |
+
- full_attention
|
| 724 |
+
- full_attention
|
| 725 |
+
- full_attention
|
| 726 |
+
- full_attention
|
| 727 |
+
- full_attention
|
| 728 |
+
max_position_embeddings: 262144
|
| 729 |
+
model_type: qwen3_vl_text
|
| 730 |
+
num_attention_heads: 32
|
| 731 |
+
num_hidden_layers: 36
|
| 732 |
+
num_key_value_heads: 8
|
| 733 |
+
pad_token_id: 0
|
| 734 |
+
rms_norm_eps: 1.0e-06
|
| 735 |
+
rope_scaling:
|
| 736 |
+
mrope_interleaved: true
|
| 737 |
+
mrope_section:
|
| 738 |
+
- 24
|
| 739 |
+
- 20
|
| 740 |
+
- 20
|
| 741 |
+
rope_type: default
|
| 742 |
+
rope_theta: 5000000
|
| 743 |
+
tie_word_embeddings: false
|
| 744 |
+
vocab_size: 151936
|
| 745 |
+
tie_word_embeddings: false
|
| 746 |
+
video_token_id: 151656
|
| 747 |
+
vision_config:
|
| 748 |
+
deepstack_visual_indexes:
|
| 749 |
+
- 8
|
| 750 |
+
- 16
|
| 751 |
+
- 24
|
| 752 |
+
depth: 27
|
| 753 |
+
hidden_act: gelu_pytorch_tanh
|
| 754 |
+
hidden_size: 1152
|
| 755 |
+
in_channels: 3
|
| 756 |
+
initializer_range: 0.02
|
| 757 |
+
intermediate_size: 4304
|
| 758 |
+
model_type: qwen3_vl
|
| 759 |
+
num_heads: 16
|
| 760 |
+
num_position_embeddings: 2304
|
| 761 |
+
out_hidden_size: 4096
|
| 762 |
+
patch_size: 16
|
| 763 |
+
spatial_merge_size: 2
|
| 764 |
+
temporal_patch_size: 2
|
| 765 |
+
vision_end_token_id: 151653
|
| 766 |
+
vision_start_token_id: 151652
|
| 767 |
+
vocab_size: 151936
|
| 768 |
+
runner:
|
| 769 |
+
change_key_name: false
|
| 770 |
+
collator:
|
| 771 |
+
list_keys:
|
| 772 |
+
- sequence_plan
|
| 773 |
+
meta_keys:
|
| 774 |
+
- task_description
|
| 775 |
+
- stats
|
| 776 |
+
- info
|
| 777 |
+
- timestamp
|
| 778 |
+
pad_id: 0
|
| 779 |
+
sequence_keys:
|
| 780 |
+
- text_token_ids
|
| 781 |
+
tensor_keys:
|
| 782 |
+
- images
|
| 783 |
+
- actions
|
| 784 |
+
- embodiment_ids
|
| 785 |
+
- raw_action_dim
|
| 786 |
+
- conditioning_fps
|
| 787 |
+
- action_fps
|
| 788 |
+
type: Cosmos3Collator
|
| 789 |
+
enable_gradient_checkpointing: true
|
| 790 |
+
enable_mixed_precision_training: true
|
| 791 |
+
grad_accumulation_steps: 8
|
| 792 |
+
lr_scheduler:
|
| 793 |
+
cycle_length: 16000
|
| 794 |
+
type: linear-warmup+linear-decay
|
| 795 |
+
warmup_steps: 500
|
| 796 |
+
max_grad_norm: 1.0
|
| 797 |
+
max_keep_ckpts: 2
|
| 798 |
+
max_steps: 5000
|
| 799 |
+
metric:
|
| 800 |
+
active_trackers:
|
| 801 |
+
- jsonl
|
| 802 |
+
- wandb
|
| 803 |
+
grad_accumulation_steps: 8
|
| 804 |
+
run_dir: work_dirs
|
| 805 |
+
type: VLAMetric
|
| 806 |
+
window_size: 1
|
| 807 |
+
mixed_precision_dtype: bf16
|
| 808 |
+
optimizer:
|
| 809 |
+
betas:
|
| 810 |
+
- 0.9
|
| 811 |
+
- 0.99
|
| 812 |
+
eps: 1.0e-08
|
| 813 |
+
exclude_1d_from_weight_decay: false
|
| 814 |
+
fused: true
|
| 815 |
+
lr: 5.0e-05
|
| 816 |
+
paramwise_learning_rate:
|
| 817 |
+
action_in_proj.: 0.00025
|
| 818 |
+
action_modality_embed: 0.00025
|
| 819 |
+
action_out_proj.: 0.00025
|
| 820 |
+
type: AdamW
|
| 821 |
+
weight_decay: 0.05
|
| 822 |
+
sampler: null
|
| 823 |
+
save_iter_interval: 500
|
| 824 |
+
sharding_strategy: full-shard
|
| 825 |
+
tokenizer:
|
| 826 |
+
model_max_length: 4096
|
| 827 |
+
model_path: ./checkpoints/Cosmos3-Nano/text_tokenizer
|
| 828 |
+
padding_side: right
|
| 829 |
+
trust_remote_code: true
|
| 830 |
+
type: PretrainedTokenizer
|
| 831 |
+
type: FSDPTrainRunner
|
| 832 |
+
train_dataloader:
|
| 833 |
+
dataset:
|
| 834 |
+
datasets:
|
| 835 |
+
- action_key: action
|
| 836 |
+
action_window_size: 16
|
| 837 |
+
data_root_path: ./datasets/libero_spatial_no_noops_lerobotv2.1
|
| 838 |
+
frame_window_size: 17
|
| 839 |
+
require_full_window: true
|
| 840 |
+
statistic_name: all_libero_no_noops
|
| 841 |
+
transforms:
|
| 842 |
+
- embodiment_id: 5
|
| 843 |
+
name_mappings:
|
| 844 |
+
actions:
|
| 845 |
+
- actions
|
| 846 |
+
observation.state:
|
| 847 |
+
- states
|
| 848 |
+
parquet_keys:
|
| 849 |
+
- observation.state
|
| 850 |
+
- timestamp
|
| 851 |
+
- actions
|
| 852 |
+
- info
|
| 853 |
+
- stats
|
| 854 |
+
- action_masks
|
| 855 |
+
type: ProcessParquetInputs
|
| 856 |
+
video_keys:
|
| 857 |
+
- observation.images.image
|
| 858 |
+
- observation.images.wrist_image
|
| 859 |
+
- type: LiberoFramewiseActionToRot6D
|
| 860 |
+
- action_metadata:
|
| 861 |
+
append_viewpoint: false
|
| 862 |
+
conditioning_fps: 20.0
|
| 863 |
+
frame_window_size: 17
|
| 864 |
+
video_height: 192
|
| 865 |
+
video_width: 320
|
| 866 |
+
cfg_dropout_rate: 0.1
|
| 867 |
+
format_prompt_as_json: true
|
| 868 |
+
max_len: 512
|
| 869 |
+
tokenizer:
|
| 870 |
+
model_max_length: 4096
|
| 871 |
+
model_path: ./checkpoints/Cosmos3-Nano/text_tokenizer
|
| 872 |
+
padding_side: right
|
| 873 |
+
trust_remote_code: true
|
| 874 |
+
type: PretrainedTokenizer
|
| 875 |
+
type: ProcessCosmos3Prompt
|
| 876 |
+
- height: 256
|
| 877 |
+
type: ResizeImages
|
| 878 |
+
width: 256
|
| 879 |
+
- type: SimpleNormalizeImages
|
| 880 |
+
- action_dim: 64
|
| 881 |
+
action_key: action
|
| 882 |
+
action_norm_type: quantile
|
| 883 |
+
state_dim: 64
|
| 884 |
+
state_key: proprio
|
| 885 |
+
state_norm_type: none
|
| 886 |
+
type: NormalizeStatesAndActions
|
| 887 |
+
- conditioning_fps: 20.0
|
| 888 |
+
frame_window_size: 17
|
| 889 |
+
mode: wam
|
| 890 |
+
prepend_state_to_action: false
|
| 891 |
+
raw_action_dim: 10
|
| 892 |
+
type: BuildCosmos3Sequence
|
| 893 |
+
- frame_window_size: 17
|
| 894 |
+
num_views: 2
|
| 895 |
+
tile_direction: horizontal
|
| 896 |
+
type: PrepareVideo
|
| 897 |
+
- height: 192
|
| 898 |
+
type: ResizeAndReflectPad
|
| 899 |
+
width: 320
|
| 900 |
+
type: ParquetDataset
|
| 901 |
+
use_delta: false
|
| 902 |
+
window_start_idx: 0
|
| 903 |
+
- action_key: action
|
| 904 |
+
action_window_size: 16
|
| 905 |
+
data_root_path: ./datasets/libero_spatial_no_noops_lerobotv2.1
|
| 906 |
+
frame_window_size: 17
|
| 907 |
+
require_full_window: true
|
| 908 |
+
statistic_name: all_libero_no_noops
|
| 909 |
+
transforms:
|
| 910 |
+
- embodiment_id: 5
|
| 911 |
+
name_mappings:
|
| 912 |
+
actions:
|
| 913 |
+
- actions
|
| 914 |
+
observation.state:
|
| 915 |
+
- states
|
| 916 |
+
parquet_keys:
|
| 917 |
+
- observation.state
|
| 918 |
+
- timestamp
|
| 919 |
+
- actions
|
| 920 |
+
- info
|
| 921 |
+
- stats
|
| 922 |
+
- action_masks
|
| 923 |
+
type: ProcessParquetInputs
|
| 924 |
+
video_keys:
|
| 925 |
+
- observation.images.image
|
| 926 |
+
- observation.images.wrist_image
|
| 927 |
+
- type: LiberoFramewiseActionToRot6D
|
| 928 |
+
- action_metadata:
|
| 929 |
+
append_viewpoint: false
|
| 930 |
+
conditioning_fps: 20.0
|
| 931 |
+
frame_window_size: 17
|
| 932 |
+
video_height: 192
|
| 933 |
+
video_width: 320
|
| 934 |
+
cfg_dropout_rate: 0.1
|
| 935 |
+
format_prompt_as_json: true
|
| 936 |
+
max_len: 512
|
| 937 |
+
tokenizer:
|
| 938 |
+
model_max_length: 4096
|
| 939 |
+
model_path: ./checkpoints/Cosmos3-Nano/text_tokenizer
|
| 940 |
+
padding_side: right
|
| 941 |
+
trust_remote_code: true
|
| 942 |
+
type: PretrainedTokenizer
|
| 943 |
+
type: ProcessCosmos3Prompt
|
| 944 |
+
- height: 256
|
| 945 |
+
type: ResizeImages
|
| 946 |
+
width: 256
|
| 947 |
+
- type: SimpleNormalizeImages
|
| 948 |
+
- action_dim: 64
|
| 949 |
+
action_key: action
|
| 950 |
+
action_norm_type: quantile
|
| 951 |
+
state_dim: 64
|
| 952 |
+
state_key: proprio
|
| 953 |
+
state_norm_type: none
|
| 954 |
+
type: NormalizeStatesAndActions
|
| 955 |
+
- conditioning_fps: 20.0
|
| 956 |
+
frame_window_size: 17
|
| 957 |
+
mode: wam
|
| 958 |
+
prepend_state_to_action: false
|
| 959 |
+
raw_action_dim: 10
|
| 960 |
+
type: BuildCosmos3Sequence
|
| 961 |
+
- frame_window_size: 17
|
| 962 |
+
num_views: 2
|
| 963 |
+
tile_direction: horizontal
|
| 964 |
+
type: PrepareVideo
|
| 965 |
+
- height: 192
|
| 966 |
+
type: ResizeAndReflectPad
|
| 967 |
+
width: 320
|
| 968 |
+
type: ParquetDataset
|
| 969 |
+
use_delta: false
|
| 970 |
+
window_start_idx: 0
|
| 971 |
+
- action_key: action
|
| 972 |
+
action_window_size: 16
|
| 973 |
+
data_root_path: ./datasets/libero_object_no_noops_lerobotv2.1
|
| 974 |
+
frame_window_size: 17
|
| 975 |
+
require_full_window: true
|
| 976 |
+
statistic_name: all_libero_no_noops
|
| 977 |
+
transforms:
|
| 978 |
+
- embodiment_id: 5
|
| 979 |
+
name_mappings:
|
| 980 |
+
actions:
|
| 981 |
+
- actions
|
| 982 |
+
observation.state:
|
| 983 |
+
- states
|
| 984 |
+
parquet_keys:
|
| 985 |
+
- observation.state
|
| 986 |
+
- timestamp
|
| 987 |
+
- actions
|
| 988 |
+
- info
|
| 989 |
+
- stats
|
| 990 |
+
- action_masks
|
| 991 |
+
type: ProcessParquetInputs
|
| 992 |
+
video_keys:
|
| 993 |
+
- observation.images.image
|
| 994 |
+
- observation.images.wrist_image
|
| 995 |
+
- type: LiberoFramewiseActionToRot6D
|
| 996 |
+
- action_metadata:
|
| 997 |
+
append_viewpoint: false
|
| 998 |
+
conditioning_fps: 20.0
|
| 999 |
+
frame_window_size: 17
|
| 1000 |
+
video_height: 192
|
| 1001 |
+
video_width: 320
|
| 1002 |
+
cfg_dropout_rate: 0.1
|
| 1003 |
+
format_prompt_as_json: true
|
| 1004 |
+
max_len: 512
|
| 1005 |
+
tokenizer:
|
| 1006 |
+
model_max_length: 4096
|
| 1007 |
+
model_path: ./checkpoints/Cosmos3-Nano/text_tokenizer
|
| 1008 |
+
padding_side: right
|
| 1009 |
+
trust_remote_code: true
|
| 1010 |
+
type: PretrainedTokenizer
|
| 1011 |
+
type: ProcessCosmos3Prompt
|
| 1012 |
+
- height: 256
|
| 1013 |
+
type: ResizeImages
|
| 1014 |
+
width: 256
|
| 1015 |
+
- type: SimpleNormalizeImages
|
| 1016 |
+
- action_dim: 64
|
| 1017 |
+
action_key: action
|
| 1018 |
+
action_norm_type: quantile
|
| 1019 |
+
state_dim: 64
|
| 1020 |
+
state_key: proprio
|
| 1021 |
+
state_norm_type: none
|
| 1022 |
+
type: NormalizeStatesAndActions
|
| 1023 |
+
- conditioning_fps: 20.0
|
| 1024 |
+
frame_window_size: 17
|
| 1025 |
+
mode: wam
|
| 1026 |
+
prepend_state_to_action: false
|
| 1027 |
+
raw_action_dim: 10
|
| 1028 |
+
type: BuildCosmos3Sequence
|
| 1029 |
+
- frame_window_size: 17
|
| 1030 |
+
num_views: 2
|
| 1031 |
+
tile_direction: horizontal
|
| 1032 |
+
type: PrepareVideo
|
| 1033 |
+
- height: 192
|
| 1034 |
+
type: ResizeAndReflectPad
|
| 1035 |
+
width: 320
|
| 1036 |
+
type: ParquetDataset
|
| 1037 |
+
use_delta: false
|
| 1038 |
+
window_start_idx: 0
|
| 1039 |
+
- action_key: action
|
| 1040 |
+
action_window_size: 16
|
| 1041 |
+
data_root_path: ./datasets/libero_object_no_noops_lerobotv2.1
|
| 1042 |
+
frame_window_size: 17
|
| 1043 |
+
require_full_window: true
|
| 1044 |
+
statistic_name: all_libero_no_noops
|
| 1045 |
+
transforms:
|
| 1046 |
+
- embodiment_id: 5
|
| 1047 |
+
name_mappings:
|
| 1048 |
+
actions:
|
| 1049 |
+
- actions
|
| 1050 |
+
observation.state:
|
| 1051 |
+
- states
|
| 1052 |
+
parquet_keys:
|
| 1053 |
+
- observation.state
|
| 1054 |
+
- timestamp
|
| 1055 |
+
- actions
|
| 1056 |
+
- info
|
| 1057 |
+
- stats
|
| 1058 |
+
- action_masks
|
| 1059 |
+
type: ProcessParquetInputs
|
| 1060 |
+
video_keys:
|
| 1061 |
+
- observation.images.image
|
| 1062 |
+
- observation.images.wrist_image
|
| 1063 |
+
- type: LiberoFramewiseActionToRot6D
|
| 1064 |
+
- action_metadata:
|
| 1065 |
+
append_viewpoint: false
|
| 1066 |
+
conditioning_fps: 20.0
|
| 1067 |
+
frame_window_size: 17
|
| 1068 |
+
video_height: 192
|
| 1069 |
+
video_width: 320
|
| 1070 |
+
cfg_dropout_rate: 0.1
|
| 1071 |
+
format_prompt_as_json: true
|
| 1072 |
+
max_len: 512
|
| 1073 |
+
tokenizer:
|
| 1074 |
+
model_max_length: 4096
|
| 1075 |
+
model_path: ./checkpoints/Cosmos3-Nano/text_tokenizer
|
| 1076 |
+
padding_side: right
|
| 1077 |
+
trust_remote_code: true
|
| 1078 |
+
type: PretrainedTokenizer
|
| 1079 |
+
type: ProcessCosmos3Prompt
|
| 1080 |
+
- height: 256
|
| 1081 |
+
type: ResizeImages
|
| 1082 |
+
width: 256
|
| 1083 |
+
- type: SimpleNormalizeImages
|
| 1084 |
+
- action_dim: 64
|
| 1085 |
+
action_key: action
|
| 1086 |
+
action_norm_type: quantile
|
| 1087 |
+
state_dim: 64
|
| 1088 |
+
state_key: proprio
|
| 1089 |
+
state_norm_type: none
|
| 1090 |
+
type: NormalizeStatesAndActions
|
| 1091 |
+
- conditioning_fps: 20.0
|
| 1092 |
+
frame_window_size: 17
|
| 1093 |
+
mode: wam
|
| 1094 |
+
prepend_state_to_action: false
|
| 1095 |
+
raw_action_dim: 10
|
| 1096 |
+
type: BuildCosmos3Sequence
|
| 1097 |
+
- frame_window_size: 17
|
| 1098 |
+
num_views: 2
|
| 1099 |
+
tile_direction: horizontal
|
| 1100 |
+
type: PrepareVideo
|
| 1101 |
+
- height: 192
|
| 1102 |
+
type: ResizeAndReflectPad
|
| 1103 |
+
width: 320
|
| 1104 |
+
type: ParquetDataset
|
| 1105 |
+
use_delta: false
|
| 1106 |
+
window_start_idx: 0
|
| 1107 |
+
- action_key: action
|
| 1108 |
+
action_window_size: 16
|
| 1109 |
+
data_root_path: ./datasets/libero_goal_no_noops_lerobotv2.1
|
| 1110 |
+
frame_window_size: 17
|
| 1111 |
+
require_full_window: true
|
| 1112 |
+
statistic_name: all_libero_no_noops
|
| 1113 |
+
transforms:
|
| 1114 |
+
- embodiment_id: 5
|
| 1115 |
+
name_mappings:
|
| 1116 |
+
actions:
|
| 1117 |
+
- actions
|
| 1118 |
+
observation.state:
|
| 1119 |
+
- states
|
| 1120 |
+
parquet_keys:
|
| 1121 |
+
- observation.state
|
| 1122 |
+
- timestamp
|
| 1123 |
+
- actions
|
| 1124 |
+
- info
|
| 1125 |
+
- stats
|
| 1126 |
+
- action_masks
|
| 1127 |
+
type: ProcessParquetInputs
|
| 1128 |
+
video_keys:
|
| 1129 |
+
- observation.images.image
|
| 1130 |
+
- observation.images.wrist_image
|
| 1131 |
+
- type: LiberoFramewiseActionToRot6D
|
| 1132 |
+
- action_metadata:
|
| 1133 |
+
append_viewpoint: false
|
| 1134 |
+
conditioning_fps: 20.0
|
| 1135 |
+
frame_window_size: 17
|
| 1136 |
+
video_height: 192
|
| 1137 |
+
video_width: 320
|
| 1138 |
+
cfg_dropout_rate: 0.1
|
| 1139 |
+
format_prompt_as_json: true
|
| 1140 |
+
max_len: 512
|
| 1141 |
+
tokenizer:
|
| 1142 |
+
model_max_length: 4096
|
| 1143 |
+
model_path: ./checkpoints/Cosmos3-Nano/text_tokenizer
|
| 1144 |
+
padding_side: right
|
| 1145 |
+
trust_remote_code: true
|
| 1146 |
+
type: PretrainedTokenizer
|
| 1147 |
+
type: ProcessCosmos3Prompt
|
| 1148 |
+
- height: 256
|
| 1149 |
+
type: ResizeImages
|
| 1150 |
+
width: 256
|
| 1151 |
+
- type: SimpleNormalizeImages
|
| 1152 |
+
- action_dim: 64
|
| 1153 |
+
action_key: action
|
| 1154 |
+
action_norm_type: quantile
|
| 1155 |
+
state_dim: 64
|
| 1156 |
+
state_key: proprio
|
| 1157 |
+
state_norm_type: none
|
| 1158 |
+
type: NormalizeStatesAndActions
|
| 1159 |
+
- conditioning_fps: 20.0
|
| 1160 |
+
frame_window_size: 17
|
| 1161 |
+
mode: wam
|
| 1162 |
+
prepend_state_to_action: false
|
| 1163 |
+
raw_action_dim: 10
|
| 1164 |
+
type: BuildCosmos3Sequence
|
| 1165 |
+
- frame_window_size: 17
|
| 1166 |
+
num_views: 2
|
| 1167 |
+
tile_direction: horizontal
|
| 1168 |
+
type: PrepareVideo
|
| 1169 |
+
- height: 192
|
| 1170 |
+
type: ResizeAndReflectPad
|
| 1171 |
+
width: 320
|
| 1172 |
+
type: ParquetDataset
|
| 1173 |
+
use_delta: false
|
| 1174 |
+
window_start_idx: 0
|
| 1175 |
+
- action_key: action
|
| 1176 |
+
action_window_size: 16
|
| 1177 |
+
data_root_path: ./datasets/libero_goal_no_noops_lerobotv2.1
|
| 1178 |
+
frame_window_size: 17
|
| 1179 |
+
require_full_window: true
|
| 1180 |
+
statistic_name: all_libero_no_noops
|
| 1181 |
+
transforms:
|
| 1182 |
+
- embodiment_id: 5
|
| 1183 |
+
name_mappings:
|
| 1184 |
+
actions:
|
| 1185 |
+
- actions
|
| 1186 |
+
observation.state:
|
| 1187 |
+
- states
|
| 1188 |
+
parquet_keys:
|
| 1189 |
+
- observation.state
|
| 1190 |
+
- timestamp
|
| 1191 |
+
- actions
|
| 1192 |
+
- info
|
| 1193 |
+
- stats
|
| 1194 |
+
- action_masks
|
| 1195 |
+
type: ProcessParquetInputs
|
| 1196 |
+
video_keys:
|
| 1197 |
+
- observation.images.image
|
| 1198 |
+
- observation.images.wrist_image
|
| 1199 |
+
- type: LiberoFramewiseActionToRot6D
|
| 1200 |
+
- action_metadata:
|
| 1201 |
+
append_viewpoint: false
|
| 1202 |
+
conditioning_fps: 20.0
|
| 1203 |
+
frame_window_size: 17
|
| 1204 |
+
video_height: 192
|
| 1205 |
+
video_width: 320
|
| 1206 |
+
cfg_dropout_rate: 0.1
|
| 1207 |
+
format_prompt_as_json: true
|
| 1208 |
+
max_len: 512
|
| 1209 |
+
tokenizer:
|
| 1210 |
+
model_max_length: 4096
|
| 1211 |
+
model_path: ./checkpoints/Cosmos3-Nano/text_tokenizer
|
| 1212 |
+
padding_side: right
|
| 1213 |
+
trust_remote_code: true
|
| 1214 |
+
type: PretrainedTokenizer
|
| 1215 |
+
type: ProcessCosmos3Prompt
|
| 1216 |
+
- height: 256
|
| 1217 |
+
type: ResizeImages
|
| 1218 |
+
width: 256
|
| 1219 |
+
- type: SimpleNormalizeImages
|
| 1220 |
+
- action_dim: 64
|
| 1221 |
+
action_key: action
|
| 1222 |
+
action_norm_type: quantile
|
| 1223 |
+
state_dim: 64
|
| 1224 |
+
state_key: proprio
|
| 1225 |
+
state_norm_type: none
|
| 1226 |
+
type: NormalizeStatesAndActions
|
| 1227 |
+
- conditioning_fps: 20.0
|
| 1228 |
+
frame_window_size: 17
|
| 1229 |
+
mode: wam
|
| 1230 |
+
prepend_state_to_action: false
|
| 1231 |
+
raw_action_dim: 10
|
| 1232 |
+
type: BuildCosmos3Sequence
|
| 1233 |
+
- frame_window_size: 17
|
| 1234 |
+
num_views: 2
|
| 1235 |
+
tile_direction: horizontal
|
| 1236 |
+
type: PrepareVideo
|
| 1237 |
+
- height: 192
|
| 1238 |
+
type: ResizeAndReflectPad
|
| 1239 |
+
width: 320
|
| 1240 |
+
type: ParquetDataset
|
| 1241 |
+
use_delta: false
|
| 1242 |
+
window_start_idx: 0
|
| 1243 |
+
- action_key: action
|
| 1244 |
+
action_window_size: 16
|
| 1245 |
+
data_root_path: ./datasets/libero_10_no_noops_lerobotv2.1
|
| 1246 |
+
frame_window_size: 17
|
| 1247 |
+
require_full_window: true
|
| 1248 |
+
statistic_name: all_libero_no_noops
|
| 1249 |
+
transforms:
|
| 1250 |
+
- embodiment_id: 5
|
| 1251 |
+
name_mappings:
|
| 1252 |
+
actions:
|
| 1253 |
+
- actions
|
| 1254 |
+
observation.state:
|
| 1255 |
+
- states
|
| 1256 |
+
parquet_keys:
|
| 1257 |
+
- observation.state
|
| 1258 |
+
- timestamp
|
| 1259 |
+
- actions
|
| 1260 |
+
- info
|
| 1261 |
+
- stats
|
| 1262 |
+
- action_masks
|
| 1263 |
+
type: ProcessParquetInputs
|
| 1264 |
+
video_keys:
|
| 1265 |
+
- observation.images.image
|
| 1266 |
+
- observation.images.wrist_image
|
| 1267 |
+
- type: LiberoFramewiseActionToRot6D
|
| 1268 |
+
- action_metadata:
|
| 1269 |
+
append_viewpoint: false
|
| 1270 |
+
conditioning_fps: 20.0
|
| 1271 |
+
frame_window_size: 17
|
| 1272 |
+
video_height: 192
|
| 1273 |
+
video_width: 320
|
| 1274 |
+
cfg_dropout_rate: 0.1
|
| 1275 |
+
format_prompt_as_json: true
|
| 1276 |
+
max_len: 512
|
| 1277 |
+
tokenizer:
|
| 1278 |
+
model_max_length: 4096
|
| 1279 |
+
model_path: ./checkpoints/Cosmos3-Nano/text_tokenizer
|
| 1280 |
+
padding_side: right
|
| 1281 |
+
trust_remote_code: true
|
| 1282 |
+
type: PretrainedTokenizer
|
| 1283 |
+
type: ProcessCosmos3Prompt
|
| 1284 |
+
- height: 256
|
| 1285 |
+
type: ResizeImages
|
| 1286 |
+
width: 256
|
| 1287 |
+
- type: SimpleNormalizeImages
|
| 1288 |
+
- action_dim: 64
|
| 1289 |
+
action_key: action
|
| 1290 |
+
action_norm_type: quantile
|
| 1291 |
+
state_dim: 64
|
| 1292 |
+
state_key: proprio
|
| 1293 |
+
state_norm_type: none
|
| 1294 |
+
type: NormalizeStatesAndActions
|
| 1295 |
+
- conditioning_fps: 20.0
|
| 1296 |
+
frame_window_size: 17
|
| 1297 |
+
mode: wam
|
| 1298 |
+
prepend_state_to_action: false
|
| 1299 |
+
raw_action_dim: 10
|
| 1300 |
+
type: BuildCosmos3Sequence
|
| 1301 |
+
- frame_window_size: 17
|
| 1302 |
+
num_views: 2
|
| 1303 |
+
tile_direction: horizontal
|
| 1304 |
+
type: PrepareVideo
|
| 1305 |
+
- height: 192
|
| 1306 |
+
type: ResizeAndReflectPad
|
| 1307 |
+
width: 320
|
| 1308 |
+
type: ParquetDataset
|
| 1309 |
+
use_delta: false
|
| 1310 |
+
window_start_idx: 0
|
| 1311 |
+
name_mappings:
|
| 1312 |
+
action:
|
| 1313 |
+
- action
|
| 1314 |
+
observation.state:
|
| 1315 |
+
- proprio
|
| 1316 |
+
statistic_keys:
|
| 1317 |
+
- observation.state
|
| 1318 |
+
- timestamp
|
| 1319 |
+
- action
|
| 1320 |
+
statistic_name: all_libero_no_noops
|
| 1321 |
+
statistics_overrides:
|
| 1322 |
+
all_libero_no_noops:
|
| 1323 |
+
action:
|
| 1324 |
+
max:
|
| 1325 |
+
- 0.9375
|
| 1326 |
+
- 0.9375
|
| 1327 |
+
- 0.9375
|
| 1328 |
+
- 1.0
|
| 1329 |
+
- 0.368853
|
| 1330 |
+
- 0.341214
|
| 1331 |
+
- 0.356395
|
| 1332 |
+
- 1.0
|
| 1333 |
+
- 0.348251
|
| 1334 |
+
- 1.0
|
| 1335 |
+
mean:
|
| 1336 |
+
- 0.050704
|
| 1337 |
+
- 0.097407
|
| 1338 |
+
- -0.094833
|
| 1339 |
+
- 0.994873
|
| 1340 |
+
- -0.004579
|
| 1341 |
+
- -0.004288
|
| 1342 |
+
- 0.004389
|
| 1343 |
+
- 0.996104
|
| 1344 |
+
- 0.001109
|
| 1345 |
+
- 0.476725
|
| 1346 |
+
min:
|
| 1347 |
+
- -0.9375
|
| 1348 |
+
- -0.9375
|
| 1349 |
+
- -0.9375
|
| 1350 |
+
- 0.902028
|
| 1351 |
+
- -0.356085
|
| 1352 |
+
- -0.367416
|
| 1353 |
+
- -0.370434
|
| 1354 |
+
- 0.921907
|
| 1355 |
+
- -0.255
|
| 1356 |
+
- 0.0
|
| 1357 |
+
q01:
|
| 1358 |
+
- -0.723214
|
| 1359 |
+
- -0.808929
|
| 1360 |
+
- -0.9375
|
| 1361 |
+
- 0.934955
|
| 1362 |
+
- -0.223431
|
| 1363 |
+
- -0.189878
|
| 1364 |
+
- -0.334735
|
| 1365 |
+
- 0.938516
|
| 1366 |
+
- -0.107736
|
| 1367 |
+
- 0.0
|
| 1368 |
+
q99:
|
| 1369 |
+
- 0.9375
|
| 1370 |
+
- 0.870536
|
| 1371 |
+
- 0.9375
|
| 1372 |
+
- 1.0
|
| 1373 |
+
- 0.331
|
| 1374 |
+
- 0.163153
|
| 1375 |
+
- 0.226216
|
| 1376 |
+
- 1.0
|
| 1377 |
+
- 0.127158
|
| 1378 |
+
- 1.0
|
| 1379 |
+
std:
|
| 1380 |
+
- 0.333621
|
| 1381 |
+
- 0.387175
|
| 1382 |
+
- 0.45714
|
| 1383 |
+
- 0.010807
|
| 1384 |
+
- 0.077802
|
| 1385 |
+
- 0.063386
|
| 1386 |
+
- 0.078571
|
| 1387 |
+
- 0.009994
|
| 1388 |
+
- 0.038504
|
| 1389 |
+
- 0.49946
|
| 1390 |
+
type: DistributedRepeatingDataset
|
| 1391 |
+
per_device_batch_size: 16
|
| 1392 |
+
per_device_num_workers: 4
|
cosmos3_nano_libero_full_finetune_bs2048/cosmos3nano_libero_full_finetune_2026_08_14_18_30_47.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
cosmos3_nano_libero_full_finetune_bs2048/dataset_statistics.json
ADDED
|
@@ -0,0 +1,138 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"all_libero_no_noops": {
|
| 3 |
+
"proprio": {
|
| 4 |
+
"mean": [
|
| 5 |
+
-0.047496414237850684,
|
| 6 |
+
0.03435983471608021,
|
| 7 |
+
0.7514258672095461,
|
| 8 |
+
2.983973905120257,
|
| 9 |
+
-0.1419395165888572,
|
| 10 |
+
-0.118218935315545,
|
| 11 |
+
0.026695325676034158,
|
| 12 |
+
-0.026933706126891162
|
| 13 |
+
],
|
| 14 |
+
"std": [
|
| 15 |
+
0.10441071071075952,
|
| 16 |
+
0.15324089039592545,
|
| 17 |
+
0.39886824990682457,
|
| 18 |
+
0.34338260700499407,
|
| 19 |
+
0.8068781715283699,
|
| 20 |
+
0.31379284611306424,
|
| 21 |
+
0.014319374319652562,
|
| 22 |
+
0.014208265894281714
|
| 23 |
+
],
|
| 24 |
+
"min": [
|
| 25 |
+
-0.48278069496154785,
|
| 26 |
+
-0.3309336006641388,
|
| 27 |
+
0.008128181099891663,
|
| 28 |
+
1.002794623374939,
|
| 29 |
+
-3.6312508583068848,
|
| 30 |
+
-1.842738389968872,
|
| 31 |
+
-0.005453015677630901,
|
| 32 |
+
-0.042015016078948975
|
| 33 |
+
],
|
| 34 |
+
"max": [
|
| 35 |
+
0.2103137969970703,
|
| 36 |
+
0.3904264271259308,
|
| 37 |
+
1.472778081893921,
|
| 38 |
+
3.7248642444610596,
|
| 39 |
+
3.5618896484375,
|
| 40 |
+
1.3863215446472168,
|
| 41 |
+
0.04232141748070717,
|
| 42 |
+
0.0013126095291227102
|
| 43 |
+
],
|
| 44 |
+
"q01": null,
|
| 45 |
+
"q99": null
|
| 46 |
+
},
|
| 47 |
+
"timestamp": {
|
| 48 |
+
"mean": [
|
| 49 |
+
4.252653686389771
|
| 50 |
+
],
|
| 51 |
+
"std": [
|
| 52 |
+
3.2577109230601815
|
| 53 |
+
],
|
| 54 |
+
"min": [
|
| 55 |
+
0.0
|
| 56 |
+
],
|
| 57 |
+
"max": [
|
| 58 |
+
25.2
|
| 59 |
+
],
|
| 60 |
+
"q01": null,
|
| 61 |
+
"q99": null
|
| 62 |
+
},
|
| 63 |
+
"action": {
|
| 64 |
+
"mean": [
|
| 65 |
+
0.050704,
|
| 66 |
+
0.097407,
|
| 67 |
+
-0.094833,
|
| 68 |
+
0.994873,
|
| 69 |
+
-0.004579,
|
| 70 |
+
-0.004288,
|
| 71 |
+
0.004389,
|
| 72 |
+
0.996104,
|
| 73 |
+
0.001109,
|
| 74 |
+
0.476725
|
| 75 |
+
],
|
| 76 |
+
"std": [
|
| 77 |
+
0.333621,
|
| 78 |
+
0.387175,
|
| 79 |
+
0.45714,
|
| 80 |
+
0.010807,
|
| 81 |
+
0.077802,
|
| 82 |
+
0.063386,
|
| 83 |
+
0.078571,
|
| 84 |
+
0.009994,
|
| 85 |
+
0.038504,
|
| 86 |
+
0.49946
|
| 87 |
+
],
|
| 88 |
+
"min": [
|
| 89 |
+
-0.9375,
|
| 90 |
+
-0.9375,
|
| 91 |
+
-0.9375,
|
| 92 |
+
0.902028,
|
| 93 |
+
-0.356085,
|
| 94 |
+
-0.367416,
|
| 95 |
+
-0.370434,
|
| 96 |
+
0.921907,
|
| 97 |
+
-0.255,
|
| 98 |
+
0.0
|
| 99 |
+
],
|
| 100 |
+
"max": [
|
| 101 |
+
0.9375,
|
| 102 |
+
0.9375,
|
| 103 |
+
0.9375,
|
| 104 |
+
1.0,
|
| 105 |
+
0.368853,
|
| 106 |
+
0.341214,
|
| 107 |
+
0.356395,
|
| 108 |
+
1.0,
|
| 109 |
+
0.348251,
|
| 110 |
+
1.0
|
| 111 |
+
],
|
| 112 |
+
"q01": [
|
| 113 |
+
-0.723214,
|
| 114 |
+
-0.808929,
|
| 115 |
+
-0.9375,
|
| 116 |
+
0.934955,
|
| 117 |
+
-0.223431,
|
| 118 |
+
-0.189878,
|
| 119 |
+
-0.334735,
|
| 120 |
+
0.938516,
|
| 121 |
+
-0.107736,
|
| 122 |
+
0.0
|
| 123 |
+
],
|
| 124 |
+
"q99": [
|
| 125 |
+
0.9375,
|
| 126 |
+
0.870536,
|
| 127 |
+
0.9375,
|
| 128 |
+
1.0,
|
| 129 |
+
0.331,
|
| 130 |
+
0.163153,
|
| 131 |
+
0.226216,
|
| 132 |
+
1.0,
|
| 133 |
+
0.127158,
|
| 134 |
+
1.0
|
| 135 |
+
]
|
| 136 |
+
}
|
| 137 |
+
}
|
| 138 |
+
}
|
cosmos3_nano_libero_full_finetune_bs2048/run-metrics.jsonl
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"hparams": "{'_ckpt_root': './checkpoints', '_cosmos3_nano_ckpt': './checkpoints/Cosmos3-Nano', '_cosmos3_nano_transformer': './checkpoints/Cosmos3-Nano/transformer', '_cosmos3_nano_vision_encoder': './checkpoints/Cosmos3-Nano/vision_encoder', '_cosmos3_nano_tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, '_wan22_vae_path': './checkpoints/Wan2.2-TI2V-5B/Wan2.2_VAE.pth', '_vision_vae': {'type': 'Cosmos3Wan22VAE', 'pretrained_name_or_path': './checkpoints/Wan2.2-TI2V-5B/Wan2.2_VAE.pth', 'encode_exact_durations': [17]}, '_action_dim': 10, '_max_action_dim': 64, '_max_state_dim': 64, '_action_horizon': 16, '_frame_window_size': 17, '_prepend_state_to_action': False, '_image_height': 256, '_image_width': 256, '_video_height': 192, '_video_width': 320, '_conditioning_fps': 20.0, '_libero_action_stats': {'mean': [0.050704, 0.097407, -0.094833, 0.994873, -0.004579, -0.004288, 0.004389, 0.996104, 0.001109, 0.476725], 'std': [0.333621, 0.387175, 0.45714, 0.010807, 0.077802, 0.063386, 0.078571, 0.009994, 0.038504, 0.49946], 'min': [-0.9375, -0.9375, -0.9375, 0.902028, -0.356085, -0.367416, -0.370434, 0.921907, -0.255, 0.0], 'max': [0.9375, 0.9375, 0.9375, 1.0, 0.368853, 0.341214, 0.356395, 1.0, 0.348251, 1.0], 'q01': [-0.723214, -0.808929, -0.9375, 0.934955, -0.223431, -0.189878, -0.334735, 0.938516, -0.107736, 0.0], 'q99': [0.9375, 0.870536, 0.9375, 1.0, 0.331, 0.163153, 0.226216, 1.0, 0.127158, 1.0]}, '_cfg_dropout_rate': 0.1, '_base_lr': 5e-05, '_action_lr': 0.00025, '_per_device_batch_size': 16, '_grad_accumulation_steps': 8, '_max_steps': 5000, '_save_iter_interval': 500, '_data_root_path': ['./datasets/libero_spatial_no_noops_lerobotv2.1', './datasets/libero_spatial_no_noops_lerobotv2.1', './datasets/libero_object_no_noops_lerobotv2.1', './datasets/libero_object_no_noops_lerobotv2.1', './datasets/libero_goal_no_noops_lerobotv2.1', './datasets/libero_goal_no_noops_lerobotv2.1', './datasets/libero_10_no_noops_lerobotv2.1'], '_statistic_name': 'all_libero_no_noops', '_cosmos3_nano_special_tokens': {'eos_token_id': 151645, 'start_of_generation': 151652, 'end_of_generation': 151653}, '_cosmos3_nano_vlm_config': {'model_type': 'qwen3_vl', 'vocab_size': 151936, 'tie_word_embeddings': False, 'image_token_id': 151655, 'video_token_id': 151656, 'vision_start_token_id': 151652, 'vision_end_token_id': 151653, 'text_config': {'model_type': 'qwen3_vl_text', 'vocab_size': 151936, 'hidden_size': 4096, 'intermediate_size': 12288, 'num_hidden_layers': 36, 'num_attention_heads': 32, 'num_key_value_heads': 8, 'head_dim': 128, 'hidden_act': 'silu', 'max_position_embeddings': 262144, 'rms_norm_eps': 1e-06, 'rope_theta': 5000000, 'attention_bias': False, 'attention_dropout': 0.0, 'bos_token_id': 151643, 'eos_token_id': 151645, 'pad_token_id': 0, 'tie_word_embeddings': False, 'rope_scaling': {'rope_type': 'default', 'mrope_interleaved': True, 'mrope_section': [24, 20, 20]}, 'layer_types': ['full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention']}, 'vision_config': {'model_type': 'qwen3_vl', 'hidden_size': 1152, 'hidden_act': 'gelu_pytorch_tanh', 'intermediate_size': 4304, 'depth': 27, 'num_heads': 16, 'in_channels': 3, 'initializer_range': 0.02, 'out_hidden_size': 4096, 'patch_size': 16, 'spatial_merge_size': 2, 'temporal_patch_size': 2, 'num_position_embeddings': 2304, 'deepstack_visual_indexes': [8, 16, 24]}}, '_cosmos3_nano_name_mapping': {'action_modality_embed': 'action_modality_embed', 'vlm_backbone.model.language_model.embed_tokens.weight': 'embed_tokens.weight', 'vlm_backbone.lm_head.weight': 'lm_head.weight', 'vlm_backbone.model.language_model.norm.weight': 'norm.weight', 'vlm_backbone.model.language_model.norm_moe_gen.weight': 'norm_moe_gen.weight', 'vlm_backbone.model.language_model.layers.': 'layers.', 'vision_in_proj.projector.': 'proj_in.', 'vision_out_proj.projector.': 'proj_out.', 'time_embedder.mlp.0.': 'time_embedder.linear_1.', 'time_embedder.mlp.2.': 'time_embedder.linear_2.', 'action_in_proj.': 'action_proj_in.', 'action_out_proj.': 'action_proj_out.', '.self_attn.q_proj.': '.self_attn.to_q.', '.self_attn.k_proj.': '.self_attn.to_k.', '.self_attn.v_proj.': '.self_attn.to_v.', '.self_attn.o_proj.': '.self_attn.to_out.', '.self_attn.q_proj_moe_gen.': '.self_attn.add_q_proj.', '.self_attn.k_proj_moe_gen.': '.self_attn.add_k_proj.', '.self_attn.v_proj_moe_gen.': '.self_attn.add_v_proj.', '.self_attn.o_proj_moe_gen.': '.self_attn.to_add_out.', '.self_attn.q_norm.': '.self_attn.norm_q.', '.self_attn.k_norm.': '.self_attn.norm_k.', '.self_attn.q_norm_moe_gen.': '.self_attn.norm_added_q.', '.self_attn.k_norm_moe_gen.': '.self_attn.norm_added_k.', 'vlm_backbone.model.visual.patch_embed.': 'patch_embed.', 'vlm_backbone.model.visual.blocks.': 'blocks.', 'vlm_backbone.model.visual.pos_embed.': 'pos_embed.', 'vlm_backbone.model.visual.merger.': 'merger.', 'vlm_backbone.model.visual.deepstack_merger_list.': 'deepstack_merger_list.'}, 'model': {'type': 'Cosmos3FlowMatching', 'vlm_backbone': {'type': 'Cosmos3MoTBackbone', 'vlm_config': {'model_type': 'qwen3_vl', 'vocab_size': 151936, 'tie_word_embeddings': False, 'image_token_id': 151655, 'video_token_id': 151656, 'vision_start_token_id': 151652, 'vision_end_token_id': 151653, 'text_config': {'model_type': 'qwen3_vl_text', 'vocab_size': 151936, 'hidden_size': 4096, 'intermediate_size': 12288, 'num_hidden_layers': 36, 'num_attention_heads': 32, 'num_key_value_heads': 8, 'head_dim': 128, 'hidden_act': 'silu', 'max_position_embeddings': 262144, 'rms_norm_eps': 1e-06, 'rope_theta': 5000000, 'attention_bias': False, 'attention_dropout': 0.0, 'bos_token_id': 151643, 'eos_token_id': 151645, 'pad_token_id': 0, 'tie_word_embeddings': False, 'rope_scaling': {'rope_type': 'default', 'mrope_interleaved': True, 'mrope_section': [24, 20, 20]}, 'layer_types': ['full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention']}, 'vision_config': {'model_type': 'qwen3_vl', 'hidden_size': 1152, 'hidden_act': 'gelu_pytorch_tanh', 'intermediate_size': 4304, 'depth': 27, 'num_heads': 16, 'in_channels': 3, 'initializer_range': 0.02, 'out_hidden_size': 4096, 'patch_size': 16, 'spatial_merge_size': 2, 'temporal_patch_size': 2, 'num_position_embeddings': 2304, 'deepstack_visual_indexes': [8, 16, 24]}}, 'include_visual': False, 'vision_encoder_path': './checkpoints/Cosmos3-Nano/vision_encoder', 'skip_init_weights': True}, 'vision_latent_dim': 48, 'latent_patch_size': 2, 'max_action_dim': 64, 'num_embodiment_domains': 32, 'vision_in_proj': {'type': 'LinearProjector', 'in_dim': 192, 'out_dim': 4096}, 'vision_out_proj': {'type': 'LinearProjector', 'in_dim': 4096, 'out_dim': 192}, 'action_in_proj': {'type': 'DomainAwareLinear', 'input_size': 64, 'output_size': 4096, 'num_domains': 32}, 'action_out_proj': {'type': 'DomainAwareLinear', 'input_size': 4096, 'output_size': 64, 'num_domains': 32}, 'rectified_flow_training_config': {'shift': {'256': 3, '480': 5, '720': 10}, 'use_dynamic_shift': False, 'train_time_image_distribution': 'logitnormal', 'train_time_video_distribution': 'waver', 'train_time_action_distribution': 'logitnormal', 'train_time_weight': 'uniform', 'vision_loss_weight': 10.0, 'independent_action_schedule': False, 'shift_action': None, 'use_high_sigma_strategy': False, 'high_sigma_ratio': 0.05, 'high_sigma_timesteps_min': 995, 'high_sigma_timesteps_max': 1000, 'use_high_sigma_strategy_action': False, 'use_discrete_rf': False, 'normalize_loss_by_active': False, 'action_loss_weight': 10.0}, 'rectified_flow_inference_config': {'num_train_timesteps': 1000, 'scheduler_type': 'unipc', 'num_steps': 30, 'shift': 10.0, 'use_dynamic_shifting': False, 'use_karras_sigmas': False}, 'timestep_scale': 0.001, 'packed_attention_backend': 'flash2', 'position_embedding_type': 'unified_3d_mrope', 'unified_3d_mrope_reset_spatial_ids': True, 'unified_3d_mrope_temporal_modality_margin': 15000, 'enable_fps_modulation': True, 'base_fps': 24.0, 'special_tokens': {'eos_token_id': 151645, 'start_of_generation': 151652, 'end_of_generation': 151653}, 'pretrained_name_or_path': './checkpoints/Cosmos3-Nano/transformer', 'name_mapping': {'action_modality_embed': 'action_modality_embed', 'vlm_backbone.model.language_model.embed_tokens.weight': 'embed_tokens.weight', 'vlm_backbone.lm_head.weight': 'lm_head.weight', 'vlm_backbone.model.language_model.norm.weight': 'norm.weight', 'vlm_backbone.model.language_model.norm_moe_gen.weight': 'norm_moe_gen.weight', 'vlm_backbone.model.language_model.layers.': 'layers.', 'vision_in_proj.projector.': 'proj_in.', 'vision_out_proj.projector.': 'proj_out.', 'time_embedder.mlp.0.': 'time_embedder.linear_1.', 'time_embedder.mlp.2.': 'time_embedder.linear_2.', 'action_in_proj.': 'action_proj_in.', 'action_out_proj.': 'action_proj_out.', '.self_attn.q_proj.': '.self_attn.to_q.', '.self_attn.k_proj.': '.self_attn.to_k.', '.self_attn.v_proj.': '.self_attn.to_v.', '.self_attn.o_proj.': '.self_attn.to_out.', '.self_attn.q_proj_moe_gen.': '.self_attn.add_q_proj.', '.self_attn.k_proj_moe_gen.': '.self_attn.add_k_proj.', '.self_attn.v_proj_moe_gen.': '.self_attn.add_v_proj.', '.self_attn.o_proj_moe_gen.': '.self_attn.to_add_out.', '.self_attn.q_norm.': '.self_attn.norm_q.', '.self_attn.k_norm.': '.self_attn.norm_k.', '.self_attn.q_norm_moe_gen.': '.self_attn.norm_added_q.', '.self_attn.k_norm_moe_gen.': '.self_attn.norm_added_k.', 'vlm_backbone.model.visual.patch_embed.': 'patch_embed.', 'vlm_backbone.model.visual.blocks.': 'blocks.', 'vlm_backbone.model.visual.pos_embed.': 'pos_embed.', 'vlm_backbone.model.visual.merger.': 'merger.', 'vlm_backbone.model.visual.deepstack_merger_list.': 'deepstack_merger_list.'}, 'vision_vae': {'type': 'Cosmos3Wan22VAE', 'pretrained_name_or_path': './checkpoints/Wan2.2-TI2V-5B/Wan2.2_VAE.pth', 'encode_exact_durations': [17]}, 'ori_action_dim': 10, 'action_horizon': 16, 'freeze_vlm_backbone': False, 'freeze_non_moe_vlm_backbone': True, 'enable_vision_loss': True}, 'inference_model': {'type': 'Cosmos3FlowMatching', 'vlm_backbone': {'type': 'Cosmos3MoTBackbone', 'vlm_config': {'model_type': 'qwen3_vl', 'vocab_size': 151936, 'tie_word_embeddings': False, 'image_token_id': 151655, 'video_token_id': 151656, 'vision_start_token_id': 151652, 'vision_end_token_id': 151653, 'text_config': {'model_type': 'qwen3_vl_text', 'vocab_size': 151936, 'hidden_size': 4096, 'intermediate_size': 12288, 'num_hidden_layers': 36, 'num_attention_heads': 32, 'num_key_value_heads': 8, 'head_dim': 128, 'hidden_act': 'silu', 'max_position_embeddings': 262144, 'rms_norm_eps': 1e-06, 'rope_theta': 5000000, 'attention_bias': False, 'attention_dropout': 0.0, 'bos_token_id': 151643, 'eos_token_id': 151645, 'pad_token_id': 0, 'tie_word_embeddings': False, 'rope_scaling': {'rope_type': 'default', 'mrope_interleaved': True, 'mrope_section': [24, 20, 20]}, 'layer_types': ['full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention']}, 'vision_config': {'model_type': 'qwen3_vl', 'hidden_size': 1152, 'hidden_act': 'gelu_pytorch_tanh', 'intermediate_size': 4304, 'depth': 27, 'num_heads': 16, 'in_channels': 3, 'initializer_range': 0.02, 'out_hidden_size': 4096, 'patch_size': 16, 'spatial_merge_size': 2, 'temporal_patch_size': 2, 'num_position_embeddings': 2304, 'deepstack_visual_indexes': [8, 16, 24]}}, 'include_visual': False, 'vision_encoder_path': './checkpoints/Cosmos3-Nano/vision_encoder', 'skip_init_weights': True}, 'vision_latent_dim': 48, 'latent_patch_size': 2, 'max_action_dim': 64, 'num_embodiment_domains': 32, 'vision_in_proj': {'type': 'LinearProjector', 'in_dim': 192, 'out_dim': 4096}, 'vision_out_proj': {'type': 'LinearProjector', 'in_dim': 4096, 'out_dim': 192}, 'action_in_proj': {'type': 'DomainAwareLinear', 'input_size': 64, 'output_size': 4096, 'num_domains': 32}, 'action_out_proj': {'type': 'DomainAwareLinear', 'input_size': 4096, 'output_size': 64, 'num_domains': 32}, 'rectified_flow_training_config': {'shift': {'256': 3, '480': 5, '720': 10}, 'use_dynamic_shift': False, 'train_time_image_distribution': 'logitnormal', 'train_time_video_distribution': 'waver', 'train_time_action_distribution': 'logitnormal', 'train_time_weight': 'uniform', 'vision_loss_weight': 10.0, 'independent_action_schedule': False, 'shift_action': None, 'use_high_sigma_strategy': False, 'high_sigma_ratio': 0.05, 'high_sigma_timesteps_min': 995, 'high_sigma_timesteps_max': 1000, 'use_high_sigma_strategy_action': False, 'use_discrete_rf': False, 'normalize_loss_by_active': False, 'action_loss_weight': 10.0}, 'rectified_flow_inference_config': {'num_train_timesteps': 1000, 'scheduler_type': 'unipc', 'num_steps': 30, 'shift': 10.0, 'use_dynamic_shifting': False, 'use_karras_sigmas': False}, 'timestep_scale': 0.001, 'packed_attention_backend': 'flash2', 'position_embedding_type': 'unified_3d_mrope', 'unified_3d_mrope_reset_spatial_ids': True, 'unified_3d_mrope_temporal_modality_margin': 15000, 'enable_fps_modulation': True, 'base_fps': 24.0, 'special_tokens': {'eos_token_id': 151645, 'start_of_generation': 151652, 'end_of_generation': 151653}, 'pretrained_name_or_path': './checkpoints/Cosmos3-Nano/transformer', 'name_mapping': {'action_modality_embed': 'action_modality_embed', 'vlm_backbone.model.language_model.embed_tokens.weight': 'embed_tokens.weight', 'vlm_backbone.lm_head.weight': 'lm_head.weight', 'vlm_backbone.model.language_model.norm.weight': 'norm.weight', 'vlm_backbone.model.language_model.norm_moe_gen.weight': 'norm_moe_gen.weight', 'vlm_backbone.model.language_model.layers.': 'layers.', 'vision_in_proj.projector.': 'proj_in.', 'vision_out_proj.projector.': 'proj_out.', 'time_embedder.mlp.0.': 'time_embedder.linear_1.', 'time_embedder.mlp.2.': 'time_embedder.linear_2.', 'action_in_proj.': 'action_proj_in.', 'action_out_proj.': 'action_proj_out.', '.self_attn.q_proj.': '.self_attn.to_q.', '.self_attn.k_proj.': '.self_attn.to_k.', '.self_attn.v_proj.': '.self_attn.to_v.', '.self_attn.o_proj.': '.self_attn.to_out.', '.self_attn.q_proj_moe_gen.': '.self_attn.add_q_proj.', '.self_attn.k_proj_moe_gen.': '.self_attn.add_k_proj.', '.self_attn.v_proj_moe_gen.': '.self_attn.add_v_proj.', '.self_attn.o_proj_moe_gen.': '.self_attn.to_add_out.', '.self_attn.q_norm.': '.self_attn.norm_q.', '.self_attn.k_norm.': '.self_attn.norm_k.', '.self_attn.q_norm_moe_gen.': '.self_attn.norm_added_q.', '.self_attn.k_norm_moe_gen.': '.self_attn.norm_added_k.', 'vlm_backbone.model.visual.patch_embed.': 'patch_embed.', 'vlm_backbone.model.visual.blocks.': 'blocks.', 'vlm_backbone.model.visual.pos_embed.': 'pos_embed.', 'vlm_backbone.model.visual.merger.': 'merger.', 'vlm_backbone.model.visual.deepstack_merger_list.': 'deepstack_merger_list.'}, 'vision_vae': {'type': 'Cosmos3Wan22VAE', 'pretrained_name_or_path': None, 'encode_exact_durations': [17]}, 'ori_action_dim': 10, 'action_horizon': 16, 'freeze_vlm_backbone': False, 'freeze_non_moe_vlm_backbone': True, 'enable_vision_loss': True}, '_transforms': [{'type': 'ProcessParquetInputs', 'parquet_keys': ['observation.state', 'timestamp', 'actions', 'info', 'stats', 'action_masks'], 'video_keys': ['observation.images.image', 'observation.images.wrist_image'], 'name_mappings': {'observation.state': ['states'], 'actions': ['actions']}, 'embodiment_id': 5}, {'type': 'LiberoFramewiseActionToRot6D'}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.1, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}}, {'type': 'ResizeImages', 'height': 256, 'width': 256}, {'type': 'SimpleNormalizeImages'}, {'type': 'NormalizeStatesAndActions', 'action_dim': 64, 'state_dim': 64, 'state_key': 'proprio', 'action_key': 'action', 'state_norm_type': 'none', 'action_norm_type': 'quantile'}, {'type': 'BuildCosmos3Sequence', 'raw_action_dim': 10, 'mode': 'wam', 'frame_window_size': 17, 'prepend_state_to_action': False, 'conditioning_fps': 20.0}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 17, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}], 'train_dataloader': {'per_device_batch_size': 16, 'per_device_num_workers': 4, 'dataset': {'type': 'DistributedRepeatingDataset', 'name_mappings': {'observation.state': ['proprio'], 'action': ['action']}, 'statistic_keys': ['observation.state', 'timestamp', 'action'], 'statistic_name': 'all_libero_no_noops', 'statistics_overrides': {'all_libero_no_noops': {'action': {'mean': [0.050704, 0.097407, -0.094833, 0.994873, -0.004579, -0.004288, 0.004389, 0.996104, 0.001109, 0.476725], 'std': [0.333621, 0.387175, 0.45714, 0.010807, 0.077802, 0.063386, 0.078571, 0.009994, 0.038504, 0.49946], 'min': [-0.9375, -0.9375, -0.9375, 0.902028, -0.356085, -0.367416, -0.370434, 0.921907, -0.255, 0.0], 'max': [0.9375, 0.9375, 0.9375, 1.0, 0.368853, 0.341214, 0.356395, 1.0, 0.348251, 1.0], 'q01': [-0.723214, -0.808929, -0.9375, 0.934955, -0.223431, -0.189878, -0.334735, 0.938516, -0.107736, 0.0], 'q99': [0.9375, 0.870536, 0.9375, 1.0, 0.331, 0.163153, 0.226216, 1.0, 0.127158, 1.0]}}}, 'datasets': [{'type': 'ParquetDataset', 'data_root_path': './datasets/libero_spatial_no_noops_lerobotv2.1', 'transforms': [{'type': 'ProcessParquetInputs', 'parquet_keys': ['observation.state', 'timestamp', 'actions', 'info', 'stats', 'action_masks'], 'video_keys': ['observation.images.image', 'observation.images.wrist_image'], 'name_mappings': {'observation.state': ['states'], 'actions': ['actions']}, 'embodiment_id': 5}, {'type': 'LiberoFramewiseActionToRot6D'}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.1, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}}, {'type': 'ResizeImages', 'height': 256, 'width': 256}, {'type': 'SimpleNormalizeImages'}, {'type': 'NormalizeStatesAndActions', 'action_dim': 64, 'state_dim': 64, 'state_key': 'proprio', 'action_key': 'action', 'state_norm_type': 'none', 'action_norm_type': 'quantile'}, {'type': 'BuildCosmos3Sequence', 'raw_action_dim': 10, 'mode': 'wam', 'frame_window_size': 17, 'prepend_state_to_action': False, 'conditioning_fps': 20.0}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 17, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}], 'action_window_size': 16, 'action_key': 'action', 'use_delta': False, 'statistic_name': 'all_libero_no_noops', 'window_start_idx': 0, 'frame_window_size': 17, 'require_full_window': True}, {'type': 'ParquetDataset', 'data_root_path': './datasets/libero_spatial_no_noops_lerobotv2.1', 'transforms': [{'type': 'ProcessParquetInputs', 'parquet_keys': ['observation.state', 'timestamp', 'actions', 'info', 'stats', 'action_masks'], 'video_keys': ['observation.images.image', 'observation.images.wrist_image'], 'name_mappings': {'observation.state': ['states'], 'actions': ['actions']}, 'embodiment_id': 5}, {'type': 'LiberoFramewiseActionToRot6D'}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.1, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}}, {'type': 'ResizeImages', 'height': 256, 'width': 256}, {'type': 'SimpleNormalizeImages'}, {'type': 'NormalizeStatesAndActions', 'action_dim': 64, 'state_dim': 64, 'state_key': 'proprio', 'action_key': 'action', 'state_norm_type': 'none', 'action_norm_type': 'quantile'}, {'type': 'BuildCosmos3Sequence', 'raw_action_dim': 10, 'mode': 'wam', 'frame_window_size': 17, 'prepend_state_to_action': False, 'conditioning_fps': 20.0}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 17, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}], 'action_window_size': 16, 'action_key': 'action', 'use_delta': False, 'statistic_name': 'all_libero_no_noops', 'window_start_idx': 0, 'frame_window_size': 17, 'require_full_window': True}, {'type': 'ParquetDataset', 'data_root_path': './datasets/libero_object_no_noops_lerobotv2.1', 'transforms': [{'type': 'ProcessParquetInputs', 'parquet_keys': ['observation.state', 'timestamp', 'actions', 'info', 'stats', 'action_masks'], 'video_keys': ['observation.images.image', 'observation.images.wrist_image'], 'name_mappings': {'observation.state': ['states'], 'actions': ['actions']}, 'embodiment_id': 5}, {'type': 'LiberoFramewiseActionToRot6D'}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.1, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}}, {'type': 'ResizeImages', 'height': 256, 'width': 256}, {'type': 'SimpleNormalizeImages'}, {'type': 'NormalizeStatesAndActions', 'action_dim': 64, 'state_dim': 64, 'state_key': 'proprio', 'action_key': 'action', 'state_norm_type': 'none', 'action_norm_type': 'quantile'}, {'type': 'BuildCosmos3Sequence', 'raw_action_dim': 10, 'mode': 'wam', 'frame_window_size': 17, 'prepend_state_to_action': False, 'conditioning_fps': 20.0}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 17, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}], 'action_window_size': 16, 'action_key': 'action', 'use_delta': False, 'statistic_name': 'all_libero_no_noops', 'window_start_idx': 0, 'frame_window_size': 17, 'require_full_window': True}, {'type': 'ParquetDataset', 'data_root_path': './datasets/libero_object_no_noops_lerobotv2.1', 'transforms': [{'type': 'ProcessParquetInputs', 'parquet_keys': ['observation.state', 'timestamp', 'actions', 'info', 'stats', 'action_masks'], 'video_keys': ['observation.images.image', 'observation.images.wrist_image'], 'name_mappings': {'observation.state': ['states'], 'actions': ['actions']}, 'embodiment_id': 5}, {'type': 'LiberoFramewiseActionToRot6D'}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.1, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}}, {'type': 'ResizeImages', 'height': 256, 'width': 256}, {'type': 'SimpleNormalizeImages'}, {'type': 'NormalizeStatesAndActions', 'action_dim': 64, 'state_dim': 64, 'state_key': 'proprio', 'action_key': 'action', 'state_norm_type': 'none', 'action_norm_type': 'quantile'}, {'type': 'BuildCosmos3Sequence', 'raw_action_dim': 10, 'mode': 'wam', 'frame_window_size': 17, 'prepend_state_to_action': False, 'conditioning_fps': 20.0}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 17, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}], 'action_window_size': 16, 'action_key': 'action', 'use_delta': False, 'statistic_name': 'all_libero_no_noops', 'window_start_idx': 0, 'frame_window_size': 17, 'require_full_window': True}, {'type': 'ParquetDataset', 'data_root_path': './datasets/libero_goal_no_noops_lerobotv2.1', 'transforms': [{'type': 'ProcessParquetInputs', 'parquet_keys': ['observation.state', 'timestamp', 'actions', 'info', 'stats', 'action_masks'], 'video_keys': ['observation.images.image', 'observation.images.wrist_image'], 'name_mappings': {'observation.state': ['states'], 'actions': ['actions']}, 'embodiment_id': 5}, {'type': 'LiberoFramewiseActionToRot6D'}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.1, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}}, {'type': 'ResizeImages', 'height': 256, 'width': 256}, {'type': 'SimpleNormalizeImages'}, {'type': 'NormalizeStatesAndActions', 'action_dim': 64, 'state_dim': 64, 'state_key': 'proprio', 'action_key': 'action', 'state_norm_type': 'none', 'action_norm_type': 'quantile'}, {'type': 'BuildCosmos3Sequence', 'raw_action_dim': 10, 'mode': 'wam', 'frame_window_size': 17, 'prepend_state_to_action': False, 'conditioning_fps': 20.0}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 17, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}], 'action_window_size': 16, 'action_key': 'action', 'use_delta': False, 'statistic_name': 'all_libero_no_noops', 'window_start_idx': 0, 'frame_window_size': 17, 'require_full_window': True}, {'type': 'ParquetDataset', 'data_root_path': './datasets/libero_goal_no_noops_lerobotv2.1', 'transforms': [{'type': 'ProcessParquetInputs', 'parquet_keys': ['observation.state', 'timestamp', 'actions', 'info', 'stats', 'action_masks'], 'video_keys': ['observation.images.image', 'observation.images.wrist_image'], 'name_mappings': {'observation.state': ['states'], 'actions': ['actions']}, 'embodiment_id': 5}, {'type': 'LiberoFramewiseActionToRot6D'}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.1, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}}, {'type': 'ResizeImages', 'height': 256, 'width': 256}, {'type': 'SimpleNormalizeImages'}, {'type': 'NormalizeStatesAndActions', 'action_dim': 64, 'state_dim': 64, 'state_key': 'proprio', 'action_key': 'action', 'state_norm_type': 'none', 'action_norm_type': 'quantile'}, {'type': 'BuildCosmos3Sequence', 'raw_action_dim': 10, 'mode': 'wam', 'frame_window_size': 17, 'prepend_state_to_action': False, 'conditioning_fps': 20.0}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 17, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}], 'action_window_size': 16, 'action_key': 'action', 'use_delta': False, 'statistic_name': 'all_libero_no_noops', 'window_start_idx': 0, 'frame_window_size': 17, 'require_full_window': True}, {'type': 'ParquetDataset', 'data_root_path': './datasets/libero_10_no_noops_lerobotv2.1', 'transforms': [{'type': 'ProcessParquetInputs', 'parquet_keys': ['observation.state', 'timestamp', 'actions', 'info', 'stats', 'action_masks'], 'video_keys': ['observation.images.image', 'observation.images.wrist_image'], 'name_mappings': {'observation.state': ['states'], 'actions': ['actions']}, 'embodiment_id': 5}, {'type': 'LiberoFramewiseActionToRot6D'}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.1, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}}, {'type': 'ResizeImages', 'height': 256, 'width': 256}, {'type': 'SimpleNormalizeImages'}, {'type': 'NormalizeStatesAndActions', 'action_dim': 64, 'state_dim': 64, 'state_key': 'proprio', 'action_key': 'action', 'state_norm_type': 'none', 'action_norm_type': 'quantile'}, {'type': 'BuildCosmos3Sequence', 'raw_action_dim': 10, 'mode': 'wam', 'frame_window_size': 17, 'prepend_state_to_action': False, 'conditioning_fps': 20.0}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 17, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}], 'action_window_size': 16, 'action_key': 'action', 'use_delta': False, 'statistic_name': 'all_libero_no_noops', 'window_start_idx': 0, 'frame_window_size': 17, 'require_full_window': True}]}}, 'runner': {'type': 'FSDPTrainRunner', 'max_steps': 5000, 'save_iter_interval': 500, 'max_keep_ckpts': 2, 'optimizer': {'lr': 5e-05, 'type': 'AdamW', 'weight_decay': 0.05, 'betas': (0.9, 0.99), 'eps': 1e-08, 'fused': True, 'exclude_1d_from_weight_decay': False, 'paramwise_learning_rate': {'action_in_proj.': 0.00025, 'action_out_proj.': 0.00025, 'action_modality_embed': 0.00025}}, 'max_grad_norm': 1.0, 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'collator': {'type': 'Cosmos3Collator', 'tensor_keys': ['images', 'actions', 'embodiment_ids', 'raw_action_dim', 'conditioning_fps', 'action_fps'], 'sequence_keys': ['text_token_ids'], 'list_keys': ['sequence_plan'], 'meta_keys': ['task_description', 'stats', 'info', 'timestamp'], 'pad_id': 0}, 'sampler': None, 'grad_accumulation_steps': 8, 'metric': {'type': 'VLAMetric', 'active_trackers': ('jsonl', 'wandb'), 'run_dir': '/mnt/data/cpfs/users/jikun/wk_dir/cosmos3/libero_align_16_9_padding', 'grad_accumulation_steps': 8, 'window_size': 1}, 'lr_scheduler': {'type': 'linear-warmup+linear-decay', 'warmup_steps': 500, 'cycle_length': 16000}, 'enable_gradient_checkpointing': True, 'enable_mixed_precision_training': True, 'mixed_precision_dtype': 'bf16', 'sharding_strategy': 'full-shard', 'change_key_name': False, 'cfg': Config (path: configs/cosmos3/cosmos3nano_libero_full_finetune.py): {'_ckpt_root': './checkpoints', '_cosmos3_nano_ckpt': './checkpoints/Cosmos3-Nano', '_cosmos3_nano_transformer': './checkpoints/Cosmos3-Nano/transformer', '_cosmos3_nano_vision_encoder': './checkpoints/Cosmos3-Nano/vision_encoder', '_cosmos3_nano_tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, '_wan22_vae_path': './checkpoints/Wan2.2-TI2V-5B/Wan2.2_VAE.pth', '_vision_vae': {'type': 'Cosmos3Wan22VAE', 'pretrained_name_or_path': './checkpoints/Wan2.2-TI2V-5B/Wan2.2_VAE.pth', 'encode_exact_durations': [17]}, '_action_dim': 10, '_max_action_dim': 64, '_max_state_dim': 64, '_action_horizon': 16, '_frame_window_size': 17, '_prepend_state_to_action': False, '_image_height': 256, '_image_width': 256, '_video_height': 192, '_video_width': 320, '_conditioning_fps': 20.0, '_libero_action_stats': {'mean': [0.050704, 0.097407, -0.094833, 0.994873, -0.004579, -0.004288, 0.004389, 0.996104, 0.001109, 0.476725], 'std': [0.333621, 0.387175, 0.45714, 0.010807, 0.077802, 0.063386, 0.078571, 0.009994, 0.038504, 0.49946], 'min': [-0.9375, -0.9375, -0.9375, 0.902028, -0.356085, -0.367416, -0.370434, 0.921907, -0.255, 0.0], 'max': [0.9375, 0.9375, 0.9375, 1.0, 0.368853, 0.341214, 0.356395, 1.0, 0.348251, 1.0], 'q01': [-0.723214, -0.808929, -0.9375, 0.934955, -0.223431, -0.189878, -0.334735, 0.938516, -0.107736, 0.0], 'q99': [0.9375, 0.870536, 0.9375, 1.0, 0.331, 0.163153, 0.226216, 1.0, 0.127158, 1.0]}, '_cfg_dropout_rate': 0.1, '_base_lr': 5e-05, '_action_lr': 0.00025, '_per_device_batch_size': 16, '_grad_accumulation_steps': 8, '_max_steps': 5000, '_save_iter_interval': 500, '_data_root_path': ['./datasets/libero_spatial_no_noops_lerobotv2.1', './datasets/libero_spatial_no_noops_lerobotv2.1', './datasets/libero_object_no_noops_lerobotv2.1', './datasets/libero_object_no_noops_lerobotv2.1', './datasets/libero_goal_no_noops_lerobotv2.1', './datasets/libero_goal_no_noops_lerobotv2.1', './datasets/libero_10_no_noops_lerobotv2.1'], '_statistic_name': 'all_libero_no_noops', '_cosmos3_nano_special_tokens': {'eos_token_id': 151645, 'start_of_generation': 151652, 'end_of_generation': 151653}, '_cosmos3_nano_vlm_config': {'model_type': 'qwen3_vl', 'vocab_size': 151936, 'tie_word_embeddings': False, 'image_token_id': 151655, 'video_token_id': 151656, 'vision_start_token_id': 151652, 'vision_end_token_id': 151653, 'text_config': {'model_type': 'qwen3_vl_text', 'vocab_size': 151936, 'hidden_size': 4096, 'intermediate_size': 12288, 'num_hidden_layers': 36, 'num_attention_heads': 32, 'num_key_value_heads': 8, 'head_dim': 128, 'hidden_act': 'silu', 'max_position_embeddings': 262144, 'rms_norm_eps': 1e-06, 'rope_theta': 5000000, 'attention_bias': False, 'attention_dropout': 0.0, 'bos_token_id': 151643, 'eos_token_id': 151645, 'pad_token_id': 0, 'tie_word_embeddings': False, 'rope_scaling': {'rope_type': 'default', 'mrope_interleaved': True, 'mrope_section': [24, 20, 20]}, 'layer_types': ['full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention']}, 'vision_config': {'model_type': 'qwen3_vl', 'hidden_size': 1152, 'hidden_act': 'gelu_pytorch_tanh', 'intermediate_size': 4304, 'depth': 27, 'num_heads': 16, 'in_channels': 3, 'initializer_range': 0.02, 'out_hidden_size': 4096, 'patch_size': 16, 'spatial_merge_size': 2, 'temporal_patch_size': 2, 'num_position_embeddings': 2304, 'deepstack_visual_indexes': [8, 16, 24]}}, '_cosmos3_nano_name_mapping': {'action_modality_embed': 'action_modality_embed', 'vlm_backbone.model.language_model.embed_tokens.weight': 'embed_tokens.weight', 'vlm_backbone.lm_head.weight': 'lm_head.weight', 'vlm_backbone.model.language_model.norm.weight': 'norm.weight', 'vlm_backbone.model.language_model.norm_moe_gen.weight': 'norm_moe_gen.weight', 'vlm_backbone.model.language_model.layers.': 'layers.', 'vision_in_proj.projector.': 'proj_in.', 'vision_out_proj.projector.': 'proj_out.', 'time_embedder.mlp.0.': 'time_embedder.linear_1.', 'time_embedder.mlp.2.': 'time_embedder.linear_2.', 'action_in_proj.': 'action_proj_in.', 'action_out_proj.': 'action_proj_out.', '.self_attn.q_proj.': '.self_attn.to_q.', '.self_attn.k_proj.': '.self_attn.to_k.', '.self_attn.v_proj.': '.self_attn.to_v.', '.self_attn.o_proj.': '.self_attn.to_out.', '.self_attn.q_proj_moe_gen.': '.self_attn.add_q_proj.', '.self_attn.k_proj_moe_gen.': '.self_attn.add_k_proj.', '.self_attn.v_proj_moe_gen.': '.self_attn.add_v_proj.', '.self_attn.o_proj_moe_gen.': '.self_attn.to_add_out.', '.self_attn.q_norm.': '.self_attn.norm_q.', '.self_attn.k_norm.': '.self_attn.norm_k.', '.self_attn.q_norm_moe_gen.': '.self_attn.norm_added_q.', '.self_attn.k_norm_moe_gen.': '.self_attn.norm_added_k.', 'vlm_backbone.model.visual.patch_embed.': 'patch_embed.', 'vlm_backbone.model.visual.blocks.': 'blocks.', 'vlm_backbone.model.visual.pos_embed.': 'pos_embed.', 'vlm_backbone.model.visual.merger.': 'merger.', 'vlm_backbone.model.visual.deepstack_merger_list.': 'deepstack_merger_list.'}, 'model': {'type': 'Cosmos3FlowMatching', 'vlm_backbone': {'type': 'Cosmos3MoTBackbone', 'vlm_config': {'model_type': 'qwen3_vl', 'vocab_size': 151936, 'tie_word_embeddings': False, 'image_token_id': 151655, 'video_token_id': 151656, 'vision_start_token_id': 151652, 'vision_end_token_id': 151653, 'text_config': {'model_type': 'qwen3_vl_text', 'vocab_size': 151936, 'hidden_size': 4096, 'intermediate_size': 12288, 'num_hidden_layers': 36, 'num_attention_heads': 32, 'num_key_value_heads': 8, 'head_dim': 128, 'hidden_act': 'silu', 'max_position_embeddings': 262144, 'rms_norm_eps': 1e-06, 'rope_theta': 5000000, 'attention_bias': False, 'attention_dropout': 0.0, 'bos_token_id': 151643, 'eos_token_id': 151645, 'pad_token_id': 0, 'tie_word_embeddings': False, 'rope_scaling': {'rope_type': 'default', 'mrope_interleaved': True, 'mrope_section': [24, 20, 20]}, 'layer_types': ['full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention']}, 'vision_config': {'model_type': 'qwen3_vl', 'hidden_size': 1152, 'hidden_act': 'gelu_pytorch_tanh', 'intermediate_size': 4304, 'depth': 27, 'num_heads': 16, 'in_channels': 3, 'initializer_range': 0.02, 'out_hidden_size': 4096, 'patch_size': 16, 'spatial_merge_size': 2, 'temporal_patch_size': 2, 'num_position_embeddings': 2304, 'deepstack_visual_indexes': [8, 16, 24]}}, 'include_visual': False, 'vision_encoder_path': './checkpoints/Cosmos3-Nano/vision_encoder', 'skip_init_weights': True}, 'vision_latent_dim': 48, 'latent_patch_size': 2, 'max_action_dim': 64, 'num_embodiment_domains': 32, 'vision_in_proj': {'type': 'LinearProjector', 'in_dim': 192, 'out_dim': 4096}, 'vision_out_proj': {'type': 'LinearProjector', 'in_dim': 4096, 'out_dim': 192}, 'action_in_proj': {'type': 'DomainAwareLinear', 'input_size': 64, 'output_size': 4096, 'num_domains': 32}, 'action_out_proj': {'type': 'DomainAwareLinear', 'input_size': 4096, 'output_size': 64, 'num_domains': 32}, 'rectified_flow_training_config': {'shift': {'256': 3, '480': 5, '720': 10}, 'use_dynamic_shift': False, 'train_time_image_distribution': 'logitnormal', 'train_time_video_distribution': 'waver', 'train_time_action_distribution': 'logitnormal', 'train_time_weight': 'uniform', 'vision_loss_weight': 10.0, 'independent_action_schedule': False, 'shift_action': None, 'use_high_sigma_strategy': False, 'high_sigma_ratio': 0.05, 'high_sigma_timesteps_min': 995, 'high_sigma_timesteps_max': 1000, 'use_high_sigma_strategy_action': False, 'use_discrete_rf': False, 'normalize_loss_by_active': False, 'action_loss_weight': 10.0}, 'rectified_flow_inference_config': {'num_train_timesteps': 1000, 'scheduler_type': 'unipc', 'num_steps': 30, 'shift': 10.0, 'use_dynamic_shifting': False, 'use_karras_sigmas': False}, 'timestep_scale': 0.001, 'packed_attention_backend': 'flash2', 'position_embedding_type': 'unified_3d_mrope', 'unified_3d_mrope_reset_spatial_ids': True, 'unified_3d_mrope_temporal_modality_margin': 15000, 'enable_fps_modulation': True, 'base_fps': 24.0, 'special_tokens': {'eos_token_id': 151645, 'start_of_generation': 151652, 'end_of_generation': 151653}, 'pretrained_name_or_path': './checkpoints/Cosmos3-Nano/transformer', 'name_mapping': {'action_modality_embed': 'action_modality_embed', 'vlm_backbone.model.language_model.embed_tokens.weight': 'embed_tokens.weight', 'vlm_backbone.lm_head.weight': 'lm_head.weight', 'vlm_backbone.model.language_model.norm.weight': 'norm.weight', 'vlm_backbone.model.language_model.norm_moe_gen.weight': 'norm_moe_gen.weight', 'vlm_backbone.model.language_model.layers.': 'layers.', 'vision_in_proj.projector.': 'proj_in.', 'vision_out_proj.projector.': 'proj_out.', 'time_embedder.mlp.0.': 'time_embedder.linear_1.', 'time_embedder.mlp.2.': 'time_embedder.linear_2.', 'action_in_proj.': 'action_proj_in.', 'action_out_proj.': 'action_proj_out.', '.self_attn.q_proj.': '.self_attn.to_q.', '.self_attn.k_proj.': '.self_attn.to_k.', '.self_attn.v_proj.': '.self_attn.to_v.', '.self_attn.o_proj.': '.self_attn.to_out.', '.self_attn.q_proj_moe_gen.': '.self_attn.add_q_proj.', '.self_attn.k_proj_moe_gen.': '.self_attn.add_k_proj.', '.self_attn.v_proj_moe_gen.': '.self_attn.add_v_proj.', '.self_attn.o_proj_moe_gen.': '.self_attn.to_add_out.', '.self_attn.q_norm.': '.self_attn.norm_q.', '.self_attn.k_norm.': '.self_attn.norm_k.', '.self_attn.q_norm_moe_gen.': '.self_attn.norm_added_q.', '.self_attn.k_norm_moe_gen.': '.self_attn.norm_added_k.', 'vlm_backbone.model.visual.patch_embed.': 'patch_embed.', 'vlm_backbone.model.visual.blocks.': 'blocks.', 'vlm_backbone.model.visual.pos_embed.': 'pos_embed.', 'vlm_backbone.model.visual.merger.': 'merger.', 'vlm_backbone.model.visual.deepstack_merger_list.': 'deepstack_merger_list.'}, 'vision_vae': {'type': 'Cosmos3Wan22VAE', 'pretrained_name_or_path': './checkpoints/Wan2.2-TI2V-5B/Wan2.2_VAE.pth', 'encode_exact_durations': [17]}, 'ori_action_dim': 10, 'action_horizon': 16, 'freeze_vlm_backbone': False, 'freeze_non_moe_vlm_backbone': True, 'enable_vision_loss': True}, 'inference_model': {'type': 'Cosmos3FlowMatching', 'vlm_backbone': {'type': 'Cosmos3MoTBackbone', 'vlm_config': {'model_type': 'qwen3_vl', 'vocab_size': 151936, 'tie_word_embeddings': False, 'image_token_id': 151655, 'video_token_id': 151656, 'vision_start_token_id': 151652, 'vision_end_token_id': 151653, 'text_config': {'model_type': 'qwen3_vl_text', 'vocab_size': 151936, 'hidden_size': 4096, 'intermediate_size': 12288, 'num_hidden_layers': 36, 'num_attention_heads': 32, 'num_key_value_heads': 8, 'head_dim': 128, 'hidden_act': 'silu', 'max_position_embeddings': 262144, 'rms_norm_eps': 1e-06, 'rope_theta': 5000000, 'attention_bias': False, 'attention_dropout': 0.0, 'bos_token_id': 151643, 'eos_token_id': 151645, 'pad_token_id': 0, 'tie_word_embeddings': False, 'rope_scaling': {'rope_type': 'default', 'mrope_interleaved': True, 'mrope_section': [24, 20, 20]}, 'layer_types': ['full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention', 'full_attention']}, 'vision_config': {'model_type': 'qwen3_vl', 'hidden_size': 1152, 'hidden_act': 'gelu_pytorch_tanh', 'intermediate_size': 4304, 'depth': 27, 'num_heads': 16, 'in_channels': 3, 'initializer_range': 0.02, 'out_hidden_size': 4096, 'patch_size': 16, 'spatial_merge_size': 2, 'temporal_patch_size': 2, 'num_position_embeddings': 2304, 'deepstack_visual_indexes': [8, 16, 24]}}, 'include_visual': False, 'vision_encoder_path': './checkpoints/Cosmos3-Nano/vision_encoder', 'skip_init_weights': True}, 'vision_latent_dim': 48, 'latent_patch_size': 2, 'max_action_dim': 64, 'num_embodiment_domains': 32, 'vision_in_proj': {'type': 'LinearProjector', 'in_dim': 192, 'out_dim': 4096}, 'vision_out_proj': {'type': 'LinearProjector', 'in_dim': 4096, 'out_dim': 192}, 'action_in_proj': {'type': 'DomainAwareLinear', 'input_size': 64, 'output_size': 4096, 'num_domains': 32}, 'action_out_proj': {'type': 'DomainAwareLinear', 'input_size': 4096, 'output_size': 64, 'num_domains': 32}, 'rectified_flow_training_config': {'shift': {'256': 3, '480': 5, '720': 10}, 'use_dynamic_shift': False, 'train_time_image_distribution': 'logitnormal', 'train_time_video_distribution': 'waver', 'train_time_action_distribution': 'logitnormal', 'train_time_weight': 'uniform', 'vision_loss_weight': 10.0, 'independent_action_schedule': False, 'shift_action': None, 'use_high_sigma_strategy': False, 'high_sigma_ratio': 0.05, 'high_sigma_timesteps_min': 995, 'high_sigma_timesteps_max': 1000, 'use_high_sigma_strategy_action': False, 'use_discrete_rf': False, 'normalize_loss_by_active': False, 'action_loss_weight': 10.0}, 'rectified_flow_inference_config': {'num_train_timesteps': 1000, 'scheduler_type': 'unipc', 'num_steps': 30, 'shift': 10.0, 'use_dynamic_shifting': False, 'use_karras_sigmas': False}, 'timestep_scale': 0.001, 'packed_attention_backend': 'flash2', 'position_embedding_type': 'unified_3d_mrope', 'unified_3d_mrope_reset_spatial_ids': True, 'unified_3d_mrope_temporal_modality_margin': 15000, 'enable_fps_modulation': True, 'base_fps': 24.0, 'special_tokens': {'eos_token_id': 151645, 'start_of_generation': 151652, 'end_of_generation': 151653}, 'pretrained_name_or_path': './checkpoints/Cosmos3-Nano/transformer', 'name_mapping': {'action_modality_embed': 'action_modality_embed', 'vlm_backbone.model.language_model.embed_tokens.weight': 'embed_tokens.weight', 'vlm_backbone.lm_head.weight': 'lm_head.weight', 'vlm_backbone.model.language_model.norm.weight': 'norm.weight', 'vlm_backbone.model.language_model.norm_moe_gen.weight': 'norm_moe_gen.weight', 'vlm_backbone.model.language_model.layers.': 'layers.', 'vision_in_proj.projector.': 'proj_in.', 'vision_out_proj.projector.': 'proj_out.', 'time_embedder.mlp.0.': 'time_embedder.linear_1.', 'time_embedder.mlp.2.': 'time_embedder.linear_2.', 'action_in_proj.': 'action_proj_in.', 'action_out_proj.': 'action_proj_out.', '.self_attn.q_proj.': '.self_attn.to_q.', '.self_attn.k_proj.': '.self_attn.to_k.', '.self_attn.v_proj.': '.self_attn.to_v.', '.self_attn.o_proj.': '.self_attn.to_out.', '.self_attn.q_proj_moe_gen.': '.self_attn.add_q_proj.', '.self_attn.k_proj_moe_gen.': '.self_attn.add_k_proj.', '.self_attn.v_proj_moe_gen.': '.self_attn.add_v_proj.', '.self_attn.o_proj_moe_gen.': '.self_attn.to_add_out.', '.self_attn.q_norm.': '.self_attn.norm_q.', '.self_attn.k_norm.': '.self_attn.norm_k.', '.self_attn.q_norm_moe_gen.': '.self_attn.norm_added_q.', '.self_attn.k_norm_moe_gen.': '.self_attn.norm_added_k.', 'vlm_backbone.model.visual.patch_embed.': 'patch_embed.', 'vlm_backbone.model.visual.blocks.': 'blocks.', 'vlm_backbone.model.visual.pos_embed.': 'pos_embed.', 'vlm_backbone.model.visual.merger.': 'merger.', 'vlm_backbone.model.visual.deepstack_merger_list.': 'deepstack_merger_list.'}, 'vision_vae': {'type': 'Cosmos3Wan22VAE', 'pretrained_name_or_path': None, 'encode_exact_durations': [17]}, 'ori_action_dim': 10, 'action_horizon': 16, 'freeze_vlm_backbone': False, 'freeze_non_moe_vlm_backbone': True, 'enable_vision_loss': True}, '_transforms': [{'type': 'ProcessParquetInputs', 'parquet_keys': ['observation.state', 'timestamp', 'actions', 'info', 'stats', 'action_masks'], 'video_keys': ['observation.images.image', 'observation.images.wrist_image'], 'name_mappings': {'observation.state': ['states'], 'actions': ['actions']}, 'embodiment_id': 5}, {'type': 'LiberoFramewiseActionToRot6D'}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.1, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}}, {'type': 'ResizeImages', 'height': 256, 'width': 256}, {'type': 'SimpleNormalizeImages'}, {'type': 'NormalizeStatesAndActions', 'action_dim': 64, 'state_dim': 64, 'state_key': 'proprio', 'action_key': 'action', 'state_norm_type': 'none', 'action_norm_type': 'quantile'}, {'type': 'BuildCosmos3Sequence', 'raw_action_dim': 10, 'mode': 'wam', 'frame_window_size': 17, 'prepend_state_to_action': False, 'conditioning_fps': 20.0}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 17, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}], 'train_dataloader': {'per_device_batch_size': 16, 'per_device_num_workers': 4, 'dataset': {'type': 'DistributedRepeatingDataset', 'name_mappings': {'observation.state': ['proprio'], 'action': ['action']}, 'statistic_keys': ['observation.state', 'timestamp', 'action'], 'statistic_name': 'all_libero_no_noops', 'statistics_overrides': {'all_libero_no_noops': {'action': {'mean': [0.050704, 0.097407, -0.094833, 0.994873, -0.004579, -0.004288, 0.004389, 0.996104, 0.001109, 0.476725], 'std': [0.333621, 0.387175, 0.45714, 0.010807, 0.077802, 0.063386, 0.078571, 0.009994, 0.038504, 0.49946], 'min': [-0.9375, -0.9375, -0.9375, 0.902028, -0.356085, -0.367416, -0.370434, 0.921907, -0.255, 0.0], 'max': [0.9375, 0.9375, 0.9375, 1.0, 0.368853, 0.341214, 0.356395, 1.0, 0.348251, 1.0], 'q01': [-0.723214, -0.808929, -0.9375, 0.934955, -0.223431, -0.189878, -0.334735, 0.938516, -0.107736, 0.0], 'q99': [0.9375, 0.870536, 0.9375, 1.0, 0.331, 0.163153, 0.226216, 1.0, 0.127158, 1.0]}}}, 'datasets': [{'type': 'ParquetDataset', 'data_root_path': './datasets/libero_spatial_no_noops_lerobotv2.1', 'transforms': [{'type': 'ProcessParquetInputs', 'parquet_keys': ['observation.state', 'timestamp', 'actions', 'info', 'stats', 'action_masks'], 'video_keys': ['observation.images.image', 'observation.images.wrist_image'], 'name_mappings': {'observation.state': ['states'], 'actions': ['actions']}, 'embodiment_id': 5}, {'type': 'LiberoFramewiseActionToRot6D'}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.1, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}}, {'type': 'ResizeImages', 'height': 256, 'width': 256}, {'type': 'SimpleNormalizeImages'}, {'type': 'NormalizeStatesAndActions', 'action_dim': 64, 'state_dim': 64, 'state_key': 'proprio', 'action_key': 'action', 'state_norm_type': 'none', 'action_norm_type': 'quantile'}, {'type': 'BuildCosmos3Sequence', 'raw_action_dim': 10, 'mode': 'wam', 'frame_window_size': 17, 'prepend_state_to_action': False, 'conditioning_fps': 20.0}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 17, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}], 'action_window_size': 16, 'action_key': 'action', 'use_delta': False, 'statistic_name': 'all_libero_no_noops', 'window_start_idx': 0, 'frame_window_size': 17, 'require_full_window': True}, {'type': 'ParquetDataset', 'data_root_path': './datasets/libero_spatial_no_noops_lerobotv2.1', 'transforms': [{'type': 'ProcessParquetInputs', 'parquet_keys': ['observation.state', 'timestamp', 'actions', 'info', 'stats', 'action_masks'], 'video_keys': ['observation.images.image', 'observation.images.wrist_image'], 'name_mappings': {'observation.state': ['states'], 'actions': ['actions']}, 'embodiment_id': 5}, {'type': 'LiberoFramewiseActionToRot6D'}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.1, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}}, {'type': 'ResizeImages', 'height': 256, 'width': 256}, {'type': 'SimpleNormalizeImages'}, {'type': 'NormalizeStatesAndActions', 'action_dim': 64, 'state_dim': 64, 'state_key': 'proprio', 'action_key': 'action', 'state_norm_type': 'none', 'action_norm_type': 'quantile'}, {'type': 'BuildCosmos3Sequence', 'raw_action_dim': 10, 'mode': 'wam', 'frame_window_size': 17, 'prepend_state_to_action': False, 'conditioning_fps': 20.0}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 17, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}], 'action_window_size': 16, 'action_key': 'action', 'use_delta': False, 'statistic_name': 'all_libero_no_noops', 'window_start_idx': 0, 'frame_window_size': 17, 'require_full_window': True}, {'type': 'ParquetDataset', 'data_root_path': './datasets/libero_object_no_noops_lerobotv2.1', 'transforms': [{'type': 'ProcessParquetInputs', 'parquet_keys': ['observation.state', 'timestamp', 'actions', 'info', 'stats', 'action_masks'], 'video_keys': ['observation.images.image', 'observation.images.wrist_image'], 'name_mappings': {'observation.state': ['states'], 'actions': ['actions']}, 'embodiment_id': 5}, {'type': 'LiberoFramewiseActionToRot6D'}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.1, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}}, {'type': 'ResizeImages', 'height': 256, 'width': 256}, {'type': 'SimpleNormalizeImages'}, {'type': 'NormalizeStatesAndActions', 'action_dim': 64, 'state_dim': 64, 'state_key': 'proprio', 'action_key': 'action', 'state_norm_type': 'none', 'action_norm_type': 'quantile'}, {'type': 'BuildCosmos3Sequence', 'raw_action_dim': 10, 'mode': 'wam', 'frame_window_size': 17, 'prepend_state_to_action': False, 'conditioning_fps': 20.0}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 17, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}], 'action_window_size': 16, 'action_key': 'action', 'use_delta': False, 'statistic_name': 'all_libero_no_noops', 'window_start_idx': 0, 'frame_window_size': 17, 'require_full_window': True}, {'type': 'ParquetDataset', 'data_root_path': './datasets/libero_object_no_noops_lerobotv2.1', 'transforms': [{'type': 'ProcessParquetInputs', 'parquet_keys': ['observation.state', 'timestamp', 'actions', 'info', 'stats', 'action_masks'], 'video_keys': ['observation.images.image', 'observation.images.wrist_image'], 'name_mappings': {'observation.state': ['states'], 'actions': ['actions']}, 'embodiment_id': 5}, {'type': 'LiberoFramewiseActionToRot6D'}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.1, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}}, {'type': 'ResizeImages', 'height': 256, 'width': 256}, {'type': 'SimpleNormalizeImages'}, {'type': 'NormalizeStatesAndActions', 'action_dim': 64, 'state_dim': 64, 'state_key': 'proprio', 'action_key': 'action', 'state_norm_type': 'none', 'action_norm_type': 'quantile'}, {'type': 'BuildCosmos3Sequence', 'raw_action_dim': 10, 'mode': 'wam', 'frame_window_size': 17, 'prepend_state_to_action': False, 'conditioning_fps': 20.0}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 17, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}], 'action_window_size': 16, 'action_key': 'action', 'use_delta': False, 'statistic_name': 'all_libero_no_noops', 'window_start_idx': 0, 'frame_window_size': 17, 'require_full_window': True}, {'type': 'ParquetDataset', 'data_root_path': './datasets/libero_goal_no_noops_lerobotv2.1', 'transforms': [{'type': 'ProcessParquetInputs', 'parquet_keys': ['observation.state', 'timestamp', 'actions', 'info', 'stats', 'action_masks'], 'video_keys': ['observation.images.image', 'observation.images.wrist_image'], 'name_mappings': {'observation.state': ['states'], 'actions': ['actions']}, 'embodiment_id': 5}, {'type': 'LiberoFramewiseActionToRot6D'}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.1, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}}, {'type': 'ResizeImages', 'height': 256, 'width': 256}, {'type': 'SimpleNormalizeImages'}, {'type': 'NormalizeStatesAndActions', 'action_dim': 64, 'state_dim': 64, 'state_key': 'proprio', 'action_key': 'action', 'state_norm_type': 'none', 'action_norm_type': 'quantile'}, {'type': 'BuildCosmos3Sequence', 'raw_action_dim': 10, 'mode': 'wam', 'frame_window_size': 17, 'prepend_state_to_action': False, 'conditioning_fps': 20.0}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 17, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}], 'action_window_size': 16, 'action_key': 'action', 'use_delta': False, 'statistic_name': 'all_libero_no_noops', 'window_start_idx': 0, 'frame_window_size': 17, 'require_full_window': True}, {'type': 'ParquetDataset', 'data_root_path': './datasets/libero_goal_no_noops_lerobotv2.1', 'transforms': [{'type': 'ProcessParquetInputs', 'parquet_keys': ['observation.state', 'timestamp', 'actions', 'info', 'stats', 'action_masks'], 'video_keys': ['observation.images.image', 'observation.images.wrist_image'], 'name_mappings': {'observation.state': ['states'], 'actions': ['actions']}, 'embodiment_id': 5}, {'type': 'LiberoFramewiseActionToRot6D'}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.1, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}}, {'type': 'ResizeImages', 'height': 256, 'width': 256}, {'type': 'SimpleNormalizeImages'}, {'type': 'NormalizeStatesAndActions', 'action_dim': 64, 'state_dim': 64, 'state_key': 'proprio', 'action_key': 'action', 'state_norm_type': 'none', 'action_norm_type': 'quantile'}, {'type': 'BuildCosmos3Sequence', 'raw_action_dim': 10, 'mode': 'wam', 'frame_window_size': 17, 'prepend_state_to_action': False, 'conditioning_fps': 20.0}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 17, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}], 'action_window_size': 16, 'action_key': 'action', 'use_delta': False, 'statistic_name': 'all_libero_no_noops', 'window_start_idx': 0, 'frame_window_size': 17, 'require_full_window': True}, {'type': 'ParquetDataset', 'data_root_path': './datasets/libero_10_no_noops_lerobotv2.1', 'transforms': [{'type': 'ProcessParquetInputs', 'parquet_keys': ['observation.state', 'timestamp', 'actions', 'info', 'stats', 'action_masks'], 'video_keys': ['observation.images.image', 'observation.images.wrist_image'], 'name_mappings': {'observation.state': ['states'], 'actions': ['actions']}, 'embodiment_id': 5}, {'type': 'LiberoFramewiseActionToRot6D'}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.1, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}}, {'type': 'ResizeImages', 'height': 256, 'width': 256}, {'type': 'SimpleNormalizeImages'}, {'type': 'NormalizeStatesAndActions', 'action_dim': 64, 'state_dim': 64, 'state_key': 'proprio', 'action_key': 'action', 'state_norm_type': 'none', 'action_norm_type': 'quantile'}, {'type': 'BuildCosmos3Sequence', 'raw_action_dim': 10, 'mode': 'wam', 'frame_window_size': 17, 'prepend_state_to_action': False, 'conditioning_fps': 20.0}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 17, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}], 'action_window_size': 16, 'action_key': 'action', 'use_delta': False, 'statistic_name': 'all_libero_no_noops', 'window_start_idx': 0, 'frame_window_size': 17, 'require_full_window': True}]}}, 'runner': {'type': 'FSDPTrainRunner', 'max_steps': 5000, 'save_iter_interval': 500, 'max_keep_ckpts': 2, 'optimizer': {'lr': 5e-05, 'type': 'AdamW', 'weight_decay': 0.05, 'betas': (0.9, 0.99), 'eps': 1e-08, 'fused': True, 'exclude_1d_from_weight_decay': False, 'paramwise_learning_rate': {'action_in_proj.': 0.00025, 'action_out_proj.': 0.00025, 'action_modality_embed': 0.00025}}, 'max_grad_norm': 1.0, 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'collator': {'type': 'Cosmos3Collator', 'tensor_keys': ['images', 'actions', 'embodiment_ids', 'raw_action_dim', 'conditioning_fps', 'action_fps'], 'sequence_keys': ['text_token_ids'], 'list_keys': ['sequence_plan'], 'meta_keys': ['task_description', 'stats', 'info', 'timestamp'], 'pad_id': 0}, 'sampler': None, 'grad_accumulation_steps': 8, 'metric': {'type': 'VLAMetric', 'active_trackers': ('jsonl', 'wandb'), 'run_dir': '/mnt/data/cpfs/users/jikun/wk_dir/cosmos3/libero_align_16_9_padding', 'grad_accumulation_steps': 8, 'window_size': 1}, 'lr_scheduler': {'type': 'linear-warmup+linear-decay', 'warmup_steps': 500, 'cycle_length': 16000}, 'enable_gradient_checkpointing': True, 'enable_mixed_precision_training': True, 'mixed_precision_dtype': 'bf16', 'sharding_strategy': 'full-shard', 'change_key_name': False, 'cfg': Config (path: configs/cosmos3/cosmos3nano_libero_full_finetune.py): {...}, 'args': Namespace(config='configs/cosmos3/cosmos3nano_libero_full_finetune.py', work_dir='/mnt/data/cpfs/users/jikun/wk_dir/cosmos3/libero_align_16_9_padding', cfg_options=None, eval_after_train=True, resume_from=None)}, 'eval': {'type': 'LiberoEvalRunner', 'task_suite_name': 'libero_10', 'model_family': 'cosmos3', 'norm_stats_key': 'all_libero_no_noops', 'eval_chunk_size': 16, 'num_trials_per_task': 50, 'num_steps_wait': 10, 'seed': 7, 'enable_mixed_precision_training': True, 'mixed_precision_dtype': 'bf16', 'dataset': {'type': 'LiberoParquetEvalDataset', 'img_buffer_len': 1, 'extra_tensor_keys': ['conditioning_fps', 'prepend_state_to_action'], 'transforms': [{'type': 'ProcessLiberoEvalInputs', 'img_keys': ['agentview_image', 'robot0_eye_in_hand_image'], 'resize_size': 256, 'embodiment_id': 5}, {'type': 'SetCosmos3ActionMetadata', 'conditioning_fps': 20.0, 'prepend_state_to_action': False}, {'type': 'TransformImage', 'image_resize_strategy': 'resize-crop', 'input_sizes': [[3, 256, 256], [3, 256, 256]], 'means': [[127.5, 127.5, 127.5], [127.5, 127.5, 127.5]], 'stds': [[127.5, 127.5, 127.5], [127.5, 127.5, 127.5]]}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.0, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}, 'output_key': 'lang_tokens', 'output_attention_mask_key': 'lang_masks'}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 1, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}]}, 'denormalize_action': {'type': 'DenormalizeLiberoFramewiseRot6DAction', 'norm_type': 'quantile_rot', 'action_dim': 10}}}, 'args': Namespace(config='configs/cosmos3/cosmos3nano_libero_full_finetune.py', work_dir='/mnt/data/cpfs/users/jikun/wk_dir/cosmos3/libero_align_16_9_padding', cfg_options=None, eval_after_train=True, resume_from=None)}, 'eval': {'type': 'LiberoEvalRunner', 'task_suite_name': 'libero_10', 'model_family': 'cosmos3', 'norm_stats_key': 'all_libero_no_noops', 'eval_chunk_size': 16, 'num_trials_per_task': 50, 'num_steps_wait': 10, 'seed': 7, 'enable_mixed_precision_training': True, 'mixed_precision_dtype': 'bf16', 'dataset': {'type': 'LiberoParquetEvalDataset', 'img_buffer_len': 1, 'extra_tensor_keys': ['conditioning_fps', 'prepend_state_to_action'], 'transforms': [{'type': 'ProcessLiberoEvalInputs', 'img_keys': ['agentview_image', 'robot0_eye_in_hand_image'], 'resize_size': 256, 'embodiment_id': 5}, {'type': 'SetCosmos3ActionMetadata', 'conditioning_fps': 20.0, 'prepend_state_to_action': False}, {'type': 'TransformImage', 'image_resize_strategy': 'resize-crop', 'input_sizes': [[3, 256, 256], [3, 256, 256]], 'means': [[127.5, 127.5, 127.5], [127.5, 127.5, 127.5]], 'stds': [[127.5, 127.5, 127.5], [127.5, 127.5, 127.5]]}, {'type': 'ProcessCosmos3Prompt', 'tokenizer': {'type': 'PretrainedTokenizer', 'model_path': './checkpoints/Cosmos3-Nano/text_tokenizer', 'model_max_length': 4096, 'padding_side': 'right', 'trust_remote_code': True}, 'max_len': 512, 'cfg_dropout_rate': 0.0, 'format_prompt_as_json': True, 'action_metadata': {'append_viewpoint': False, 'frame_window_size': 17, 'conditioning_fps': 20.0, 'video_height': 192, 'video_width': 320}, 'output_key': 'lang_tokens', 'output_attention_mask_key': 'lang_masks'}, {'type': 'PrepareVideo', 'num_views': 2, 'frame_window_size': 1, 'tile_direction': 'horizontal'}, {'type': 'ResizeAndReflectPad', 'height': 192, 'width': 320}]}, 'denormalize_action': {'type': 'DenormalizeLiberoFramewiseRot6DAction', 'norm_type': 'quantile_rot', 'action_dim': 10}}}", "run_id": "cosmos3nano_libero_full_finetune_2026_08_14_18_30_47"}
|
cosmos3_nano_libero_full_finetune_bs2048/tokenizer/added_tokens.json
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"</think>": 151668,
|
| 3 |
+
"</tool_call>": 151658,
|
| 4 |
+
"</tool_response>": 151666,
|
| 5 |
+
"<think>": 151667,
|
| 6 |
+
"<tool_call>": 151657,
|
| 7 |
+
"<tool_response>": 151665,
|
| 8 |
+
"<|box_end|>": 151649,
|
| 9 |
+
"<|box_start|>": 151648,
|
| 10 |
+
"<|endoftext|>": 151643,
|
| 11 |
+
"<|file_sep|>": 151664,
|
| 12 |
+
"<|fim_middle|>": 151660,
|
| 13 |
+
"<|fim_pad|>": 151662,
|
| 14 |
+
"<|fim_prefix|>": 151659,
|
| 15 |
+
"<|fim_suffix|>": 151661,
|
| 16 |
+
"<|im_end|>": 151645,
|
| 17 |
+
"<|im_start|>": 151644,
|
| 18 |
+
"<|image_pad|>": 151655,
|
| 19 |
+
"<|object_ref_end|>": 151647,
|
| 20 |
+
"<|object_ref_start|>": 151646,
|
| 21 |
+
"<|quad_end|>": 151651,
|
| 22 |
+
"<|quad_start|>": 151650,
|
| 23 |
+
"<|repo_name|>": 151663,
|
| 24 |
+
"<|video_pad|>": 151656,
|
| 25 |
+
"<|vision_end|>": 151653,
|
| 26 |
+
"<|vision_pad|>": 151654,
|
| 27 |
+
"<|vision_start|>": 151652
|
| 28 |
+
}
|
cosmos3_nano_libero_full_finetune_bs2048/tokenizer/chat_template.jinja
ADDED
|
@@ -0,0 +1,120 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{%- if tools %}
|
| 2 |
+
{{- '<|im_start|>system\n' }}
|
| 3 |
+
{%- if messages[0].role == 'system' %}
|
| 4 |
+
{%- if messages[0].content is string %}
|
| 5 |
+
{{- messages[0].content }}
|
| 6 |
+
{%- else %}
|
| 7 |
+
{%- for content in messages[0].content %}
|
| 8 |
+
{%- if 'text' in content %}
|
| 9 |
+
{{- content.text }}
|
| 10 |
+
{%- endif %}
|
| 11 |
+
{%- endfor %}
|
| 12 |
+
{%- endif %}
|
| 13 |
+
{{- '\n\n' }}
|
| 14 |
+
{%- endif %}
|
| 15 |
+
{{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
|
| 16 |
+
{%- for tool in tools %}
|
| 17 |
+
{{- "\n" }}
|
| 18 |
+
{{- tool | tojson }}
|
| 19 |
+
{%- endfor %}
|
| 20 |
+
{{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
|
| 21 |
+
{%- else %}
|
| 22 |
+
{%- if messages[0].role == 'system' %}
|
| 23 |
+
{{- '<|im_start|>system\n' }}
|
| 24 |
+
{%- if messages[0].content is string %}
|
| 25 |
+
{{- messages[0].content }}
|
| 26 |
+
{%- else %}
|
| 27 |
+
{%- for content in messages[0].content %}
|
| 28 |
+
{%- if 'text' in content %}
|
| 29 |
+
{{- content.text }}
|
| 30 |
+
{%- endif %}
|
| 31 |
+
{%- endfor %}
|
| 32 |
+
{%- endif %}
|
| 33 |
+
{{- '<|im_end|>\n' }}
|
| 34 |
+
{%- endif %}
|
| 35 |
+
{%- endif %}
|
| 36 |
+
{%- set image_count = namespace(value=0) %}
|
| 37 |
+
{%- set video_count = namespace(value=0) %}
|
| 38 |
+
{%- for message in messages %}
|
| 39 |
+
{%- if message.role == "user" %}
|
| 40 |
+
{{- '<|im_start|>' + message.role + '\n' }}
|
| 41 |
+
{%- if message.content is string %}
|
| 42 |
+
{{- message.content }}
|
| 43 |
+
{%- else %}
|
| 44 |
+
{%- for content in message.content %}
|
| 45 |
+
{%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
|
| 46 |
+
{%- set image_count.value = image_count.value + 1 %}
|
| 47 |
+
{%- if add_vision_id %}Picture {{ image_count.value }}: {% endif -%}
|
| 48 |
+
<|vision_start|><|image_pad|><|vision_end|>
|
| 49 |
+
{%- elif content.type == 'video' or 'video' in content %}
|
| 50 |
+
{%- set video_count.value = video_count.value + 1 %}
|
| 51 |
+
{%- if add_vision_id %}Video {{ video_count.value }}: {% endif -%}
|
| 52 |
+
<|vision_start|><|video_pad|><|vision_end|>
|
| 53 |
+
{%- elif 'text' in content %}
|
| 54 |
+
{{- content.text }}
|
| 55 |
+
{%- endif %}
|
| 56 |
+
{%- endfor %}
|
| 57 |
+
{%- endif %}
|
| 58 |
+
{{- '<|im_end|>\n' }}
|
| 59 |
+
{%- elif message.role == "assistant" %}
|
| 60 |
+
{{- '<|im_start|>' + message.role + '\n' }}
|
| 61 |
+
{%- if message.content is string %}
|
| 62 |
+
{{- message.content }}
|
| 63 |
+
{%- else %}
|
| 64 |
+
{%- for content_item in message.content %}
|
| 65 |
+
{%- if 'text' in content_item %}
|
| 66 |
+
{{- content_item.text }}
|
| 67 |
+
{%- endif %}
|
| 68 |
+
{%- endfor %}
|
| 69 |
+
{%- endif %}
|
| 70 |
+
{%- if message.tool_calls %}
|
| 71 |
+
{%- for tool_call in message.tool_calls %}
|
| 72 |
+
{%- if (loop.first and message.content) or (not loop.first) %}
|
| 73 |
+
{{- '\n' }}
|
| 74 |
+
{%- endif %}
|
| 75 |
+
{%- if tool_call.function %}
|
| 76 |
+
{%- set tool_call = tool_call.function %}
|
| 77 |
+
{%- endif %}
|
| 78 |
+
{{- '<tool_call>\n{"name": "' }}
|
| 79 |
+
{{- tool_call.name }}
|
| 80 |
+
{{- '", "arguments": ' }}
|
| 81 |
+
{%- if tool_call.arguments is string %}
|
| 82 |
+
{{- tool_call.arguments }}
|
| 83 |
+
{%- else %}
|
| 84 |
+
{{- tool_call.arguments | tojson }}
|
| 85 |
+
{%- endif %}
|
| 86 |
+
{{- '}\n</tool_call>' }}
|
| 87 |
+
{%- endfor %}
|
| 88 |
+
{%- endif %}
|
| 89 |
+
{{- '<|im_end|>\n' }}
|
| 90 |
+
{%- elif message.role == "tool" %}
|
| 91 |
+
{%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
|
| 92 |
+
{{- '<|im_start|>user' }}
|
| 93 |
+
{%- endif %}
|
| 94 |
+
{{- '\n<tool_response>\n' }}
|
| 95 |
+
{%- if message.content is string %}
|
| 96 |
+
{{- message.content }}
|
| 97 |
+
{%- else %}
|
| 98 |
+
{%- for content in message.content %}
|
| 99 |
+
{%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
|
| 100 |
+
{%- set image_count.value = image_count.value + 1 %}
|
| 101 |
+
{%- if add_vision_id %}Picture {{ image_count.value }}: {% endif -%}
|
| 102 |
+
<|vision_start|><|image_pad|><|vision_end|>
|
| 103 |
+
{%- elif content.type == 'video' or 'video' in content %}
|
| 104 |
+
{%- set video_count.value = video_count.value + 1 %}
|
| 105 |
+
{%- if add_vision_id %}Video {{ video_count.value }}: {% endif -%}
|
| 106 |
+
<|vision_start|><|video_pad|><|vision_end|>
|
| 107 |
+
{%- elif 'text' in content %}
|
| 108 |
+
{{- content.text }}
|
| 109 |
+
{%- endif %}
|
| 110 |
+
{%- endfor %}
|
| 111 |
+
{%- endif %}
|
| 112 |
+
{{- '\n</tool_response>' }}
|
| 113 |
+
{%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
|
| 114 |
+
{{- '<|im_end|>\n' }}
|
| 115 |
+
{%- endif %}
|
| 116 |
+
{%- endif %}
|
| 117 |
+
{%- endfor %}
|
| 118 |
+
{%- if add_generation_prompt %}
|
| 119 |
+
{{- '<|im_start|>assistant\n' }}
|
| 120 |
+
{%- endif %}
|
cosmos3_nano_libero_full_finetune_bs2048/tokenizer/merges.txt
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
cosmos3_nano_libero_full_finetune_bs2048/tokenizer/special_tokens_map.json
ADDED
|
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"additional_special_tokens": [
|
| 3 |
+
"<|im_start|>",
|
| 4 |
+
"<|im_end|>",
|
| 5 |
+
"<|object_ref_start|>",
|
| 6 |
+
"<|object_ref_end|>",
|
| 7 |
+
"<|box_start|>",
|
| 8 |
+
"<|box_end|>",
|
| 9 |
+
"<|quad_start|>",
|
| 10 |
+
"<|quad_end|>",
|
| 11 |
+
"<|vision_start|>",
|
| 12 |
+
"<|vision_end|>",
|
| 13 |
+
"<|vision_pad|>",
|
| 14 |
+
"<|image_pad|>",
|
| 15 |
+
"<|video_pad|>"
|
| 16 |
+
],
|
| 17 |
+
"eos_token": {
|
| 18 |
+
"content": "<|im_end|>",
|
| 19 |
+
"lstrip": false,
|
| 20 |
+
"normalized": false,
|
| 21 |
+
"rstrip": false,
|
| 22 |
+
"single_word": false
|
| 23 |
+
},
|
| 24 |
+
"pad_token": {
|
| 25 |
+
"content": "<|endoftext|>",
|
| 26 |
+
"lstrip": false,
|
| 27 |
+
"normalized": false,
|
| 28 |
+
"rstrip": false,
|
| 29 |
+
"single_word": false
|
| 30 |
+
}
|
| 31 |
+
}
|
cosmos3_nano_libero_full_finetune_bs2048/tokenizer/tokenizer.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:aeb13307a71acd8fe81861d94ad54ab689df773318809eed3cbe794b4492dae4
|
| 3 |
+
size 11422654
|
cosmos3_nano_libero_full_finetune_bs2048/tokenizer/tokenizer_config.json
ADDED
|
@@ -0,0 +1,239 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_bos_token": false,
|
| 3 |
+
"add_prefix_space": false,
|
| 4 |
+
"added_tokens_decoder": {
|
| 5 |
+
"151643": {
|
| 6 |
+
"content": "<|endoftext|>",
|
| 7 |
+
"lstrip": false,
|
| 8 |
+
"normalized": false,
|
| 9 |
+
"rstrip": false,
|
| 10 |
+
"single_word": false,
|
| 11 |
+
"special": true
|
| 12 |
+
},
|
| 13 |
+
"151644": {
|
| 14 |
+
"content": "<|im_start|>",
|
| 15 |
+
"lstrip": false,
|
| 16 |
+
"normalized": false,
|
| 17 |
+
"rstrip": false,
|
| 18 |
+
"single_word": false,
|
| 19 |
+
"special": true
|
| 20 |
+
},
|
| 21 |
+
"151645": {
|
| 22 |
+
"content": "<|im_end|>",
|
| 23 |
+
"lstrip": false,
|
| 24 |
+
"normalized": false,
|
| 25 |
+
"rstrip": false,
|
| 26 |
+
"single_word": false,
|
| 27 |
+
"special": true
|
| 28 |
+
},
|
| 29 |
+
"151646": {
|
| 30 |
+
"content": "<|object_ref_start|>",
|
| 31 |
+
"lstrip": false,
|
| 32 |
+
"normalized": false,
|
| 33 |
+
"rstrip": false,
|
| 34 |
+
"single_word": false,
|
| 35 |
+
"special": true
|
| 36 |
+
},
|
| 37 |
+
"151647": {
|
| 38 |
+
"content": "<|object_ref_end|>",
|
| 39 |
+
"lstrip": false,
|
| 40 |
+
"normalized": false,
|
| 41 |
+
"rstrip": false,
|
| 42 |
+
"single_word": false,
|
| 43 |
+
"special": true
|
| 44 |
+
},
|
| 45 |
+
"151648": {
|
| 46 |
+
"content": "<|box_start|>",
|
| 47 |
+
"lstrip": false,
|
| 48 |
+
"normalized": false,
|
| 49 |
+
"rstrip": false,
|
| 50 |
+
"single_word": false,
|
| 51 |
+
"special": true
|
| 52 |
+
},
|
| 53 |
+
"151649": {
|
| 54 |
+
"content": "<|box_end|>",
|
| 55 |
+
"lstrip": false,
|
| 56 |
+
"normalized": false,
|
| 57 |
+
"rstrip": false,
|
| 58 |
+
"single_word": false,
|
| 59 |
+
"special": true
|
| 60 |
+
},
|
| 61 |
+
"151650": {
|
| 62 |
+
"content": "<|quad_start|>",
|
| 63 |
+
"lstrip": false,
|
| 64 |
+
"normalized": false,
|
| 65 |
+
"rstrip": false,
|
| 66 |
+
"single_word": false,
|
| 67 |
+
"special": true
|
| 68 |
+
},
|
| 69 |
+
"151651": {
|
| 70 |
+
"content": "<|quad_end|>",
|
| 71 |
+
"lstrip": false,
|
| 72 |
+
"normalized": false,
|
| 73 |
+
"rstrip": false,
|
| 74 |
+
"single_word": false,
|
| 75 |
+
"special": true
|
| 76 |
+
},
|
| 77 |
+
"151652": {
|
| 78 |
+
"content": "<|vision_start|>",
|
| 79 |
+
"lstrip": false,
|
| 80 |
+
"normalized": false,
|
| 81 |
+
"rstrip": false,
|
| 82 |
+
"single_word": false,
|
| 83 |
+
"special": true
|
| 84 |
+
},
|
| 85 |
+
"151653": {
|
| 86 |
+
"content": "<|vision_end|>",
|
| 87 |
+
"lstrip": false,
|
| 88 |
+
"normalized": false,
|
| 89 |
+
"rstrip": false,
|
| 90 |
+
"single_word": false,
|
| 91 |
+
"special": true
|
| 92 |
+
},
|
| 93 |
+
"151654": {
|
| 94 |
+
"content": "<|vision_pad|>",
|
| 95 |
+
"lstrip": false,
|
| 96 |
+
"normalized": false,
|
| 97 |
+
"rstrip": false,
|
| 98 |
+
"single_word": false,
|
| 99 |
+
"special": true
|
| 100 |
+
},
|
| 101 |
+
"151655": {
|
| 102 |
+
"content": "<|image_pad|>",
|
| 103 |
+
"lstrip": false,
|
| 104 |
+
"normalized": false,
|
| 105 |
+
"rstrip": false,
|
| 106 |
+
"single_word": false,
|
| 107 |
+
"special": true
|
| 108 |
+
},
|
| 109 |
+
"151656": {
|
| 110 |
+
"content": "<|video_pad|>",
|
| 111 |
+
"lstrip": false,
|
| 112 |
+
"normalized": false,
|
| 113 |
+
"rstrip": false,
|
| 114 |
+
"single_word": false,
|
| 115 |
+
"special": true
|
| 116 |
+
},
|
| 117 |
+
"151657": {
|
| 118 |
+
"content": "<tool_call>",
|
| 119 |
+
"lstrip": false,
|
| 120 |
+
"normalized": false,
|
| 121 |
+
"rstrip": false,
|
| 122 |
+
"single_word": false,
|
| 123 |
+
"special": false
|
| 124 |
+
},
|
| 125 |
+
"151658": {
|
| 126 |
+
"content": "</tool_call>",
|
| 127 |
+
"lstrip": false,
|
| 128 |
+
"normalized": false,
|
| 129 |
+
"rstrip": false,
|
| 130 |
+
"single_word": false,
|
| 131 |
+
"special": false
|
| 132 |
+
},
|
| 133 |
+
"151659": {
|
| 134 |
+
"content": "<|fim_prefix|>",
|
| 135 |
+
"lstrip": false,
|
| 136 |
+
"normalized": false,
|
| 137 |
+
"rstrip": false,
|
| 138 |
+
"single_word": false,
|
| 139 |
+
"special": false
|
| 140 |
+
},
|
| 141 |
+
"151660": {
|
| 142 |
+
"content": "<|fim_middle|>",
|
| 143 |
+
"lstrip": false,
|
| 144 |
+
"normalized": false,
|
| 145 |
+
"rstrip": false,
|
| 146 |
+
"single_word": false,
|
| 147 |
+
"special": false
|
| 148 |
+
},
|
| 149 |
+
"151661": {
|
| 150 |
+
"content": "<|fim_suffix|>",
|
| 151 |
+
"lstrip": false,
|
| 152 |
+
"normalized": false,
|
| 153 |
+
"rstrip": false,
|
| 154 |
+
"single_word": false,
|
| 155 |
+
"special": false
|
| 156 |
+
},
|
| 157 |
+
"151662": {
|
| 158 |
+
"content": "<|fim_pad|>",
|
| 159 |
+
"lstrip": false,
|
| 160 |
+
"normalized": false,
|
| 161 |
+
"rstrip": false,
|
| 162 |
+
"single_word": false,
|
| 163 |
+
"special": false
|
| 164 |
+
},
|
| 165 |
+
"151663": {
|
| 166 |
+
"content": "<|repo_name|>",
|
| 167 |
+
"lstrip": false,
|
| 168 |
+
"normalized": false,
|
| 169 |
+
"rstrip": false,
|
| 170 |
+
"single_word": false,
|
| 171 |
+
"special": false
|
| 172 |
+
},
|
| 173 |
+
"151664": {
|
| 174 |
+
"content": "<|file_sep|>",
|
| 175 |
+
"lstrip": false,
|
| 176 |
+
"normalized": false,
|
| 177 |
+
"rstrip": false,
|
| 178 |
+
"single_word": false,
|
| 179 |
+
"special": false
|
| 180 |
+
},
|
| 181 |
+
"151665": {
|
| 182 |
+
"content": "<tool_response>",
|
| 183 |
+
"lstrip": false,
|
| 184 |
+
"normalized": false,
|
| 185 |
+
"rstrip": false,
|
| 186 |
+
"single_word": false,
|
| 187 |
+
"special": false
|
| 188 |
+
},
|
| 189 |
+
"151666": {
|
| 190 |
+
"content": "</tool_response>",
|
| 191 |
+
"lstrip": false,
|
| 192 |
+
"normalized": false,
|
| 193 |
+
"rstrip": false,
|
| 194 |
+
"single_word": false,
|
| 195 |
+
"special": false
|
| 196 |
+
},
|
| 197 |
+
"151667": {
|
| 198 |
+
"content": "<think>",
|
| 199 |
+
"lstrip": false,
|
| 200 |
+
"normalized": false,
|
| 201 |
+
"rstrip": false,
|
| 202 |
+
"single_word": false,
|
| 203 |
+
"special": false
|
| 204 |
+
},
|
| 205 |
+
"151668": {
|
| 206 |
+
"content": "</think>",
|
| 207 |
+
"lstrip": false,
|
| 208 |
+
"normalized": false,
|
| 209 |
+
"rstrip": false,
|
| 210 |
+
"single_word": false,
|
| 211 |
+
"special": false
|
| 212 |
+
}
|
| 213 |
+
},
|
| 214 |
+
"additional_special_tokens": [
|
| 215 |
+
"<|im_start|>",
|
| 216 |
+
"<|im_end|>",
|
| 217 |
+
"<|object_ref_start|>",
|
| 218 |
+
"<|object_ref_end|>",
|
| 219 |
+
"<|box_start|>",
|
| 220 |
+
"<|box_end|>",
|
| 221 |
+
"<|quad_start|>",
|
| 222 |
+
"<|quad_end|>",
|
| 223 |
+
"<|vision_start|>",
|
| 224 |
+
"<|vision_end|>",
|
| 225 |
+
"<|vision_pad|>",
|
| 226 |
+
"<|image_pad|>",
|
| 227 |
+
"<|video_pad|>"
|
| 228 |
+
],
|
| 229 |
+
"bos_token": null,
|
| 230 |
+
"clean_up_tokenization_spaces": false,
|
| 231 |
+
"eos_token": "<|im_end|>",
|
| 232 |
+
"errors": "replace",
|
| 233 |
+
"extra_special_tokens": {},
|
| 234 |
+
"model_max_length": 262144,
|
| 235 |
+
"pad_token": "<|endoftext|>",
|
| 236 |
+
"split_special_tokens": false,
|
| 237 |
+
"tokenizer_class": "Qwen2Tokenizer",
|
| 238 |
+
"unk_token": null
|
| 239 |
+
}
|
cosmos3_nano_libero_full_finetune_bs2048/tokenizer/vocab.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
cosmos3_nano_libero_full_finetune_bs2048/vlm_backbone_config.json
ADDED
|
@@ -0,0 +1,100 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"image_token_id": 151655,
|
| 3 |
+
"model_type": "qwen3_vl",
|
| 4 |
+
"text_config": {
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"bos_token_id": 151643,
|
| 8 |
+
"eos_token_id": 151645,
|
| 9 |
+
"head_dim": 128,
|
| 10 |
+
"hidden_act": "silu",
|
| 11 |
+
"hidden_size": 4096,
|
| 12 |
+
"initializer_range": 0.02,
|
| 13 |
+
"intermediate_size": 12288,
|
| 14 |
+
"layer_types": [
|
| 15 |
+
"full_attention",
|
| 16 |
+
"full_attention",
|
| 17 |
+
"full_attention",
|
| 18 |
+
"full_attention",
|
| 19 |
+
"full_attention",
|
| 20 |
+
"full_attention",
|
| 21 |
+
"full_attention",
|
| 22 |
+
"full_attention",
|
| 23 |
+
"full_attention",
|
| 24 |
+
"full_attention",
|
| 25 |
+
"full_attention",
|
| 26 |
+
"full_attention",
|
| 27 |
+
"full_attention",
|
| 28 |
+
"full_attention",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"full_attention",
|
| 31 |
+
"full_attention",
|
| 32 |
+
"full_attention",
|
| 33 |
+
"full_attention",
|
| 34 |
+
"full_attention",
|
| 35 |
+
"full_attention",
|
| 36 |
+
"full_attention",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"full_attention",
|
| 39 |
+
"full_attention",
|
| 40 |
+
"full_attention",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"full_attention",
|
| 43 |
+
"full_attention",
|
| 44 |
+
"full_attention",
|
| 45 |
+
"full_attention",
|
| 46 |
+
"full_attention",
|
| 47 |
+
"full_attention",
|
| 48 |
+
"full_attention",
|
| 49 |
+
"full_attention",
|
| 50 |
+
"full_attention"
|
| 51 |
+
],
|
| 52 |
+
"max_position_embeddings": 262144,
|
| 53 |
+
"model_type": "qwen3_vl_text",
|
| 54 |
+
"num_attention_heads": 32,
|
| 55 |
+
"num_hidden_layers": 36,
|
| 56 |
+
"num_key_value_heads": 8,
|
| 57 |
+
"packed_attention_backend": "flash2",
|
| 58 |
+
"pad_token_id": 0,
|
| 59 |
+
"rms_norm_eps": 1e-06,
|
| 60 |
+
"rope_parameters": {
|
| 61 |
+
"mrope_interleaved": true,
|
| 62 |
+
"mrope_section": [
|
| 63 |
+
24,
|
| 64 |
+
20,
|
| 65 |
+
20
|
| 66 |
+
],
|
| 67 |
+
"rope_theta": 5000000,
|
| 68 |
+
"rope_type": "default"
|
| 69 |
+
},
|
| 70 |
+
"tie_word_embeddings": false,
|
| 71 |
+
"use_cache": true,
|
| 72 |
+
"vocab_size": 151936
|
| 73 |
+
},
|
| 74 |
+
"tie_word_embeddings": false,
|
| 75 |
+
"transformers_version": "5.3.0",
|
| 76 |
+
"video_token_id": 151656,
|
| 77 |
+
"vision_config": {
|
| 78 |
+
"deepstack_visual_indexes": [
|
| 79 |
+
8,
|
| 80 |
+
16,
|
| 81 |
+
24
|
| 82 |
+
],
|
| 83 |
+
"depth": 27,
|
| 84 |
+
"hidden_act": "gelu_pytorch_tanh",
|
| 85 |
+
"hidden_size": 1152,
|
| 86 |
+
"in_channels": 3,
|
| 87 |
+
"initializer_range": 0.02,
|
| 88 |
+
"intermediate_size": 4304,
|
| 89 |
+
"model_type": "qwen3_vl",
|
| 90 |
+
"num_heads": 16,
|
| 91 |
+
"num_position_embeddings": 2304,
|
| 92 |
+
"out_hidden_size": 4096,
|
| 93 |
+
"patch_size": 16,
|
| 94 |
+
"spatial_merge_size": 2,
|
| 95 |
+
"temporal_patch_size": 2
|
| 96 |
+
},
|
| 97 |
+
"vision_end_token_id": 151653,
|
| 98 |
+
"vision_start_token_id": 151652,
|
| 99 |
+
"vocab_size": 151936
|
| 100 |
+
}
|