Whalswp commited on
Commit
36d8c2a
·
verified ·
1 Parent(s): 2ecdf76

Add files using upload-large-folder tool

Browse files
Files changed (49) hide show
  1. .gitattributes +2 -0
  2. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.3/phase2/model-00001-of-00002.safetensors +3 -0
  3. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.3/phase2/model-00002-of-00002.safetensors +3 -0
  4. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.3/phase2/training_args.bin +3 -0
  5. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.3/phase3/model-00001-of-00002.safetensors +3 -0
  6. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.3/phase3/model-00002-of-00002.safetensors +3 -0
  7. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.3/phase3/training_args.bin +3 -0
  8. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.5/phase2/model-00001-of-00002.safetensors +3 -0
  9. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.5/phase2/model-00002-of-00002.safetensors +3 -0
  10. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.5/phase2/training_args.bin +3 -0
  11. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.5/phase3/model-00001-of-00002.safetensors +3 -0
  12. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.5/phase3/model-00002-of-00002.safetensors +3 -0
  13. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.5/phase3/training_args.bin +3 -0
  14. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.7/phase2/model-00001-of-00002.safetensors +3 -0
  15. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.7/phase2/model-00002-of-00002.safetensors +3 -0
  16. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.7/phase2/training_args.bin +3 -0
  17. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.7/phase3/model-00001-of-00002.safetensors +3 -0
  18. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.7/phase3/model-00002-of-00002.safetensors +3 -0
  19. VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.7/phase3/training_args.bin +3 -0
  20. VLM_Only/MGD_Raw_VLM/train.log +3 -0
  21. VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase2/config.json +70 -0
  22. VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase2/experiment_cfg/metadata.json +431 -0
  23. VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase2/model-00001-of-00002.safetensors +3 -0
  24. VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase2/model-00002-of-00002.safetensors +3 -0
  25. VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase2/training_args.bin +3 -0
  26. VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase3/config.json +70 -0
  27. VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase3/model-00001-of-00002.safetensors +3 -0
  28. VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase3/model-00002-of-00002.safetensors +3 -0
  29. VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase3/model.safetensors.index.json +0 -0
  30. VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase3/trainer_state.json +0 -0
  31. VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase3/training_args.bin +3 -0
  32. VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/resolved_config.yaml +50 -0
  33. VLM_Only/RKD_Raw_VLM/rkd_temp_0.1/phase2/model-00001-of-00002.safetensors +3 -0
  34. VLM_Only/RKD_Raw_VLM/rkd_temp_0.1/phase2/model-00002-of-00002.safetensors +3 -0
  35. VLM_Only/RKD_Raw_VLM/rkd_temp_0.1/phase2/training_args.bin +3 -0
  36. VLM_Only/RKD_Raw_VLM/rkd_temp_0.1/phase3/model-00001-of-00002.safetensors +3 -0
  37. VLM_Only/RKD_Raw_VLM/rkd_temp_0.1/phase3/model-00002-of-00002.safetensors +3 -0
  38. VLM_Only/RKD_Raw_VLM/rkd_temp_0.1/phase3/training_args.bin +3 -0
  39. VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase2/config.json +70 -0
  40. VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase2/model-00001-of-00002.safetensors +3 -0
  41. VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase2/model-00002-of-00002.safetensors +3 -0
  42. VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase2/model.safetensors.index.json +0 -0
  43. VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase2/training_args.bin +3 -0
  44. VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase3/config.json +70 -0
  45. VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase3/model-00001-of-00002.safetensors +3 -0
  46. VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase3/model-00002-of-00002.safetensors +3 -0
  47. VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase3/model.safetensors.index.json +0 -0
  48. VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase3/training_args.bin +3 -0
  49. VLM_Only/RKD_Raw_VLM/train.log +3 -0
.gitattributes CHANGED
@@ -34,3 +34,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  RKD_Processing_line_Only/train.log filter=lfs diff=lfs merge=lfs -text
 
 
 
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  RKD_Processing_line_Only/train.log filter=lfs diff=lfs merge=lfs -text
37
+ VLM_Only/RKD_Raw_VLM/train.log filter=lfs diff=lfs merge=lfs -text
38
+ VLM_Only/MGD_Raw_VLM/train.log filter=lfs diff=lfs merge=lfs -text
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.3/phase2/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6544d9cacbb212108427cd6bb3903606146fc09123c30fed0f22883ffdff6e39
3
+ size 4999367032
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.3/phase2/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a432c78c936157e1ed50aff0999948296bcd8177accc07cb9f0f3aa0bb7e5f40
3
+ size 2643339748
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.3/phase2/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:71ded126380adf64cfba6621c3695daed80f91f626224ec15e12d3e945508e38
3
+ size 5905
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.3/phase3/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c4c74d353785ca1970b1b8e2f0b9a9c9364614a926871a87b9e1b5bdeee6861
3
+ size 4999367032
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.3/phase3/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:38b9afdf103b44a45538fc975fec6c871cb831f3054d0324e9564389d6386f02
3
+ size 2643339748
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.3/phase3/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:04c6e48c1bc19c5a4ee644acc82a98584aa6d7e02b7f006f0b6a8df3211bbf1f
3
+ size 5905
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.5/phase2/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e3807d9ade847ae77bd20bd82f566028856cce39721c9ba8795f9b5a8c483683
3
+ size 4999367032
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.5/phase2/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6df31710e97c6910db12b8d9cdc25a8ca027404487289f68fd1a992c5b765e58
3
+ size 2643339748
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.5/phase2/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a05f64f78cf87da8e10d41e8d51cf70297e8b54424cef3de5eafc89a28c80cda
3
+ size 5905
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.5/phase3/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:27b10894cc3d87df8e4655eed1ac5c84f6a4f070e3c9eda51b0db10f27821b21
3
+ size 4999367032
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.5/phase3/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe6b655423829a484c7e709636a257ab6a2a7f8189d842af92f6614c7b062bb7
3
+ size 2643339748
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.5/phase3/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e71f7524a1a8171fedc4b96e117af36f48808151536420395eab7451ece6979
3
+ size 5905
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.7/phase2/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:43f4f0e7d58a77f6dc39de538403413f8edc713aa2173a65c8dbf10e82465aaf
3
+ size 4999367032
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.7/phase2/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61e3fc3cf8905b9a81df6961aa2a310d651de38af8ddff3dbefe267f8e7eec6e
3
+ size 2643339748
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.7/phase2/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9cfdf69797bc5e7b88bb3de827b6e742d52c337ccc3eeeda86086bc866463c62
3
+ size 5905
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.7/phase3/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:71ba74bb250dc56ee7123e62d3505a396bd296db5408a301629aea52ccc489d6
3
+ size 4999367032
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.7/phase3/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2afd1818864fcb053fc29963a495bc99cb8c982a403b57d4c2e3cb813d895795
3
+ size 2643339748
VLM_Only/MGD_Raw_VLM/mgd_token_mask_ratio_0.7/phase3/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e73d91b214b99910b985ecad6e179faffac133916b6f33fa49566db43b22c5f8
3
+ size 5905
VLM_Only/MGD_Raw_VLM/train.log ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c31179c1bde8d16630dc731a2a97f41d838a3b04853b63c8194b33c7ffb0a0f
3
+ size 32765352
VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase2/config.json ADDED
@@ -0,0 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5_RKD"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "rkd_action_temp": 0.05,
63
+ "rkd_enabled": true,
64
+ "rkd_fm_loss_weight": 0.0,
65
+ "rkd_loss_weight": 1.0,
66
+ "rkd_relation_mode": "token_pair_mean",
67
+ "rkd_vlm_temp": 0.05,
68
+ "torch_dtype": "bfloat16",
69
+ "transformers_version": "4.51.3"
70
+ }
VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase2/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.3139495849609375,
8
+ 0.4876587688922882,
9
+ 0.7196521759033203
10
+ ],
11
+ "min": [
12
+ -4.94293737411499,
13
+ -6.890198230743408,
14
+ 0.6996827125549316
15
+ ],
16
+ "mean": [
17
+ 2.749288365724937,
18
+ -2.0455216239909983,
19
+ 0.700532551020533
20
+ ],
21
+ "std": [
22
+ 1.6786295897046262,
23
+ 1.312897969434244,
24
+ 0.00133751758906258
25
+ ],
26
+ "q01": [
27
+ -1.4241907881148037,
28
+ -5.1716547935950885,
29
+ 0.6999670898768614
30
+ ],
31
+ "q99": [
32
+ 6.198111545831555,
33
+ -0.5873531334104489,
34
+ 0.7038833014242587
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.23084999587450247,
54
+ 0.6238431088581847
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6556404371644047,
60
+ 0.3572252105732042
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9999999999999999,
66
+ 1.7482354593273349e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9999999999999999,
72
+ 0.9999999999999999
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.9014528393745422,
78
+ 0.8003877401351929,
79
+ 0.9829942584037781
80
+ ],
81
+ "min": [
82
+ -0.37471598386764526,
83
+ -0.8472502827644348,
84
+ -0.25070279836654663
85
+ ],
86
+ "mean": [
87
+ 0.28667332601863704,
88
+ -0.038315473170876704,
89
+ 0.45974658906777444
90
+ ],
91
+ "std": [
92
+ 0.17067058828305154,
93
+ 0.23569952037784195,
94
+ 0.21950702354181487
95
+ ],
96
+ "q01": [
97
+ 0.015172948963784228,
98
+ -0.44450502063220615,
99
+ 0.23314444103329338
100
+ ],
101
+ "q99": [
102
+ 0.5609069689572412,
103
+ 0.38454179128009547,
104
+ 0.7028102736473852
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999998807907104,
110
+ 0.9984455108642578,
111
+ 0.9434727430343628,
112
+ 0.9062229990959167
113
+ ],
114
+ "min": [
115
+ -0.9999930262565613,
116
+ -0.9988337755203247,
117
+ -0.9618149995803833,
118
+ 2.1872274658107926e-07
119
+ ],
120
+ "mean": [
121
+ -0.25672297401336,
122
+ 0.024425335806687216,
123
+ -0.08740977164483847,
124
+ 0.16608739644018397
125
+ ],
126
+ "std": [
127
+ 0.7932585208392383,
128
+ 0.3102260195580416,
129
+ 0.3772472576315148,
130
+ 0.17451959300158773
131
+ ],
132
+ "q01": [
133
+ -0.99379006987882,
134
+ -0.5478102050038828,
135
+ -0.6101777952971636,
136
+ 0.002274085014134748
137
+ ],
138
+ "q99": [
139
+ 0.8804856038896288,
140
+ 0.5910611092873037,
141
+ 0.5216058042993562,
142
+ 0.5231965974012255
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.055869169533252716,
148
+ 0.010916369967162609
149
+ ],
150
+ "min": [
151
+ -0.011436971835792065,
152
+ -0.05664053186774254
153
+ ],
154
+ "mean": [
155
+ 0.031651551516052555,
156
+ -0.03161481387920421
157
+ ],
158
+ "std": [
159
+ 0.013119769222217526,
160
+ 0.01306078225192402
161
+ ],
162
+ "q01": [
163
+ 0.0060999246446777336,
164
+ -0.04062601653964198
165
+ ],
166
+ "q99": [
167
+ 0.04054891621517283,
168
+ -0.0059163567113456345
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.008144896753718373,
188
+ -0.00018893135146078263,
189
+ -0.0008739062727221845,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11030045424364411,
194
+ 0.10148082570313594,
195
+ 0.08975698373467506,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.07287373067581518,
200
+ -0.09446994199569948,
201
+ -0.07702818259249572,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.04485485414407898,
206
+ 0.09626941605965177,
207
+ 0.07610665906122742,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9221974313425939
220
+ ],
221
+ "std": [
222
+ 0.38670194342612674
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.384693946883099
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ 0.01976129923017491,
244
+ -0.020870631344490034,
245
+ -0.05993476517349181
246
+ ],
247
+ "std": [
248
+ 0.4347024990804053,
249
+ 0.4221662976569997,
250
+ 0.3845506843044066
251
+ ],
252
+ "q01": [
253
+ -0.8835792939767807,
254
+ -0.9280919251713778,
255
+ -0.783361071470956
256
+ ],
257
+ "q99": [
258
+ 0.7622084438449307,
259
+ 0.8625125252609077,
260
+ 0.7470974753250706
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.008089673068544335,
276
+ -0.026498254627927348,
277
+ 0.0029083551743059005
278
+ ],
279
+ "std": [
280
+ 0.11216734503733773,
281
+ 0.13206002613798629,
282
+ 0.12615870412692864
283
+ ],
284
+ "q01": [
285
+ -0.2814627972846672,
286
+ -0.4342386516115439,
287
+ -0.34489755628340574
288
+ ],
289
+ "q99": [
290
+ 0.31879353266939386,
291
+ 0.28411797683220563,
292
+ 0.3597718671669894
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3824228874769045
304
+ ],
305
+ "std": [
306
+ 0.9240015096469786
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.5597272305813592
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase2/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8edce7d319943a95124d943f4016ce4f87158fd80a3585ed9a66a9156fa6fd2a
3
+ size 4999367032
VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase2/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c966ae82934e4f401e1138f03ea5ab851e33e7b9f0ac68c5e0765ae0dd1903a
3
+ size 2586705312
VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase2/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59f32c7b9ac91a9a0e3336413ba397a6c5d40640bf64e49b1867d2c49e81bb0f
3
+ size 5905
VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase3/config.json ADDED
@@ -0,0 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5_RKD"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "rkd_action_temp": 0.05,
63
+ "rkd_enabled": false,
64
+ "rkd_fm_loss_weight": 1.0,
65
+ "rkd_loss_weight": 0.0,
66
+ "rkd_relation_mode": "token_pair_mean",
67
+ "rkd_vlm_temp": 0.05,
68
+ "torch_dtype": "bfloat16",
69
+ "transformers_version": "4.51.3"
70
+ }
VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase3/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8b5f2bad465a2bc661c5adb37dcfa30e3175cec7ddf13e611c4dab3c034bd490
3
+ size 4999367032
VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase3/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a08b8400d196f00e4c5266699b7e5e91e497f06e523c76e6f375d915defd0f5
3
+ size 2586705312
VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase3/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase3/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/phase3/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a491612ce95fff9085b91605d0bf6e4ad94fc6c17177e5fe40bbd7746e47920
3
+ size 5905
VLM_Only/RKD_Raw_VLM/rkd_temp_0.05/resolved_config.yaml ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: rkd_temp_ablation_from_trained_base
2
+ policy_type: groot_rkd_raw
3
+ base_model_path: ${HOME}/groot_robocasa/Atomic26_baseline/checkpoint-60000
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/VLM_Only/RKD_Raw_VLM
5
+ sweep:
6
+ rkd_temp:
7
+ - 0.05
8
+ - 0.1
9
+ - 0.2
10
+ dataset:
11
+ dataset_soup: my_atomic26_human
12
+ training:
13
+ num_gpus: 2
14
+ batch_size: 32
15
+ model: null
16
+ phases:
17
+ - name: phase2_rkd_only
18
+ max_steps: 60000
19
+ save_steps: 0
20
+ trainable:
21
+ preset: freeze_processing_line
22
+ tune_llm: true
23
+ tune_visual: true
24
+ tune_projector: false
25
+ tune_diffusion_model: false
26
+ losses:
27
+ rkd_enabled: true
28
+ rkd_fm_loss_weight: 0.0
29
+ rkd_loss_weight: 1.0
30
+ rkd_action_temp: 0.05
31
+ rkd_vlm_temp: 0.05
32
+ rkd_relation_mode: token_pair_mean
33
+ - name: phase3_rkd_fm
34
+ max_steps: 30000
35
+ batch_size: 64
36
+ save_steps: 0
37
+ trainable:
38
+ tune_llm: false
39
+ tune_visual: false
40
+ tune_projector: true
41
+ tune_diffusion_model: true
42
+ losses:
43
+ rkd_enabled: false
44
+ rkd_fm_loss_weight: 1.0
45
+ rkd_loss_weight: 0.0
46
+ rkd_action_temp: 0.05
47
+ rkd_vlm_temp: 0.05
48
+ rkd_relation_mode: token_pair_mean
49
+ resolved_sweep:
50
+ rkd_temp: 0.05
VLM_Only/RKD_Raw_VLM/rkd_temp_0.1/phase2/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d725066559ed665a547d4255088c1442ffd5fd750b5ab0d65986b698915e63c6
3
+ size 4999367032
VLM_Only/RKD_Raw_VLM/rkd_temp_0.1/phase2/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c966ae82934e4f401e1138f03ea5ab851e33e7b9f0ac68c5e0765ae0dd1903a
3
+ size 2586705312
VLM_Only/RKD_Raw_VLM/rkd_temp_0.1/phase2/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:568379c4488a522a1dde97d01c39d6c431c5f384cf683d930f5fd27aa72ddab4
3
+ size 5841
VLM_Only/RKD_Raw_VLM/rkd_temp_0.1/phase3/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6be767e43c2e5cae059b147fee2f07dc3a1f8ba85235566823fc939b2f07bc29
3
+ size 4999367032
VLM_Only/RKD_Raw_VLM/rkd_temp_0.1/phase3/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:174ad1ac28316111731e5e5e058d8b69cad672ab202bb8523f783eb0380803a1
3
+ size 2586705312
VLM_Only/RKD_Raw_VLM/rkd_temp_0.1/phase3/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:def6eafca994c637f993a90c49f61cd70f2e49e708f676e7596d5954e6c59704
3
+ size 5841
VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase2/config.json ADDED
@@ -0,0 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5_RKD"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "rkd_action_temp": 0.2,
63
+ "rkd_enabled": true,
64
+ "rkd_fm_loss_weight": 0.0,
65
+ "rkd_loss_weight": 1.0,
66
+ "rkd_relation_mode": "token_pair_mean",
67
+ "rkd_vlm_temp": 0.2,
68
+ "torch_dtype": "bfloat16",
69
+ "transformers_version": "4.51.3"
70
+ }
VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase2/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a9da60c81bf9fb511a0856e800d5df7f8e12f98a3282602a85950c822721d901
3
+ size 4999367032
VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase2/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c966ae82934e4f401e1138f03ea5ab851e33e7b9f0ac68c5e0765ae0dd1903a
3
+ size 2586705312
VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase2/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase2/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:793bf25c7c18b561ce7b1ad5abcfb900e75a37967072ead2e172593925115ef9
3
+ size 5841
VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase3/config.json ADDED
@@ -0,0 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5_RKD"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "rkd_action_temp": 0.2,
63
+ "rkd_enabled": false,
64
+ "rkd_fm_loss_weight": 1.0,
65
+ "rkd_loss_weight": 0.0,
66
+ "rkd_relation_mode": "token_pair_mean",
67
+ "rkd_vlm_temp": 0.2,
68
+ "torch_dtype": "bfloat16",
69
+ "transformers_version": "4.51.3"
70
+ }
VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase3/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dabc67f917463bd508b73008f442b51ce671fa84ec9048de0cf0539a4380582c
3
+ size 4999367032
VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase3/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b3c913fc285d1f4734d32e9e5d69a363fc8cb0bdd34b41101f199caeb6d5690
3
+ size 2586705312
VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase3/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
VLM_Only/RKD_Raw_VLM/rkd_temp_0.2/phase3/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:15923e7c6cd7919eec5ab8bc6a064d0f1f0baa50e38ce6ffccc250c1845980c6
3
+ size 5841
VLM_Only/RKD_Raw_VLM/train.log ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ea08e7b94b52a52b537352ac7e0fad0744535ed4df4b269e034f2bc24964f85
3
+ size 31618395