Hugging Face's logo

MinghaoFu
/
rlwf-ckpt

Safetensors
Model card Files
xet
 Community
MinghaoFu commited on
Commit
30a1822
·
verified ·
1 Parent(s): f197359

upload mimicgen-core-14b-lora-step80000 (LoRA weights only, no optimizer state)

Browse files
Files changed (4) hide show
  1. mimicgen-core-14b-lora-step80000/config.json +114 -0
  2. mimicgen-core-14b-lora-step80000/experiment_cfg/conf.yaml +0 -0
  3. mimicgen-core-14b-lora-step80000/experiment_cfg/metadata.json +213 -0
  4. mimicgen-core-14b-lora-step80000/model.safetensors +3 -0
mimicgen-core-14b-lora-step80000/config.json ADDED
@@ -0,0 +1,114 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "_convert_": "object",
5
+ "_target_": "groot.vla.model.dreamzero.action_head.wan_flow_matching_action_tf.WANPolicyHead",
6
+ "config": {
7
+ "_recursive_": false,
8
+ "_target_": "groot.vla.model.dreamzero.action_head.wan_flow_matching_action_tf.WANPolicyHeadConfig",
9
+ "action_dim": 32,
10
+ "action_horizon": 24,
11
+ "action_loss_embodiment_ids": [
12
+ 26,
13
+ 17,
14
+ 32,
15
+ 33
16
+ ],
17
+ "add_pos_embed": true,
18
+ "backbone_embedding_dim": 0,
19
+ "backbone_features_projector_cfg": null,
20
+ "decouple_video_action_noise": false,
21
+ "diffusion_model_cfg": {
22
+ "_convert_": "object",
23
+ "_target_": "groot.vla.model.dreamzero.modules.wan_video_dit_action_casual_chunk.CausalWanModel",
24
+ "diffusion_model_pretrained_path": "/workspace/minghao/DreamZero-Libero/checkpoints/Wan2.1-I2V-14B-480P",
25
+ "dim": 5120,
26
+ "eps": 1e-06,
27
+ "ffn_dim": 13824,
28
+ "frame_seqlen": 512,
29
+ "freq_dim": 256,
30
+ "in_dim": 36,
31
+ "max_chunk_size": 4,
32
+ "model_type": "i2v",
33
+ "num_action_per_block": 24,
34
+ "num_frame_per_block": 2,
35
+ "num_heads": 40,
36
+ "num_layers": 40,
37
+ "num_state_per_block": 1,
38
+ "out_dim": 16
39
+ },
40
+ "expand_batch": null,
41
+ "freeze_decode_layer": false,
42
+ "freeze_text_image_vae": true,
43
+ "hidden_size": 64,
44
+ "image_encoder_cfg": {
45
+ "_convert_": "object",
46
+ "_target_": "groot.vla.model.dreamzero.modules.wan_video_image_encoder.WanImageEncoder",
47
+ "image_encoder_pretrained_path": "/workspace/minghao/DreamZero-Libero/checkpoints/Wan2.1-I2V-14B-480P/models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth"
48
+ },
49
+ "init_lora_weights": "kaiming",
50
+ "input_embedding_dim": 1536,
51
+ "load_pretrained_det_decode_layer_path": null,
52
+ "lora_alpha": 4,
53
+ "lora_rank": 4,
54
+ "lora_target_modules": "q,k,v,o,ffn.0,ffn.2",
55
+ "max_action_dim": 32,
56
+ "max_state_dim": 64,
57
+ "model_dtype": "float32",
58
+ "noise_beta_alpha": 1.5,
59
+ "noise_beta_beta": 1.0,
60
+ "noise_s": 0.999,
61
+ "num_frame_per_block": 2,
62
+ "num_frames": 33,
63
+ "num_inference_timesteps": 4,
64
+ "num_timestep_buckets": 1000,
65
+ "repa_coeff": 1.0,
66
+ "repa_layer": 8,
67
+ "skip_all_component_loading": false,
68
+ "text_encoder_cfg": {
69
+ "_convert_": "object",
70
+ "_target_": "groot.vla.model.dreamzero.modules.wan_video_text_encoder.WanTextEncoder",
71
+ "text_encoder_pretrained_path": "/workspace/minghao/DreamZero-Libero/checkpoints/Wan2.1-I2V-14B-480P/models_t5_umt5-xxl-enc-bf16.pth"
72
+ },
73
+ "tile_size_height": 34,
74
+ "tile_size_width": 34,
75
+ "tile_stride_height": 18,
76
+ "tile_stride_width": 16,
77
+ "tiled": false,
78
+ "train_architecture": "lora",
79
+ "tune_diffusion_model": true,
80
+ "tune_projector": true,
81
+ "use_gradient_checkpointing": true,
82
+ "use_vlln": true,
83
+ "vae_cfg": {
84
+ "_convert_": "object",
85
+ "_target_": "groot.vla.model.dreamzero.modules.wan_video_vae.WanVideoVAE",
86
+ "vae_pretrained_path": "/workspace/minghao/DreamZero-Libero/checkpoints/Wan2.1-I2V-14B-480P/Wan2.1_VAE.pth"
87
+ },
88
+ "video_noise_beta_alpha": 3.0,
89
+ "video_noise_beta_beta": 1.0,
90
+ "vl_self_attention_cfg": {
91
+ "_target_": "groot.vla.model.n1_5.modules.cross_attention_dit.SelfAttentionTransformer",
92
+ "attention_head_dim": 64,
93
+ "dropout": 0.2,
94
+ "final_dropout": true,
95
+ "num_attention_heads": 24,
96
+ "num_layers": 4,
97
+ "positional_embeddings": null
98
+ }
99
+ }
100
+ },
101
+ "action_horizon": 24,
102
+ "architectures": [
103
+ "VLA"
104
+ ],
105
+ "backbone_cfg": {
106
+ "_target_": "groot.vla.model.dreamzero.backbone.identity.IdentityBackbone"
107
+ },
108
+ "hidden_size": 0,
109
+ "model_dtype": "float32",
110
+ "model_type": "vla",
111
+ "resume_path": "/workspace/minghao/DreamZero-Libero/checkpoints/dreamzero_mimicgen_core_14b_lora",
112
+ "torch_dtype": "bfloat16",
113
+ "transformers_version": "4.53.2"
114
+ }
mimicgen-core-14b-lora-step80000/experiment_cfg/conf.yaml ADDED
The diff for this file is too large to render. See raw diff
 
mimicgen-core-14b-lora-step80000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,213 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "mimicgen_sim": {
3
+ "statistics": {
4
+ "state": {
5
+ "eef_state": {
6
+ "max": [
7
+ 0.3781653253899684,
8
+ 0.5555333260011485,
9
+ 1.223503939696811,
10
+ 3.141571766729021,
11
+ 3.14091583583906,
12
+ 1.9111917518750583
13
+ ],
14
+ "min": [
15
+ -0.5484811260165126,
16
+ -0.47873402319250064,
17
+ 0.8084958111356036,
18
+ -3.1415754826269975,
19
+ -3.140540736976143,
20
+ -1.8040912167796939
21
+ ],
22
+ "mean": [
23
+ -0.0177051364409928,
24
+ -0.004981000174034581,
25
+ 0.9706267985637327,
26
+ 0.1844200134547997,
27
+ -0.12205160101266108,
28
+ -0.002227242358943213
29
+ ],
30
+ "std": [
31
+ 0.12111342372387424,
32
+ 0.13549865760316396,
33
+ 0.07447554508819595,
34
+ 2.858073076760002,
35
+ 1.0624186251331758,
36
+ 0.17310904723441267
37
+ ],
38
+ "q01": [
39
+ -0.3234043401351073,
40
+ -0.3020909569103607,
41
+ 0.8174326577729326,
42
+ -3.1339872558115913,
43
+ -2.6965523669590237,
44
+ -0.5625783582305172
45
+ ],
46
+ "q99": [
47
+ 0.23348672202666101,
48
+ 0.29738309079264025,
49
+ 1.1317326011334667,
50
+ 3.1340257107853837,
51
+ 2.6179584127357147,
52
+ 0.3739185832694553
53
+ ]
54
+ },
55
+ "gripper_state": {
56
+ "max": [
57
+ 0.04479708965398592,
58
+ 0.014002309454187324
59
+ ],
60
+ "min": [
61
+ -0.013601018557230699,
62
+ -0.04476327842184798
63
+ ],
64
+ "mean": [
65
+ 0.026640730366728437,
66
+ -0.026341034527659554
67
+ ],
68
+ "std": [
69
+ 0.013965685716068156,
70
+ 0.014091458989452393
71
+ ],
72
+ "q01": [
73
+ -0.004554157407015276,
74
+ -0.04051676591568512
75
+ ],
76
+ "q99": [
77
+ 0.04047115507141684,
78
+ 0.004661336343960254
79
+ ]
80
+ }
81
+ },
82
+ "action": {
83
+ "pose_delta": {
84
+ "max": [
85
+ 1.0,
86
+ 1.0,
87
+ 1.0,
88
+ 1.0,
89
+ 1.0,
90
+ 1.0
91
+ ],
92
+ "min": [
93
+ -1.0,
94
+ -1.0,
95
+ -1.0,
96
+ -1.0,
97
+ -1.0,
98
+ -1.0
99
+ ],
100
+ "mean": [
101
+ 0.03960332447147033,
102
+ 0.00471503964658854,
103
+ -0.01797380100731096,
104
+ 0.0009091264346900674,
105
+ 0.006353555250611444,
106
+ 0.009815371033736992
107
+ ],
108
+ "std": [
109
+ 0.4040929305692675,
110
+ 0.45100468035625474,
111
+ 0.35916913490773844,
112
+ 0.06847438458646551,
113
+ 0.08666168306300813,
114
+ 0.24475533809103922
115
+ ],
116
+ "q01": [
117
+ -1.0,
118
+ -1.0,
119
+ -0.7628942409285452,
120
+ -0.1682020859737398,
121
+ -0.19670659331125595,
122
+ -0.9992047220611601
123
+ ],
124
+ "q99": [
125
+ 1.0,
126
+ 1.0,
127
+ 1.0,
128
+ 0.16815162314605947,
129
+ 0.23174351675724905,
130
+ 1.0
131
+ ]
132
+ },
133
+ "gripper_position": {
134
+ "max": [
135
+ 1.0
136
+ ],
137
+ "min": [
138
+ 0.0
139
+ ],
140
+ "mean": [
141
+ 0.5932405664162677
142
+ ],
143
+ "std": [
144
+ 0.4912292710797897
145
+ ],
146
+ "q01": [
147
+ 0.0
148
+ ],
149
+ "q99": [
150
+ 1.0
151
+ ]
152
+ }
153
+ }
154
+ },
155
+ "modalities": {
156
+ "video": {
157
+ "agentview_rgb": {
158
+ "resolution": [
159
+ 256,
160
+ 256
161
+ ],
162
+ "channels": 3,
163
+ "fps": 20.0
164
+ },
165
+ "eye_in_hand_rgb": {
166
+ "resolution": [
167
+ 256,
168
+ 256
169
+ ],
170
+ "channels": 3,
171
+ "fps": 20.0
172
+ }
173
+ },
174
+ "state": {
175
+ "eef_state": {
176
+ "absolute": true,
177
+ "rotation_type": null,
178
+ "shape": [
179
+ 6
180
+ ],
181
+ "continuous": true
182
+ },
183
+ "gripper_state": {
184
+ "absolute": true,
185
+ "rotation_type": null,
186
+ "shape": [
187
+ 2
188
+ ],
189
+ "continuous": true
190
+ }
191
+ },
192
+ "action": {
193
+ "pose_delta": {
194
+ "absolute": true,
195
+ "rotation_type": null,
196
+ "shape": [
197
+ 6
198
+ ],
199
+ "continuous": true
200
+ },
201
+ "gripper_position": {
202
+ "absolute": true,
203
+ "rotation_type": null,
204
+ "shape": [
205
+ 1
206
+ ],
207
+ "continuous": true
208
+ }
209
+ }
210
+ },
211
+ "embodiment_tag": "mimicgen_sim"
212
+ }
213
+ }
mimicgen-core-14b-lora-step80000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e2f99aa706044171c4a9ff1514160e60ba3e59a198857efd23b0028dcf03ee8
3
+ size 217288600