dwko commited on
Commit
2c5c417
·
verified ·
1 Parent(s): c77f9d6

Upload folder using huggingface_hub

Browse files
config.json ADDED
@@ -0,0 +1,133 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_in_proj_cfg": {
3
+ "_target_": "alpamayo_r1.models.action_in_proj.PerWaypointActionInProjV2",
4
+ "hidden_size": 512,
5
+ "max_freq": 100.0,
6
+ "num_enc_layers": 2,
7
+ "num_fourier_feats": 20
8
+ },
9
+ "action_out_proj_cfg": {
10
+ "_target_": "torch.nn.Linear"
11
+ },
12
+ "action_space_cfg": {
13
+ "_target_": "alpamayo_r1.action_space.UnicycleAccelCurvatureActionSpace",
14
+ "a_lambda": 0.0001,
15
+ "a_ridge": 0.0001,
16
+ "accel_bounds": [
17
+ -9.8,
18
+ 9.8
19
+ ],
20
+ "accel_mean": 0.02902694707164455,
21
+ "accel_std": 0.6810426736454882,
22
+ "curvature_bounds": [
23
+ -0.33,
24
+ 0.33
25
+ ],
26
+ "curvature_mean": 0.0002692167976330542,
27
+ "curvature_std": 0.026148280660833106,
28
+ "dt": 0.1,
29
+ "kappa_lambda": 0.0001,
30
+ "kappa_ridge": 0.0001,
31
+ "n_waypoints": 64,
32
+ "theta_lambda": 1e-06,
33
+ "theta_ridge": 1e-08,
34
+ "v_lambda": 1e-06,
35
+ "v_ridge": 0.0001
36
+ },
37
+ "add_special_tokens": true,
38
+ "architectures": [
39
+ "AlpamayoR1"
40
+ ],
41
+ "attn_implementation": null,
42
+ "diffusion_cfg": {
43
+ "_target_": "alpamayo_r1.diffusion.flow_matching.FlowMatching",
44
+ "int_method": "euler",
45
+ "x_dims": "???"
46
+ },
47
+ "dtype": "float16",
48
+ "expert_cfg": {
49
+ "dtype": "bfloat16",
50
+ "head_dim": 128,
51
+ "hidden_size": 2048,
52
+ "intermediate_size": 8256,
53
+ "num_attention_heads": 16
54
+ },
55
+ "expert_non_causal_attention": true,
56
+ "hist_traj_tokenizer_cfg": {
57
+ "_target_": "alpamayo_r1.models.delta_tokenizer.DeltaTrajectoryTokenizer"
58
+ },
59
+ "keep_same_dtype": true,
60
+ "max_pixels": 196608,
61
+ "min_pixels": 163840,
62
+ "model_dtype": "bfloat16",
63
+ "model_type": "alpamayo_r1",
64
+ "quantization_config": {
65
+ "_load_in_4bit": true,
66
+ "_load_in_8bit": false,
67
+ "bnb_4bit_compute_dtype": "float32",
68
+ "bnb_4bit_quant_storage": "uint8",
69
+ "bnb_4bit_quant_type": "fp4",
70
+ "bnb_4bit_use_double_quant": false,
71
+ "llm_int8_enable_fp32_cpu_offload": false,
72
+ "llm_int8_has_fp16_weight": false,
73
+ "llm_int8_skip_modules": null,
74
+ "llm_int8_threshold": 6.0,
75
+ "load_in_4bit": true,
76
+ "load_in_8bit": false,
77
+ "quant_method": "bitsandbytes"
78
+ },
79
+ "tokens_per_future_traj": 128,
80
+ "tokens_per_history_traj": 48,
81
+ "traj_token_ids": {
82
+ "future": 155685,
83
+ "future_end": 155683,
84
+ "future_start": 155681,
85
+ "history": 155684,
86
+ "history_end": 155676,
87
+ "history_start": 155674
88
+ },
89
+ "traj_token_start_idx": 151669,
90
+ "traj_tokenizer_cfg": {
91
+ "_recursive_": false,
92
+ "_target_": "alpamayo_r1.action_space.discrete_action_space.DiscreteTrajectoryTokenizer",
93
+ "action_space_cfg": {
94
+ "_target_": "alpamayo_r1.action_space.UnicycleAccelCurvatureActionSpace",
95
+ "a_lambda": 0.0001,
96
+ "a_ridge": 0.0001,
97
+ "accel_bounds": [
98
+ -9.8,
99
+ 9.8
100
+ ],
101
+ "accel_mean": 0.02902694707164455,
102
+ "accel_std": 0.6810426736454882,
103
+ "curvature_bounds": [
104
+ -0.33,
105
+ 0.33
106
+ ],
107
+ "curvature_mean": 0.0002692167976330542,
108
+ "curvature_std": 0.026148280660833106,
109
+ "dt": 0.1,
110
+ "kappa_lambda": 0.0001,
111
+ "kappa_ridge": 0.0001,
112
+ "n_waypoints": 64,
113
+ "theta_lambda": 1e-06,
114
+ "theta_ridge": 1e-08,
115
+ "v_lambda": 1e-06,
116
+ "v_ridge": 0.0001
117
+ },
118
+ "dims_max": [
119
+ 10,
120
+ 10
121
+ ],
122
+ "dims_min": [
123
+ -10,
124
+ -10
125
+ ],
126
+ "num_bins": 3000
127
+ },
128
+ "traj_vocab_size": 4000,
129
+ "transformers_version": "4.57.3",
130
+ "vlm_backend": "qwenvl3",
131
+ "vlm_name_or_path": "Qwen/Qwen3-VL-8B-Instruct",
132
+ "vocab_size": 155697
133
+ }
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b2e18a0db15204a3c5f9ac15398ba13a2ac0977b732736baf9cde9c982b0710d
3
+ size 4995460982
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a564abc09885a67aa60372ca3d910958e5d7e2dea95589577b9a49212ace3659
3
+ size 3077883685
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff