Hugging Face's logo

open-gigaai
/
GigaBrain-0.7-3.5B-Base

Robotics
Diffusers
Safetensors
English
vla
embodied-ai
vision-language-action
Model card Files
xet
 Community
GigaBrain-0.7-3.5B-Base
File size: 1,903 Bytes
0fa1075
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
{
  "_class_name": "GigaBrain0Policy",
  "_diffusers_version": "0.34.0",
  "action_loss_vlm_gradient_weight": 0.1,
  "agent_pos_config": {
    "state": 32
  },
  "compact_language_logits": true,
  "crossdit_attention_head_dim": 64,
  "crossdit_num_target_vision_tokens": 32,
  "detach_history_frames": true,
  "enable_knowledge_insulation": true,
  "enable_learnable_traj_token": false,
  "enable_next_token_prediction": true,
  "expert_intermediate_size": null,
  "expert_type": "dual_stream",
  "fast_token_tail_skip_tokens": 128,
  "fast_token_tail_vocab_size": null,
  "fast_token_vocab_mode": "tail",
  "fast_vocab_size": 2048,
  "flow_action_horizon": 50,
  "gemma3_config_dict": null,
  "gemma3_expert_final_norm_ada": false,
  "gemma3_expert_full_attends_prefix": false,
  "gemma3_gate_self_mlp": true,
  "gemma4_config_dict": null,
  "gemma4_expert_cross_attend_shared_layers": false,
  "gemma4_expert_final_norm_ada": true,
  "gemma4_expert_full_attends_prefix": false,
  "has_action_expert": true,
  "language_loss_checkpoint": true,
  "language_loss_chunk_size": 0,
  "max_action_dim": 32,
  "max_state_dim": 32,
  "max_traj_dim": 4,
  "n_action_steps": 50,
  "num_embodiments": 8,
  "num_steps": 10,
  "num_traj_tokens": 10,
  "observation_memory_size": 1,
  "proj_width": 1024,
  "proj_with_mask": true,
  "propri_token_id": 257153,
  "qwen2_5_vl_config_dict": null,
  "qwen3_5_config_dict": null,
  "qwen3_vl_config_dict": null,
  "qwen3_vl_expert_adarms_base": false,
  "robomanip_attention_head_dim": 64,
  "robomanip_cross_attn_modality_alternating": true,
  "robomanip_ff_dim": null,
  "robomanip_num_layers": 10,
  "robomanip_num_query_tokens": 0,
  "state_hidden_size": 2048,
  "state_input_mode": "proprio_anchor",
  "temporal_attention_interval": 4,
  "traj_hidden_dim": 256,
  "use_cache": true,
  "vision_in_channels": 3,
  "vlm_hidden_size": 2304,
  "vlm_type": "paligemma2"
}