Instructions to use open-gigaai/GigaBrain-0.7-3.5B-Base with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use open-gigaai/GigaBrain-0.7-3.5B-Base with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("open-gigaai/GigaBrain-0.7-3.5B-Base", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
Upload GigaBrain-0.7 config
Browse files- config.json +61 -0
config.json
ADDED
|
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_class_name": "GigaBrain0Policy",
|
| 3 |
+
"_diffusers_version": "0.34.0",
|
| 4 |
+
"action_loss_vlm_gradient_weight": 0.1,
|
| 5 |
+
"agent_pos_config": {
|
| 6 |
+
"state": 32
|
| 7 |
+
},
|
| 8 |
+
"compact_language_logits": true,
|
| 9 |
+
"crossdit_attention_head_dim": 64,
|
| 10 |
+
"crossdit_num_target_vision_tokens": 32,
|
| 11 |
+
"detach_history_frames": true,
|
| 12 |
+
"enable_knowledge_insulation": true,
|
| 13 |
+
"enable_learnable_traj_token": false,
|
| 14 |
+
"enable_next_token_prediction": true,
|
| 15 |
+
"expert_intermediate_size": null,
|
| 16 |
+
"expert_type": "dual_stream",
|
| 17 |
+
"fast_token_tail_skip_tokens": 128,
|
| 18 |
+
"fast_token_tail_vocab_size": null,
|
| 19 |
+
"fast_token_vocab_mode": "tail",
|
| 20 |
+
"fast_vocab_size": 2048,
|
| 21 |
+
"flow_action_horizon": 50,
|
| 22 |
+
"gemma3_config_dict": null,
|
| 23 |
+
"gemma3_expert_final_norm_ada": false,
|
| 24 |
+
"gemma3_expert_full_attends_prefix": false,
|
| 25 |
+
"gemma3_gate_self_mlp": true,
|
| 26 |
+
"gemma4_config_dict": null,
|
| 27 |
+
"gemma4_expert_cross_attend_shared_layers": false,
|
| 28 |
+
"gemma4_expert_final_norm_ada": true,
|
| 29 |
+
"gemma4_expert_full_attends_prefix": false,
|
| 30 |
+
"has_action_expert": true,
|
| 31 |
+
"language_loss_checkpoint": true,
|
| 32 |
+
"language_loss_chunk_size": 0,
|
| 33 |
+
"max_action_dim": 32,
|
| 34 |
+
"max_state_dim": 32,
|
| 35 |
+
"max_traj_dim": 4,
|
| 36 |
+
"n_action_steps": 50,
|
| 37 |
+
"num_embodiments": 8,
|
| 38 |
+
"num_steps": 10,
|
| 39 |
+
"num_traj_tokens": 10,
|
| 40 |
+
"observation_memory_size": 1,
|
| 41 |
+
"proj_width": 1024,
|
| 42 |
+
"proj_with_mask": true,
|
| 43 |
+
"propri_token_id": 257153,
|
| 44 |
+
"qwen2_5_vl_config_dict": null,
|
| 45 |
+
"qwen3_5_config_dict": null,
|
| 46 |
+
"qwen3_vl_config_dict": null,
|
| 47 |
+
"qwen3_vl_expert_adarms_base": false,
|
| 48 |
+
"robomanip_attention_head_dim": 64,
|
| 49 |
+
"robomanip_cross_attn_modality_alternating": true,
|
| 50 |
+
"robomanip_ff_dim": null,
|
| 51 |
+
"robomanip_num_layers": 10,
|
| 52 |
+
"robomanip_num_query_tokens": 0,
|
| 53 |
+
"state_hidden_size": 2048,
|
| 54 |
+
"state_input_mode": "proprio_anchor",
|
| 55 |
+
"temporal_attention_interval": 4,
|
| 56 |
+
"traj_hidden_dim": 256,
|
| 57 |
+
"use_cache": true,
|
| 58 |
+
"vision_in_channels": 3,
|
| 59 |
+
"vlm_hidden_size": 2304,
|
| 60 |
+
"vlm_type": "paligemma2"
|
| 61 |
+
}
|