Upload BASE_seed0_40k/config.json with huggingface_hub
Browse files- BASE_seed0_40k/config.json +65 -0
BASE_seed0_40k/config.json
ADDED
|
@@ -0,0 +1,65 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_head_type": "flowmatching",
|
| 6 |
+
"action_horizon": 16,
|
| 7 |
+
"add_pos_embed": true,
|
| 8 |
+
"backbone_embedding_dim": 2048,
|
| 9 |
+
"diffusion_model_cfg": {
|
| 10 |
+
"attention_head_dim": 48,
|
| 11 |
+
"cross_attention_dim": 2048,
|
| 12 |
+
"dropout": 0.2,
|
| 13 |
+
"final_dropout": true,
|
| 14 |
+
"interleave_self_attention": true,
|
| 15 |
+
"norm_type": "ada_norm",
|
| 16 |
+
"num_attention_heads": 32,
|
| 17 |
+
"num_layers": 16,
|
| 18 |
+
"output_dim": 1024,
|
| 19 |
+
"positional_embeddings": null
|
| 20 |
+
},
|
| 21 |
+
"hidden_size": 1024,
|
| 22 |
+
"input_embedding_dim": 1536,
|
| 23 |
+
"max_action_dim": 32,
|
| 24 |
+
"max_state_dim": 64,
|
| 25 |
+
"model_dtype": "float32",
|
| 26 |
+
"noise_beta_alpha": 1.5,
|
| 27 |
+
"noise_beta_beta": 1.0,
|
| 28 |
+
"noise_s": 0.999,
|
| 29 |
+
"num_inference_timesteps": 4,
|
| 30 |
+
"num_target_vision_tokens": 32,
|
| 31 |
+
"num_timestep_buckets": 1000,
|
| 32 |
+
"tune_diffusion_model": true,
|
| 33 |
+
"tune_projector": true,
|
| 34 |
+
"use_vlln": true,
|
| 35 |
+
"vl_self_attention_cfg": {
|
| 36 |
+
"attention_head_dim": 64,
|
| 37 |
+
"dropout": 0.2,
|
| 38 |
+
"final_dropout": true,
|
| 39 |
+
"num_attention_heads": 32,
|
| 40 |
+
"num_layers": 4,
|
| 41 |
+
"positional_embeddings": null
|
| 42 |
+
}
|
| 43 |
+
},
|
| 44 |
+
"action_horizon": 16,
|
| 45 |
+
"architectures": [
|
| 46 |
+
"GR00T_N1_5"
|
| 47 |
+
],
|
| 48 |
+
"attn_implementation": null,
|
| 49 |
+
"backbone_cfg": {
|
| 50 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 51 |
+
"load_bf16": false,
|
| 52 |
+
"project_to_dim": null,
|
| 53 |
+
"reproject_vision": false,
|
| 54 |
+
"select_layer": 12,
|
| 55 |
+
"tune_llm": false,
|
| 56 |
+
"tune_visual": true,
|
| 57 |
+
"use_flash_attention": true
|
| 58 |
+
},
|
| 59 |
+
"compute_dtype": "bfloat16",
|
| 60 |
+
"hidden_size": 2048,
|
| 61 |
+
"model_dtype": "float32",
|
| 62 |
+
"model_type": "gr00t_n1_5",
|
| 63 |
+
"torch_dtype": "bfloat16",
|
| 64 |
+
"transformers_version": "4.51.3"
|
| 65 |
+
}
|