Upload experiment_cfg/final_model_config.json with huggingface_hub

experiment_cfg/final_model_config.json

@@ -0,0 +1,207 @@
+{
+  "model_type": "RLDX-1",
+  "model_dtype": "bfloat16",
+  "model_name": "RLWRLD/RLDX-1-VLM",
+  "backbone_model_type": "vtc_qwen3_vl",
+  "model_revision": null,
+  "tune_top_llm_layers": 0,
+  "backbone_embedding_dim": 4096,
+  "tune_llm": false,
+  "tune_visual": false,
+  "select_layer": 18,
+  "reproject_vision": false,
+  "use_flash_attention": true,
+  "load_bf16": true,
+  "backbone_trainable_params_fp32": true,
+  "freeze_cog_tokens": false,
+  "keep_full_lm": false,
+  "vlm_lang_loss": false,
+  "vlm_lang_weight": 0.5,
+  "image_max_area": 65536,
+  "image_resize_m": 32,
+  "apply_sincos_state_encoding": false,
+  "use_relative_action": true,
+  "use_percentiles": true,
+  "conversation_image_first": false,
+  "max_state_dim": 64,
+  "max_action_dim": 64,
+  "action_horizon": 16,
+  "hidden_size": 1024,
+  "input_embedding_dim": 1536,
+  "general_embodiment_train_ratio": 0,
+  "add_pos_embed": true,
+  "attn_dropout": 0.2,
+  "use_vlln": false,
+  "max_seq_len": 1024,
+  "n_cog_tokens": 64,
+  "diffusion_model_cfg": {
+    "attention_head_dim": 64,
+    "depth_multi_stream": 4,
+    "depth_single_stream": 8,
+    "dropout": 0.2,
+    "final_dropout": true,
+    "num_attention_heads": 24,
+    "output_dim": 1024,
+    "positional_embeddings": "rope_sa_only",
+    "rope_theta": 10000.0,
+    "temb_type": "input_token",
+    "use_swiglu": true,
+    "action_model_max_seq_len": 512,
+    "pre_norm": "layer_norm",
+    "qk_norm": "rms_norm",
+    "sa_dim": 1536,
+    "vl_dim": 4096,
+    "use_physics": false,
+    "physics_dim": 0
+  },
+  "num_inference_timesteps": 4,
+  "noise_beta_alpha": 1.5,
+  "noise_beta_beta": 1.0,
+  "noise_s": 0.999,
+  "num_timestep_buckets": 1000,
+  "tune_projector": true,
+  "tune_diffusion_model": true,
+  "tune_vlln": true,
+  "action_model_use_lora": false,
+  "action_model_lora_rank": 16,
+  "action_model_lora_alpha": 32,
+  "action_model_lora_dropout": 0.0,
+  "action_model_lora_target_modules": [
+    "vl_qkv",
+    "vl_proj",
+    "sa_qkv",
+    "sa_proj",
+    "p_qkv",
+    "p_proj",
+    "linear1",
+    "linear2"
+  ],
+  "backbone_use_lora": false,
+  "backbone_lora_rank": 16,
+  "backbone_lora_alpha": 32,
+  "backbone_lora_dropout": 0.0,
+  "backbone_lora_num_layers": -1,
+  "backbone_lora_target_modules": [
+    "q_proj",
+    "k_proj",
+    "v_proj",
+    "o_proj",
+    "gate_proj",
+    "up_proj",
+    "down_proj"
+  ],
+  "state_dropout_prob": 0.0,
+  "state_additive_noise_scale": 0.0,
+  "rtc_training_max_delay": 0,
+  "rtc_inference_mode": "none",
+  "rtc_inference_delay": 0,
+  "rtc_inference_exec_horizon": 0,
+  "rtc_jacobian_steps_only": 3,
+  "use_memory": false,
+  "memory_length": 4,
+  "memory_n_cog_tokens": null,
+  "concat_memory": false,
+  "memory_dropout_prob": 0.0,
+  "memory_stride": 16,
+  "memory_cfg": {
+    "hidden_size": 4096,
+    "intermediate_size": 16384,
+    "num_hidden_layers": 2,
+    "num_attention_heads": 16,
+    "num_key_value_heads": 16,
+    "max_position_embeddings": 32,
+    "rms_norm_eps": 1e-05,
+    "use_causal_attn": true,
+    "use_rope": true
+  },
+  "use_motion": false,
+  "motion_insert_layer": 9,
+  "motion_d_hid": 512,
+  "motion_window": [
+    5,
+    9,
+    9
+  ],
+  "motion_ext_chnls": [
+    256
+  ],
+  "motion_int_chnls": [
+    256,
+    256,
+    512
+  ],
+  "motion_corr_func": "cosine",
+  "motion_n_encoders": 1,
+  "motion_use_layerscale": false,
+  "motion_layerscale_init": 1e-05,
+  "motion_use_layernorm": false,
+  "motion_use_syncbn": false,
+  "motion_injection_point": "vision_encoder",
+  "motion_pool_type": "avg",
+  "motion_drop": true,
+  "motion_gradient_check": false,
+  "motion_int_mode": "lite",
+  "use_video": true,
+  "video_length": 1,
+  "video_stride": 2,
+  "use_physics": false,
+  "physics_keys": [],
+  "physics_dims": [],
+  "physics_loss_weight": 0.1,
+  "allow_missing_physics": false,
+  "physics_delta_indices": null,
+  "physics_use_flow_matching": true,
+  "physics_dropout_prob": 0.0,
+  "use_finger_latent": false,
+  "finger_wrist_mode": "delta",
+  "finger_bimanual": false,
+  "finger_alignment_npz": "",
+  "finger_fingertip_norm_npz": "",
+  "finger_wrist_delta_stats_npz": "",
+  "finger_wrist_abs_stats_npz": "",
+  "embod_reg_mode": "",
+  "embod_reg_weight": 0.05,
+  "embod_reg_layer": 2,
+  "embod_reg_gather": false,
+  "embod_reg_tbins": 0,
+  "embod_reg_vic_var": 1.0,
+  "embod_reg_vic_cov": 0.04,
+  "embod_reg_hand_bins": 0,
+  "embod_reg_hand_edges": "",
+  "anchor_dim_norm": "",
+  "anchor_dim_stats_path": "",
+  "anchor_tips_dims": 15,
+  "anchor_human_tips_weight": 1.0,
+  "actlat_mode": false,
+  "actlat_tokenizer_path": "",
+  "actlat_embodiment_id": "openarm_prq",
+  "actlat_prq_stats_path": "",
+  "actlat_standardize": false,
+  "actlat_z_stats_path": "",
+  "anchor_readout_weight": 0.0,
+  "anchor_readout_taps": "cog,d2,s3,s7",
+  "anchor_readout_dim": 24,
+  "anchor_readout_tweight": true,
+  "anchor_readout_shuffle": false,
+  "anchor_readout_asym": false,
+  "gr1hp_align_weight": 0.2,
+  "gr1hp_align_only": true,
+  "gr1hp_align_dim": 24,
+  "gr1hp_align_shuffle": false,
+  "geom_bn_mode": "",
+  "geom_bn_weight": 0.0,
+  "geom_bn_dim": 24,
+  "geom_bn_asym": true,
+  "geom_bn_ema": 0.0,
+  "geom_bn_shuffle": false,
+  "geom_bn_lift_hidden": 0,
+  "geom_bn_route_warmup_steps": 0,
+  "geom_effort_dim": 0,
+  "geom_effort_weight": 0.0,
+  "egopi_split_embodiment": false,
+  "gr1hp_split_embodiment": false,
+  "train_stage": "",
+  "stage_from_checkpoint": "",
+  "freeze_modules": "",
+  "max_num_embodiments": 36
+}