{
    "type": "xr1_stage2",
    "n_obs_steps": 2,
    "action_sample_factor": 1,
    "normalization_mapping": {
        "VISUAL": "IDENTITY",
        "STATE": "MEAN_STD",
        "ACTION": "MEAN_STD"
    },
    "input_features": {
        "observation.state": {
            "type": "STATE",
            "shape": [
                32
            ]
        },
        "observation.images.image_0": {
            "type": "VISUAL",
            "shape": [
                224,
                3,
                224
            ]
        },
        "observation.images.image_1": {
            "type": "VISUAL",
            "shape": [
                224,
                3,
                224
            ]
        },
        "observation.images.image_2": {
            "type": "VISUAL",
            "shape": [
                224,
                3,
                224
            ]
        },
        "observation.images.image_3": {
            "type": "VISUAL",
            "shape": [
                224,
                3,
                224
            ]
        },
        "observation.images.image_wrist_0": {
            "type": "VISUAL",
            "shape": [
                224,
                3,
                224
            ]
        },
        "observation.images.image_wrist_1": {
            "type": "VISUAL",
            "shape": [
                224,
                3,
                224
            ]
        },
        "observation.state_is_pad": {
            "type": "STATE",
            "shape": [
                2
            ]
        },
        "observation.images.image_0_is_pad": {
            "type": "STATE",
            "shape": [
                2
            ]
        },
        "observation.images.image_1_is_pad": {
            "type": "STATE",
            "shape": [
                2
            ]
        },
        "observation.images.image_2_is_pad": {
            "type": "STATE",
            "shape": [
                2
            ]
        },
        "observation.images.image_3_is_pad": {
            "type": "STATE",
            "shape": [
                2
            ]
        },
        "observation.images.image_wrist_0_is_pad": {
            "type": "STATE",
            "shape": [
                2
            ]
        },
        "observation.images.image_wrist_1_is_pad": {
            "type": "STATE",
            "shape": [
                2
            ]
        }
    },
    "output_features": {
        "action": {
            "type": "ACTION",
            "shape": [
                32
            ]
        },
        "action_is_pad": {
            "type": "ACTION",
            "shape": [
                50
            ]
        }
    },
    "stage1_pretrained_path": "/media/jushen/bamboo-fan/Save/crossvla/stage1_4dataset_10_8_bs12_kl_cb256_0528_stat/checkpoints/275000/pretrained_model/",
    "stage2_pretrained_path": "None",
    "stage2_latent_image_token_check": false,
    "dataset_stats_generate": true,
    "heterogeneous": true,
    "split_dataset": true,
    "real_robot_dev": false,
    "image_interval_steps": 50,
    "action_latent_token_num": 13,
    "mformer_hidden_size": 768,
    "decoder_hidden_size": 768,
    "codebook_embed_dim": 256,
    "codebook_k_size": 256,
    "action_chunk_size": 50,
    "chunk_size": 50,
    "n_action_steps": 50,
    "resampler": true,
    "resampler_dim": 2048,
    "resampler_depth": 3,
    "resampler_dim_head": 128,
    "resampler_heads": 4,
    "resampler_num_media_embeds": 1,
    "resampler_num_latents": 9,
    "max_state_dim": 32,
    "max_action_dim": 32,
    "resize_imgs_with_padding": [
        224,
        224
    ],
    "empty_cameras": 0,
    "adapt_to_pi_aloha": false,
    "use_delta_joint_actions_aloha": false,
    "tokenizer_max_length": 48,
    "proj_width": 1024,
    "num_steps": 10,
    "use_cache": true,
    "attention_implementation": "eager",
    "freeze_vision_encoder": false,
    "freeze_language_encoder": true,
    "train_expert_only": false,
    "train_state_proj": true,
    "optimizer_lr": 0.0001,
    "optimizer_betas": [
        0.9,
        0.95
    ],
    "optimizer_eps": 1e-08,
    "optimizer_weight_decay": 1e-10,
    "scheduler_warmup_steps": 5000,
    "scheduler_decay_steps": 300000,
    "scheduler_decay_lr": 1e-06
}