Add files using upload-large-folder tool
Browse files- config.json +173 -0
- model.safetensors +3 -0
- train_config.json +289 -0
config.json
ADDED
|
@@ -0,0 +1,173 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"type": "xr1_stage2",
|
| 3 |
+
"n_obs_steps": 2,
|
| 4 |
+
"action_sample_factor": 1,
|
| 5 |
+
"normalization_mapping": {
|
| 6 |
+
"VISUAL": "IDENTITY",
|
| 7 |
+
"STATE": "MEAN_STD",
|
| 8 |
+
"ACTION": "MEAN_STD"
|
| 9 |
+
},
|
| 10 |
+
"input_features": {
|
| 11 |
+
"observation.state": {
|
| 12 |
+
"type": "STATE",
|
| 13 |
+
"shape": [
|
| 14 |
+
32
|
| 15 |
+
]
|
| 16 |
+
},
|
| 17 |
+
"observation.images.image_0": {
|
| 18 |
+
"type": "VISUAL",
|
| 19 |
+
"shape": [
|
| 20 |
+
224,
|
| 21 |
+
3,
|
| 22 |
+
224
|
| 23 |
+
]
|
| 24 |
+
},
|
| 25 |
+
"observation.images.image_1": {
|
| 26 |
+
"type": "VISUAL",
|
| 27 |
+
"shape": [
|
| 28 |
+
224,
|
| 29 |
+
3,
|
| 30 |
+
224
|
| 31 |
+
]
|
| 32 |
+
},
|
| 33 |
+
"observation.images.image_2": {
|
| 34 |
+
"type": "VISUAL",
|
| 35 |
+
"shape": [
|
| 36 |
+
224,
|
| 37 |
+
3,
|
| 38 |
+
224
|
| 39 |
+
]
|
| 40 |
+
},
|
| 41 |
+
"observation.images.image_3": {
|
| 42 |
+
"type": "VISUAL",
|
| 43 |
+
"shape": [
|
| 44 |
+
224,
|
| 45 |
+
3,
|
| 46 |
+
224
|
| 47 |
+
]
|
| 48 |
+
},
|
| 49 |
+
"observation.images.image_wrist_0": {
|
| 50 |
+
"type": "VISUAL",
|
| 51 |
+
"shape": [
|
| 52 |
+
224,
|
| 53 |
+
3,
|
| 54 |
+
224
|
| 55 |
+
]
|
| 56 |
+
},
|
| 57 |
+
"observation.images.image_wrist_1": {
|
| 58 |
+
"type": "VISUAL",
|
| 59 |
+
"shape": [
|
| 60 |
+
224,
|
| 61 |
+
3,
|
| 62 |
+
224
|
| 63 |
+
]
|
| 64 |
+
},
|
| 65 |
+
"observation.state_is_pad": {
|
| 66 |
+
"type": "STATE",
|
| 67 |
+
"shape": [
|
| 68 |
+
2
|
| 69 |
+
]
|
| 70 |
+
},
|
| 71 |
+
"observation.images.image_0_is_pad": {
|
| 72 |
+
"type": "STATE",
|
| 73 |
+
"shape": [
|
| 74 |
+
2
|
| 75 |
+
]
|
| 76 |
+
},
|
| 77 |
+
"observation.images.image_1_is_pad": {
|
| 78 |
+
"type": "STATE",
|
| 79 |
+
"shape": [
|
| 80 |
+
2
|
| 81 |
+
]
|
| 82 |
+
},
|
| 83 |
+
"observation.images.image_2_is_pad": {
|
| 84 |
+
"type": "STATE",
|
| 85 |
+
"shape": [
|
| 86 |
+
2
|
| 87 |
+
]
|
| 88 |
+
},
|
| 89 |
+
"observation.images.image_3_is_pad": {
|
| 90 |
+
"type": "STATE",
|
| 91 |
+
"shape": [
|
| 92 |
+
2
|
| 93 |
+
]
|
| 94 |
+
},
|
| 95 |
+
"observation.images.image_wrist_0_is_pad": {
|
| 96 |
+
"type": "STATE",
|
| 97 |
+
"shape": [
|
| 98 |
+
2
|
| 99 |
+
]
|
| 100 |
+
},
|
| 101 |
+
"observation.images.image_wrist_1_is_pad": {
|
| 102 |
+
"type": "STATE",
|
| 103 |
+
"shape": [
|
| 104 |
+
2
|
| 105 |
+
]
|
| 106 |
+
}
|
| 107 |
+
},
|
| 108 |
+
"output_features": {
|
| 109 |
+
"action": {
|
| 110 |
+
"type": "ACTION",
|
| 111 |
+
"shape": [
|
| 112 |
+
32
|
| 113 |
+
]
|
| 114 |
+
},
|
| 115 |
+
"action_is_pad": {
|
| 116 |
+
"type": "ACTION",
|
| 117 |
+
"shape": [
|
| 118 |
+
50
|
| 119 |
+
]
|
| 120 |
+
}
|
| 121 |
+
},
|
| 122 |
+
"stage1_pretrained_path": "/media/jushen/bamboo-fan/Save/crossvla/stage1_4dataset_10_8_bs12_kl_cb256_0528_stat/checkpoints/275000/pretrained_model/",
|
| 123 |
+
"stage2_pretrained_path": "None",
|
| 124 |
+
"stage2_latent_image_token_check": false,
|
| 125 |
+
"dataset_stats_generate": true,
|
| 126 |
+
"heterogeneous": true,
|
| 127 |
+
"split_dataset": true,
|
| 128 |
+
"real_robot_dev": false,
|
| 129 |
+
"image_interval_steps": 50,
|
| 130 |
+
"action_latent_token_num": 13,
|
| 131 |
+
"mformer_hidden_size": 768,
|
| 132 |
+
"decoder_hidden_size": 768,
|
| 133 |
+
"codebook_embed_dim": 256,
|
| 134 |
+
"codebook_k_size": 256,
|
| 135 |
+
"action_chunk_size": 50,
|
| 136 |
+
"chunk_size": 50,
|
| 137 |
+
"n_action_steps": 50,
|
| 138 |
+
"resampler": true,
|
| 139 |
+
"resampler_dim": 2048,
|
| 140 |
+
"resampler_depth": 3,
|
| 141 |
+
"resampler_dim_head": 128,
|
| 142 |
+
"resampler_heads": 4,
|
| 143 |
+
"resampler_num_media_embeds": 1,
|
| 144 |
+
"resampler_num_latents": 9,
|
| 145 |
+
"max_state_dim": 32,
|
| 146 |
+
"max_action_dim": 32,
|
| 147 |
+
"resize_imgs_with_padding": [
|
| 148 |
+
224,
|
| 149 |
+
224
|
| 150 |
+
],
|
| 151 |
+
"empty_cameras": 0,
|
| 152 |
+
"adapt_to_pi_aloha": false,
|
| 153 |
+
"use_delta_joint_actions_aloha": false,
|
| 154 |
+
"tokenizer_max_length": 48,
|
| 155 |
+
"proj_width": 1024,
|
| 156 |
+
"num_steps": 10,
|
| 157 |
+
"use_cache": true,
|
| 158 |
+
"attention_implementation": "eager",
|
| 159 |
+
"freeze_vision_encoder": false,
|
| 160 |
+
"freeze_language_encoder": true,
|
| 161 |
+
"train_expert_only": false,
|
| 162 |
+
"train_state_proj": true,
|
| 163 |
+
"optimizer_lr": 0.0001,
|
| 164 |
+
"optimizer_betas": [
|
| 165 |
+
0.9,
|
| 166 |
+
0.95
|
| 167 |
+
],
|
| 168 |
+
"optimizer_eps": 1e-08,
|
| 169 |
+
"optimizer_weight_decay": 1e-10,
|
| 170 |
+
"scheduler_warmup_steps": 5000,
|
| 171 |
+
"scheduler_decay_steps": 300000,
|
| 172 |
+
"scheduler_decay_lr": 1e-06
|
| 173 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9ad846e3601d2449bfd9038e88a779c3b9e72947f69e956e761c74f58a3c95ce
|
| 3 |
+
size 16423335872
|
train_config.json
ADDED
|
@@ -0,0 +1,289 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"dataset": {
|
| 3 |
+
"select_dataset": "HumanoidStation_in_jushen",
|
| 4 |
+
"repo_ids": null,
|
| 5 |
+
"roots": null,
|
| 6 |
+
"repo_id": null,
|
| 7 |
+
"root": null,
|
| 8 |
+
"episodes": null,
|
| 9 |
+
"image_transforms": {
|
| 10 |
+
"enable": false,
|
| 11 |
+
"max_num_transforms": 3,
|
| 12 |
+
"random_order": false,
|
| 13 |
+
"tfs": {
|
| 14 |
+
"brightness": {
|
| 15 |
+
"weight": 1.0,
|
| 16 |
+
"type": "ColorJitter",
|
| 17 |
+
"kwargs": {
|
| 18 |
+
"brightness": [
|
| 19 |
+
0.8,
|
| 20 |
+
1.2
|
| 21 |
+
]
|
| 22 |
+
}
|
| 23 |
+
},
|
| 24 |
+
"contrast": {
|
| 25 |
+
"weight": 1.0,
|
| 26 |
+
"type": "ColorJitter",
|
| 27 |
+
"kwargs": {
|
| 28 |
+
"contrast": [
|
| 29 |
+
0.8,
|
| 30 |
+
1.2
|
| 31 |
+
]
|
| 32 |
+
}
|
| 33 |
+
},
|
| 34 |
+
"saturation": {
|
| 35 |
+
"weight": 1.0,
|
| 36 |
+
"type": "ColorJitter",
|
| 37 |
+
"kwargs": {
|
| 38 |
+
"saturation": [
|
| 39 |
+
0.5,
|
| 40 |
+
1.5
|
| 41 |
+
]
|
| 42 |
+
}
|
| 43 |
+
},
|
| 44 |
+
"hue": {
|
| 45 |
+
"weight": 1.0,
|
| 46 |
+
"type": "ColorJitter",
|
| 47 |
+
"kwargs": {
|
| 48 |
+
"hue": [
|
| 49 |
+
-0.05,
|
| 50 |
+
0.05
|
| 51 |
+
]
|
| 52 |
+
}
|
| 53 |
+
},
|
| 54 |
+
"sharpness": {
|
| 55 |
+
"weight": 1.0,
|
| 56 |
+
"type": "SharpnessJitter",
|
| 57 |
+
"kwargs": {
|
| 58 |
+
"sharpness": [
|
| 59 |
+
0.5,
|
| 60 |
+
1.5
|
| 61 |
+
]
|
| 62 |
+
}
|
| 63 |
+
}
|
| 64 |
+
}
|
| 65 |
+
},
|
| 66 |
+
"local_files_only": false,
|
| 67 |
+
"use_imagenet_stats": true,
|
| 68 |
+
"video_backend": "pyav"
|
| 69 |
+
},
|
| 70 |
+
"env": null,
|
| 71 |
+
"policy": {
|
| 72 |
+
"type": "crossvla_stage2_plus",
|
| 73 |
+
"n_obs_steps": 2,
|
| 74 |
+
"action_sample_factor": 1,
|
| 75 |
+
"normalization_mapping": {
|
| 76 |
+
"VISUAL": "IDENTITY",
|
| 77 |
+
"STATE": "MEAN_STD",
|
| 78 |
+
"ACTION": "MEAN_STD"
|
| 79 |
+
},
|
| 80 |
+
"input_features": {
|
| 81 |
+
"observation.state": {
|
| 82 |
+
"type": "STATE",
|
| 83 |
+
"shape": [
|
| 84 |
+
32
|
| 85 |
+
]
|
| 86 |
+
},
|
| 87 |
+
"observation.images.image_0": {
|
| 88 |
+
"type": "VISUAL",
|
| 89 |
+
"shape": [
|
| 90 |
+
224,
|
| 91 |
+
3,
|
| 92 |
+
224
|
| 93 |
+
]
|
| 94 |
+
},
|
| 95 |
+
"observation.images.image_1": {
|
| 96 |
+
"type": "VISUAL",
|
| 97 |
+
"shape": [
|
| 98 |
+
224,
|
| 99 |
+
3,
|
| 100 |
+
224
|
| 101 |
+
]
|
| 102 |
+
},
|
| 103 |
+
"observation.images.image_2": {
|
| 104 |
+
"type": "VISUAL",
|
| 105 |
+
"shape": [
|
| 106 |
+
224,
|
| 107 |
+
3,
|
| 108 |
+
224
|
| 109 |
+
]
|
| 110 |
+
},
|
| 111 |
+
"observation.images.image_3": {
|
| 112 |
+
"type": "VISUAL",
|
| 113 |
+
"shape": [
|
| 114 |
+
224,
|
| 115 |
+
3,
|
| 116 |
+
224
|
| 117 |
+
]
|
| 118 |
+
},
|
| 119 |
+
"observation.images.image_wrist_0": {
|
| 120 |
+
"type": "VISUAL",
|
| 121 |
+
"shape": [
|
| 122 |
+
224,
|
| 123 |
+
3,
|
| 124 |
+
224
|
| 125 |
+
]
|
| 126 |
+
},
|
| 127 |
+
"observation.images.image_wrist_1": {
|
| 128 |
+
"type": "VISUAL",
|
| 129 |
+
"shape": [
|
| 130 |
+
224,
|
| 131 |
+
3,
|
| 132 |
+
224
|
| 133 |
+
]
|
| 134 |
+
},
|
| 135 |
+
"observation.state_is_pad": {
|
| 136 |
+
"type": "STATE",
|
| 137 |
+
"shape": [
|
| 138 |
+
2
|
| 139 |
+
]
|
| 140 |
+
},
|
| 141 |
+
"observation.images.image_0_is_pad": {
|
| 142 |
+
"type": "STATE",
|
| 143 |
+
"shape": [
|
| 144 |
+
2
|
| 145 |
+
]
|
| 146 |
+
},
|
| 147 |
+
"observation.images.image_1_is_pad": {
|
| 148 |
+
"type": "STATE",
|
| 149 |
+
"shape": [
|
| 150 |
+
2
|
| 151 |
+
]
|
| 152 |
+
},
|
| 153 |
+
"observation.images.image_2_is_pad": {
|
| 154 |
+
"type": "STATE",
|
| 155 |
+
"shape": [
|
| 156 |
+
2
|
| 157 |
+
]
|
| 158 |
+
},
|
| 159 |
+
"observation.images.image_3_is_pad": {
|
| 160 |
+
"type": "STATE",
|
| 161 |
+
"shape": [
|
| 162 |
+
2
|
| 163 |
+
]
|
| 164 |
+
},
|
| 165 |
+
"observation.images.image_wrist_0_is_pad": {
|
| 166 |
+
"type": "STATE",
|
| 167 |
+
"shape": [
|
| 168 |
+
2
|
| 169 |
+
]
|
| 170 |
+
},
|
| 171 |
+
"observation.images.image_wrist_1_is_pad": {
|
| 172 |
+
"type": "STATE",
|
| 173 |
+
"shape": [
|
| 174 |
+
2
|
| 175 |
+
]
|
| 176 |
+
}
|
| 177 |
+
},
|
| 178 |
+
"output_features": {
|
| 179 |
+
"action": {
|
| 180 |
+
"type": "ACTION",
|
| 181 |
+
"shape": [
|
| 182 |
+
32
|
| 183 |
+
]
|
| 184 |
+
},
|
| 185 |
+
"action_is_pad": {
|
| 186 |
+
"type": "ACTION",
|
| 187 |
+
"shape": [
|
| 188 |
+
50
|
| 189 |
+
]
|
| 190 |
+
}
|
| 191 |
+
},
|
| 192 |
+
"stage1_pretrained_path": "/media/jushen/bamboo-fan/Save/crossvla/stage1_4dataset_10_8_bs12_kl_cb256_0528_stat/checkpoints/275000/pretrained_model/",
|
| 193 |
+
"stage2_pretrained_path": "None",
|
| 194 |
+
"stage2_latent_image_token_check": false,
|
| 195 |
+
"dataset_stats_generate": true,
|
| 196 |
+
"heterogeneous": true,
|
| 197 |
+
"split_dataset": true,
|
| 198 |
+
"real_robot_dev": false,
|
| 199 |
+
"image_interval_steps": 50,
|
| 200 |
+
"action_latent_token_num": 13,
|
| 201 |
+
"mformer_hidden_size": 768,
|
| 202 |
+
"decoder_hidden_size": 768,
|
| 203 |
+
"codebook_embed_dim": 256,
|
| 204 |
+
"codebook_k_size": 256,
|
| 205 |
+
"action_chunk_size": 50,
|
| 206 |
+
"chunk_size": 50,
|
| 207 |
+
"n_action_steps": 50,
|
| 208 |
+
"resampler": true,
|
| 209 |
+
"resampler_dim": 2048,
|
| 210 |
+
"resampler_depth": 3,
|
| 211 |
+
"resampler_dim_head": 128,
|
| 212 |
+
"resampler_heads": 4,
|
| 213 |
+
"resampler_num_media_embeds": 1,
|
| 214 |
+
"resampler_num_latents": 9,
|
| 215 |
+
"max_state_dim": 32,
|
| 216 |
+
"max_action_dim": 32,
|
| 217 |
+
"resize_imgs_with_padding": [
|
| 218 |
+
224,
|
| 219 |
+
224
|
| 220 |
+
],
|
| 221 |
+
"empty_cameras": 0,
|
| 222 |
+
"adapt_to_pi_aloha": false,
|
| 223 |
+
"use_delta_joint_actions_aloha": false,
|
| 224 |
+
"tokenizer_max_length": 48,
|
| 225 |
+
"proj_width": 1024,
|
| 226 |
+
"num_steps": 10,
|
| 227 |
+
"use_cache": true,
|
| 228 |
+
"attention_implementation": "eager",
|
| 229 |
+
"freeze_vision_encoder": false,
|
| 230 |
+
"freeze_language_encoder": true,
|
| 231 |
+
"train_expert_only": false,
|
| 232 |
+
"train_state_proj": true,
|
| 233 |
+
"optimizer_lr": 0.0001,
|
| 234 |
+
"optimizer_betas": [
|
| 235 |
+
0.9,
|
| 236 |
+
0.95
|
| 237 |
+
],
|
| 238 |
+
"optimizer_eps": 1e-08,
|
| 239 |
+
"optimizer_weight_decay": 1e-10,
|
| 240 |
+
"scheduler_warmup_steps": 5000,
|
| 241 |
+
"scheduler_decay_steps": 300000,
|
| 242 |
+
"scheduler_decay_lr": 1e-06
|
| 243 |
+
},
|
| 244 |
+
"output_dir": "/media/jushen/bamboo-fan/Save/crossvla/dev/stage2_dev_pretrain_10_8_bs8_stage1_275k",
|
| 245 |
+
"job_name": "stage2_dev_pretrain_10_8_bs8_stage1_275k",
|
| 246 |
+
"resume": true,
|
| 247 |
+
"device": "cuda",
|
| 248 |
+
"use_amp": false,
|
| 249 |
+
"seed": 1000,
|
| 250 |
+
"num_workers": 4,
|
| 251 |
+
"batch_size": 8,
|
| 252 |
+
"gradient_accumulation_steps": 1,
|
| 253 |
+
"steps": 300000,
|
| 254 |
+
"eval_freq": 20000,
|
| 255 |
+
"log_freq": 200,
|
| 256 |
+
"save_checkpoint": true,
|
| 257 |
+
"save_freq": 10000,
|
| 258 |
+
"use_policy_training_preset": true,
|
| 259 |
+
"optimizer": {
|
| 260 |
+
"type": "adamw",
|
| 261 |
+
"lr": 0.0001,
|
| 262 |
+
"weight_decay": 1e-10,
|
| 263 |
+
"grad_clip_norm": 10.0,
|
| 264 |
+
"betas": [
|
| 265 |
+
0.9,
|
| 266 |
+
0.95
|
| 267 |
+
],
|
| 268 |
+
"eps": 1e-08
|
| 269 |
+
},
|
| 270 |
+
"scheduler": {
|
| 271 |
+
"type": "cosine_decay_with_warmup",
|
| 272 |
+
"num_warmup_steps": 5000,
|
| 273 |
+
"num_decay_steps": 300000,
|
| 274 |
+
"peak_lr": 0.0001,
|
| 275 |
+
"decay_lr": 1e-06
|
| 276 |
+
},
|
| 277 |
+
"eval": {
|
| 278 |
+
"n_episodes": 50,
|
| 279 |
+
"batch_size": 50,
|
| 280 |
+
"use_async_envs": false
|
| 281 |
+
},
|
| 282 |
+
"wandb": {
|
| 283 |
+
"enable": true,
|
| 284 |
+
"disable_artifact": true,
|
| 285 |
+
"project": "crossvla_dev",
|
| 286 |
+
"entity": null,
|
| 287 |
+
"notes": null
|
| 288 |
+
}
|
| 289 |
+
}
|