Instructions to use trytoycon/diffusion_pick_place_50_30ksteps with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- LeRobot
How to use trytoycon/diffusion_pick_place_50_30ksteps with LeRobot:
- Notebooks
- Google Colab
- Kaggle
Upload policy weights, train config and readme
Browse files- config.json +10 -9
- model.safetensors +2 -2
- train_config.json +22 -19
config.json
CHANGED
|
@@ -42,8 +42,8 @@
|
|
| 42 |
"tags": null,
|
| 43 |
"license": null,
|
| 44 |
"pretrained_path": null,
|
| 45 |
-
"horizon":
|
| 46 |
-
"n_action_steps":
|
| 47 |
"normalization_mapping": {
|
| 48 |
"VISUAL": "MEAN_STD",
|
| 49 |
"STATE": "MIN_MAX",
|
|
@@ -51,15 +51,14 @@
|
|
| 51 |
},
|
| 52 |
"drop_n_last_frames": 7,
|
| 53 |
"vision_backbone": "resnet18",
|
| 54 |
-
"
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
],
|
| 58 |
"crop_is_random": true,
|
| 59 |
-
"pretrained_backbone_weights":
|
| 60 |
-
"use_group_norm":
|
| 61 |
"spatial_softmax_num_keypoints": 32,
|
| 62 |
-
"use_separate_rgb_encoder_per_camera":
|
| 63 |
"down_dims": [
|
| 64 |
512,
|
| 65 |
1024,
|
|
@@ -78,6 +77,8 @@
|
|
| 78 |
"clip_sample": true,
|
| 79 |
"clip_sample_range": 1.0,
|
| 80 |
"num_inference_steps": null,
|
|
|
|
|
|
|
| 81 |
"do_mask_loss_for_padding": false,
|
| 82 |
"optimizer_lr": 0.0001,
|
| 83 |
"optimizer_betas": [
|
|
|
|
| 42 |
"tags": null,
|
| 43 |
"license": null,
|
| 44 |
"pretrained_path": null,
|
| 45 |
+
"horizon": 64,
|
| 46 |
+
"n_action_steps": 32,
|
| 47 |
"normalization_mapping": {
|
| 48 |
"VISUAL": "MEAN_STD",
|
| 49 |
"STATE": "MIN_MAX",
|
|
|
|
| 51 |
},
|
| 52 |
"drop_n_last_frames": 7,
|
| 53 |
"vision_backbone": "resnet18",
|
| 54 |
+
"resize_shape": null,
|
| 55 |
+
"crop_ratio": 1.0,
|
| 56 |
+
"crop_shape": null,
|
|
|
|
| 57 |
"crop_is_random": true,
|
| 58 |
+
"pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
|
| 59 |
+
"use_group_norm": false,
|
| 60 |
"spatial_softmax_num_keypoints": 32,
|
| 61 |
+
"use_separate_rgb_encoder_per_camera": true,
|
| 62 |
"down_dims": [
|
| 63 |
512,
|
| 64 |
1024,
|
|
|
|
| 77 |
"clip_sample": true,
|
| 78 |
"clip_sample_range": 1.0,
|
| 79 |
"num_inference_steps": null,
|
| 80 |
+
"compile_model": false,
|
| 81 |
+
"compile_mode": "reduce-overhead",
|
| 82 |
"do_mask_loss_for_padding": false,
|
| 83 |
"optimizer_lr": 0.0001,
|
| 84 |
"optimizer_betas": [
|
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:86b6e43eeeca88611d6af208f3bc364959075897aa07fc6b2eed542d0597474e
|
| 3 |
+
size 1111408608
|
train_config.json
CHANGED
|
@@ -77,6 +77,7 @@
|
|
| 77 |
"revision": null,
|
| 78 |
"use_imagenet_stats": true,
|
| 79 |
"video_backend": "torchcodec",
|
|
|
|
| 80 |
"streaming": false
|
| 81 |
},
|
| 82 |
"env": null,
|
|
@@ -124,8 +125,8 @@
|
|
| 124 |
"tags": null,
|
| 125 |
"license": null,
|
| 126 |
"pretrained_path": null,
|
| 127 |
-
"horizon":
|
| 128 |
-
"n_action_steps":
|
| 129 |
"normalization_mapping": {
|
| 130 |
"VISUAL": "MEAN_STD",
|
| 131 |
"STATE": "MIN_MAX",
|
|
@@ -133,15 +134,14 @@
|
|
| 133 |
},
|
| 134 |
"drop_n_last_frames": 7,
|
| 135 |
"vision_backbone": "resnet18",
|
| 136 |
-
"
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
],
|
| 140 |
"crop_is_random": true,
|
| 141 |
-
"pretrained_backbone_weights":
|
| 142 |
-
"use_group_norm":
|
| 143 |
"spatial_softmax_num_keypoints": 32,
|
| 144 |
-
"use_separate_rgb_encoder_per_camera":
|
| 145 |
"down_dims": [
|
| 146 |
512,
|
| 147 |
1024,
|
|
@@ -160,6 +160,8 @@
|
|
| 160 |
"clip_sample": true,
|
| 161 |
"clip_sample_range": 1.0,
|
| 162 |
"num_inference_steps": null,
|
|
|
|
|
|
|
| 163 |
"do_mask_loss_for_padding": false,
|
| 164 |
"optimizer_lr": 0.0001,
|
| 165 |
"optimizer_betas": [
|
|
@@ -171,12 +173,16 @@
|
|
| 171 |
"scheduler_name": "cosine",
|
| 172 |
"scheduler_warmup_steps": 500
|
| 173 |
},
|
| 174 |
-
"
|
| 175 |
-
"
|
|
|
|
| 176 |
"resume": false,
|
| 177 |
"seed": 1000,
|
|
|
|
| 178 |
"num_workers": 4,
|
| 179 |
"batch_size": 4,
|
|
|
|
|
|
|
| 180 |
"steps": 30000,
|
| 181 |
"eval_freq": 20000,
|
| 182 |
"log_freq": 200,
|
|
@@ -202,8 +208,8 @@
|
|
| 202 |
},
|
| 203 |
"eval": {
|
| 204 |
"n_episodes": 50,
|
| 205 |
-
"batch_size":
|
| 206 |
-
"use_async_envs":
|
| 207 |
},
|
| 208 |
"wandb": {
|
| 209 |
"enable": false,
|
|
@@ -212,14 +218,11 @@
|
|
| 212 |
"entity": null,
|
| 213 |
"notes": null,
|
| 214 |
"run_id": null,
|
| 215 |
-
"mode": null
|
|
|
|
| 216 |
},
|
| 217 |
"peft": null,
|
| 218 |
-
"
|
| 219 |
-
"rabc_progress_path": null,
|
| 220 |
-
"rabc_kappa": 0.01,
|
| 221 |
-
"rabc_epsilon": 1e-06,
|
| 222 |
-
"rabc_head_mode": "sparse",
|
| 223 |
"rename_map": {},
|
| 224 |
"checkpoint_path": null
|
| 225 |
}
|
|
|
|
| 77 |
"revision": null,
|
| 78 |
"use_imagenet_stats": true,
|
| 79 |
"video_backend": "torchcodec",
|
| 80 |
+
"return_uint8": false,
|
| 81 |
"streaming": false
|
| 82 |
},
|
| 83 |
"env": null,
|
|
|
|
| 125 |
"tags": null,
|
| 126 |
"license": null,
|
| 127 |
"pretrained_path": null,
|
| 128 |
+
"horizon": 64,
|
| 129 |
+
"n_action_steps": 32,
|
| 130 |
"normalization_mapping": {
|
| 131 |
"VISUAL": "MEAN_STD",
|
| 132 |
"STATE": "MIN_MAX",
|
|
|
|
| 134 |
},
|
| 135 |
"drop_n_last_frames": 7,
|
| 136 |
"vision_backbone": "resnet18",
|
| 137 |
+
"resize_shape": null,
|
| 138 |
+
"crop_ratio": 1.0,
|
| 139 |
+
"crop_shape": null,
|
|
|
|
| 140 |
"crop_is_random": true,
|
| 141 |
+
"pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
|
| 142 |
+
"use_group_norm": false,
|
| 143 |
"spatial_softmax_num_keypoints": 32,
|
| 144 |
+
"use_separate_rgb_encoder_per_camera": true,
|
| 145 |
"down_dims": [
|
| 146 |
512,
|
| 147 |
1024,
|
|
|
|
| 160 |
"clip_sample": true,
|
| 161 |
"clip_sample_range": 1.0,
|
| 162 |
"num_inference_steps": null,
|
| 163 |
+
"compile_model": false,
|
| 164 |
+
"compile_mode": "reduce-overhead",
|
| 165 |
"do_mask_loss_for_padding": false,
|
| 166 |
"optimizer_lr": 0.0001,
|
| 167 |
"optimizer_betas": [
|
|
|
|
| 173 |
"scheduler_name": "cosine",
|
| 174 |
"scheduler_warmup_steps": 500
|
| 175 |
},
|
| 176 |
+
"reward_model": null,
|
| 177 |
+
"output_dir": "outputs/diffusion_pick_place_dish_50_30ksteps_lerobot0.5.2",
|
| 178 |
+
"job_name": "diffusion_pick_place_dish_v2_30k",
|
| 179 |
"resume": false,
|
| 180 |
"seed": 1000,
|
| 181 |
+
"cudnn_deterministic": false,
|
| 182 |
"num_workers": 4,
|
| 183 |
"batch_size": 4,
|
| 184 |
+
"prefetch_factor": 4,
|
| 185 |
+
"persistent_workers": true,
|
| 186 |
"steps": 30000,
|
| 187 |
"eval_freq": 20000,
|
| 188 |
"log_freq": 200,
|
|
|
|
| 208 |
},
|
| 209 |
"eval": {
|
| 210 |
"n_episodes": 50,
|
| 211 |
+
"batch_size": 16,
|
| 212 |
+
"use_async_envs": true
|
| 213 |
},
|
| 214 |
"wandb": {
|
| 215 |
"enable": false,
|
|
|
|
| 218 |
"entity": null,
|
| 219 |
"notes": null,
|
| 220 |
"run_id": null,
|
| 221 |
+
"mode": null,
|
| 222 |
+
"add_tags": true
|
| 223 |
},
|
| 224 |
"peft": null,
|
| 225 |
+
"sample_weighting": null,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 226 |
"rename_map": {},
|
| 227 |
"checkpoint_path": null
|
| 228 |
}
|