Upload policy weights, train config and readme
Browse files- README.md +62 -0
- config.json +103 -0
- model.safetensors +3 -0
- train_config.json +215 -0
README.md
ADDED
|
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
datasets: li-whale/fold_bag
|
| 3 |
+
library_name: lerobot
|
| 4 |
+
license: apache-2.0
|
| 5 |
+
model_name: pi0fast
|
| 6 |
+
pipeline_tag: robotics
|
| 7 |
+
tags:
|
| 8 |
+
- lerobot
|
| 9 |
+
- pi0fast
|
| 10 |
+
- robotics
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
# Model Card for pi0fast
|
| 14 |
+
|
| 15 |
+
<!-- Provide a quick summary of what the model is/does. -->
|
| 16 |
+
|
| 17 |
+
|
| 18 |
+
[Pi0-Fast](https://huggingface.co/papers/2501.09747) is a variant of Pi0 that uses a new tokenization method called FAST, which enables training of an autoregressive vision-language-action policy for high-frequency robotic tasks with improved performance and reduced training time.
|
| 19 |
+
|
| 20 |
+
|
| 21 |
+
This policy has been trained and pushed to the Hub using [LeRobot](https://github.com/huggingface/lerobot).
|
| 22 |
+
See the full documentation at [LeRobot Docs](https://huggingface.co/docs/lerobot/index).
|
| 23 |
+
|
| 24 |
+
---
|
| 25 |
+
|
| 26 |
+
## How to Get Started with the Model
|
| 27 |
+
|
| 28 |
+
For a complete walkthrough, see the [training guide](https://huggingface.co/docs/lerobot/il_robots#train-a-policy).
|
| 29 |
+
Below is the short version on how to train and run inference/eval:
|
| 30 |
+
|
| 31 |
+
### Train from scratch
|
| 32 |
+
|
| 33 |
+
```bash
|
| 34 |
+
lerobot-train \
|
| 35 |
+
--dataset.repo_id=${HF_USER}/<dataset> \
|
| 36 |
+
--policy.type=act \
|
| 37 |
+
--output_dir=outputs/train/<desired_policy_repo_id> \
|
| 38 |
+
--job_name=lerobot_training \
|
| 39 |
+
--policy.device=cuda \
|
| 40 |
+
--policy.repo_id=${HF_USER}/<desired_policy_repo_id>
|
| 41 |
+
--wandb.enable=true
|
| 42 |
+
```
|
| 43 |
+
|
| 44 |
+
_Writes checkpoints to `outputs/train/<desired_policy_repo_id>/checkpoints/`._
|
| 45 |
+
|
| 46 |
+
### Evaluate the policy/run inference
|
| 47 |
+
|
| 48 |
+
```bash
|
| 49 |
+
lerobot-record \
|
| 50 |
+
--robot.type=so100_follower \
|
| 51 |
+
--dataset.repo_id=<hf_user>/eval_<dataset> \
|
| 52 |
+
--policy.path=<hf_user>/<desired_policy_repo_id> \
|
| 53 |
+
--episodes=10
|
| 54 |
+
```
|
| 55 |
+
|
| 56 |
+
Prefix the dataset repo with **eval\_** and supply `--policy.path` pointing to a local or hub checkpoint.
|
| 57 |
+
|
| 58 |
+
---
|
| 59 |
+
|
| 60 |
+
## Model Details
|
| 61 |
+
|
| 62 |
+
- **License:** apache-2.0
|
config.json
ADDED
|
@@ -0,0 +1,103 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"type": "pi0fast",
|
| 3 |
+
"n_obs_steps": 1,
|
| 4 |
+
"normalization_mapping": {
|
| 5 |
+
"VISUAL": "IDENTITY",
|
| 6 |
+
"STATE": "MEAN_STD",
|
| 7 |
+
"ACTION": "MEAN_STD"
|
| 8 |
+
},
|
| 9 |
+
"input_features": {
|
| 10 |
+
"observation.state": {
|
| 11 |
+
"type": "STATE",
|
| 12 |
+
"shape": [
|
| 13 |
+
32
|
| 14 |
+
]
|
| 15 |
+
},
|
| 16 |
+
"observations.images.left_wrist": {
|
| 17 |
+
"type": "VISUAL",
|
| 18 |
+
"shape": [
|
| 19 |
+
480,
|
| 20 |
+
640,
|
| 21 |
+
3
|
| 22 |
+
]
|
| 23 |
+
},
|
| 24 |
+
"observations.images.right_wrist": {
|
| 25 |
+
"type": "VISUAL",
|
| 26 |
+
"shape": [
|
| 27 |
+
480,
|
| 28 |
+
640,
|
| 29 |
+
3
|
| 30 |
+
]
|
| 31 |
+
},
|
| 32 |
+
"observations.images.top": {
|
| 33 |
+
"type": "VISUAL",
|
| 34 |
+
"shape": [
|
| 35 |
+
480,
|
| 36 |
+
640,
|
| 37 |
+
3
|
| 38 |
+
]
|
| 39 |
+
},
|
| 40 |
+
"observations.qpos": {
|
| 41 |
+
"type": "STATE",
|
| 42 |
+
"shape": [
|
| 43 |
+
16
|
| 44 |
+
]
|
| 45 |
+
},
|
| 46 |
+
"observations.qvel": {
|
| 47 |
+
"type": "STATE",
|
| 48 |
+
"shape": [
|
| 49 |
+
16
|
| 50 |
+
]
|
| 51 |
+
}
|
| 52 |
+
},
|
| 53 |
+
"output_features": {
|
| 54 |
+
"action": {
|
| 55 |
+
"type": "ACTION",
|
| 56 |
+
"shape": [
|
| 57 |
+
16
|
| 58 |
+
]
|
| 59 |
+
}
|
| 60 |
+
},
|
| 61 |
+
"device": "cuda",
|
| 62 |
+
"use_amp": false,
|
| 63 |
+
"push_to_hub": true,
|
| 64 |
+
"repo_id": "li-whale/pi0_fold_bag",
|
| 65 |
+
"private": null,
|
| 66 |
+
"tags": null,
|
| 67 |
+
"license": null,
|
| 68 |
+
"chunk_size": 10,
|
| 69 |
+
"n_action_steps": 5,
|
| 70 |
+
"max_state_dim": 32,
|
| 71 |
+
"max_action_dim": 32,
|
| 72 |
+
"resize_imgs_with_padding": [
|
| 73 |
+
224,
|
| 74 |
+
224
|
| 75 |
+
],
|
| 76 |
+
"interpolate_like_pi": false,
|
| 77 |
+
"empty_cameras": 0,
|
| 78 |
+
"adapt_to_pi_aloha": false,
|
| 79 |
+
"use_delta_joint_actions_aloha": false,
|
| 80 |
+
"tokenizer_max_length": 48,
|
| 81 |
+
"proj_width": 1024,
|
| 82 |
+
"max_decoding_steps": 256,
|
| 83 |
+
"fast_skip_tokens": 128,
|
| 84 |
+
"max_input_seq_len": 256,
|
| 85 |
+
"use_cache": true,
|
| 86 |
+
"freeze_vision_encoder": true,
|
| 87 |
+
"freeze_lm_head": true,
|
| 88 |
+
"optimizer_lr": 0.0001,
|
| 89 |
+
"optimizer_betas": [
|
| 90 |
+
0.9,
|
| 91 |
+
0.95
|
| 92 |
+
],
|
| 93 |
+
"optimizer_eps": 1e-08,
|
| 94 |
+
"optimizer_weight_decay": 1e-05,
|
| 95 |
+
"scheduler_warmup_steps": 1000,
|
| 96 |
+
"scheduler_decay_steps": 30000,
|
| 97 |
+
"scheduler_decay_lr": 2.5e-06,
|
| 98 |
+
"checkpoint_path": null,
|
| 99 |
+
"padding_side": "right",
|
| 100 |
+
"precision": "bfloat16",
|
| 101 |
+
"grad_clip_norm": 1,
|
| 102 |
+
"relaxed_action_decoding": true
|
| 103 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7eeebd9288bcfe046bae84a71dd87d9faa2da758eb2cc9ca77c43baa5b7bbcd7
|
| 3 |
+
size 5846766808
|
train_config.json
ADDED
|
@@ -0,0 +1,215 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"dataset": {
|
| 3 |
+
"repo_id": "li-whale/fold_bag",
|
| 4 |
+
"root": null,
|
| 5 |
+
"episodes": null,
|
| 6 |
+
"image_transforms": {
|
| 7 |
+
"enable": false,
|
| 8 |
+
"max_num_transforms": 3,
|
| 9 |
+
"random_order": false,
|
| 10 |
+
"tfs": {
|
| 11 |
+
"brightness": {
|
| 12 |
+
"weight": 1.0,
|
| 13 |
+
"type": "ColorJitter",
|
| 14 |
+
"kwargs": {
|
| 15 |
+
"brightness": [
|
| 16 |
+
0.8,
|
| 17 |
+
1.2
|
| 18 |
+
]
|
| 19 |
+
}
|
| 20 |
+
},
|
| 21 |
+
"contrast": {
|
| 22 |
+
"weight": 1.0,
|
| 23 |
+
"type": "ColorJitter",
|
| 24 |
+
"kwargs": {
|
| 25 |
+
"contrast": [
|
| 26 |
+
0.8,
|
| 27 |
+
1.2
|
| 28 |
+
]
|
| 29 |
+
}
|
| 30 |
+
},
|
| 31 |
+
"saturation": {
|
| 32 |
+
"weight": 1.0,
|
| 33 |
+
"type": "ColorJitter",
|
| 34 |
+
"kwargs": {
|
| 35 |
+
"saturation": [
|
| 36 |
+
0.5,
|
| 37 |
+
1.5
|
| 38 |
+
]
|
| 39 |
+
}
|
| 40 |
+
},
|
| 41 |
+
"hue": {
|
| 42 |
+
"weight": 1.0,
|
| 43 |
+
"type": "ColorJitter",
|
| 44 |
+
"kwargs": {
|
| 45 |
+
"hue": [
|
| 46 |
+
-0.05,
|
| 47 |
+
0.05
|
| 48 |
+
]
|
| 49 |
+
}
|
| 50 |
+
},
|
| 51 |
+
"sharpness": {
|
| 52 |
+
"weight": 1.0,
|
| 53 |
+
"type": "SharpnessJitter",
|
| 54 |
+
"kwargs": {
|
| 55 |
+
"sharpness": [
|
| 56 |
+
0.5,
|
| 57 |
+
1.5
|
| 58 |
+
]
|
| 59 |
+
}
|
| 60 |
+
}
|
| 61 |
+
}
|
| 62 |
+
},
|
| 63 |
+
"revision": null,
|
| 64 |
+
"use_imagenet_stats": true,
|
| 65 |
+
"video_backend": "torchcodec"
|
| 66 |
+
},
|
| 67 |
+
"env": null,
|
| 68 |
+
"policy": {
|
| 69 |
+
"type": "pi0fast",
|
| 70 |
+
"n_obs_steps": 1,
|
| 71 |
+
"normalization_mapping": {
|
| 72 |
+
"VISUAL": "IDENTITY",
|
| 73 |
+
"STATE": "MEAN_STD",
|
| 74 |
+
"ACTION": "MEAN_STD"
|
| 75 |
+
},
|
| 76 |
+
"input_features": {
|
| 77 |
+
"observation.state": {
|
| 78 |
+
"type": "STATE",
|
| 79 |
+
"shape": [
|
| 80 |
+
32
|
| 81 |
+
]
|
| 82 |
+
},
|
| 83 |
+
"observations.images.left_wrist": {
|
| 84 |
+
"type": "VISUAL",
|
| 85 |
+
"shape": [
|
| 86 |
+
480,
|
| 87 |
+
640,
|
| 88 |
+
3
|
| 89 |
+
]
|
| 90 |
+
},
|
| 91 |
+
"observations.images.right_wrist": {
|
| 92 |
+
"type": "VISUAL",
|
| 93 |
+
"shape": [
|
| 94 |
+
480,
|
| 95 |
+
640,
|
| 96 |
+
3
|
| 97 |
+
]
|
| 98 |
+
},
|
| 99 |
+
"observations.images.top": {
|
| 100 |
+
"type": "VISUAL",
|
| 101 |
+
"shape": [
|
| 102 |
+
480,
|
| 103 |
+
640,
|
| 104 |
+
3
|
| 105 |
+
]
|
| 106 |
+
},
|
| 107 |
+
"observations.qpos": {
|
| 108 |
+
"type": "STATE",
|
| 109 |
+
"shape": [
|
| 110 |
+
16
|
| 111 |
+
]
|
| 112 |
+
},
|
| 113 |
+
"observations.qvel": {
|
| 114 |
+
"type": "STATE",
|
| 115 |
+
"shape": [
|
| 116 |
+
16
|
| 117 |
+
]
|
| 118 |
+
}
|
| 119 |
+
},
|
| 120 |
+
"output_features": {
|
| 121 |
+
"action": {
|
| 122 |
+
"type": "ACTION",
|
| 123 |
+
"shape": [
|
| 124 |
+
16
|
| 125 |
+
]
|
| 126 |
+
}
|
| 127 |
+
},
|
| 128 |
+
"device": "cuda",
|
| 129 |
+
"use_amp": false,
|
| 130 |
+
"push_to_hub": true,
|
| 131 |
+
"repo_id": "li-whale/pi0_fold_bag",
|
| 132 |
+
"private": null,
|
| 133 |
+
"tags": null,
|
| 134 |
+
"license": null,
|
| 135 |
+
"chunk_size": 10,
|
| 136 |
+
"n_action_steps": 5,
|
| 137 |
+
"max_state_dim": 32,
|
| 138 |
+
"max_action_dim": 32,
|
| 139 |
+
"resize_imgs_with_padding": [
|
| 140 |
+
224,
|
| 141 |
+
224
|
| 142 |
+
],
|
| 143 |
+
"interpolate_like_pi": false,
|
| 144 |
+
"empty_cameras": 0,
|
| 145 |
+
"adapt_to_pi_aloha": false,
|
| 146 |
+
"use_delta_joint_actions_aloha": false,
|
| 147 |
+
"tokenizer_max_length": 48,
|
| 148 |
+
"proj_width": 1024,
|
| 149 |
+
"max_decoding_steps": 256,
|
| 150 |
+
"fast_skip_tokens": 128,
|
| 151 |
+
"max_input_seq_len": 256,
|
| 152 |
+
"use_cache": true,
|
| 153 |
+
"freeze_vision_encoder": true,
|
| 154 |
+
"freeze_lm_head": true,
|
| 155 |
+
"optimizer_lr": 0.0001,
|
| 156 |
+
"optimizer_betas": [
|
| 157 |
+
0.9,
|
| 158 |
+
0.95
|
| 159 |
+
],
|
| 160 |
+
"optimizer_eps": 1e-08,
|
| 161 |
+
"optimizer_weight_decay": 1e-05,
|
| 162 |
+
"scheduler_warmup_steps": 1000,
|
| 163 |
+
"scheduler_decay_steps": 30000,
|
| 164 |
+
"scheduler_decay_lr": 2.5e-06,
|
| 165 |
+
"checkpoint_path": null,
|
| 166 |
+
"padding_side": "right",
|
| 167 |
+
"precision": "bfloat16",
|
| 168 |
+
"grad_clip_norm": 1,
|
| 169 |
+
"relaxed_action_decoding": true
|
| 170 |
+
},
|
| 171 |
+
"output_dir": "outputs/train/pi0_fold_bag",
|
| 172 |
+
"job_name": "fold_bag",
|
| 173 |
+
"resume": false,
|
| 174 |
+
"seed": 1000,
|
| 175 |
+
"num_workers": 4,
|
| 176 |
+
"batch_size": 1,
|
| 177 |
+
"steps": 10000,
|
| 178 |
+
"eval_freq": 20000,
|
| 179 |
+
"log_freq": 200,
|
| 180 |
+
"save_checkpoint": true,
|
| 181 |
+
"save_freq": 20000,
|
| 182 |
+
"use_policy_training_preset": true,
|
| 183 |
+
"optimizer": {
|
| 184 |
+
"type": "adamw",
|
| 185 |
+
"lr": 0.0001,
|
| 186 |
+
"weight_decay": 1e-05,
|
| 187 |
+
"grad_clip_norm": 1,
|
| 188 |
+
"betas": [
|
| 189 |
+
0.9,
|
| 190 |
+
0.95
|
| 191 |
+
],
|
| 192 |
+
"eps": 1e-08
|
| 193 |
+
},
|
| 194 |
+
"scheduler": {
|
| 195 |
+
"type": "cosine_decay_with_warmup",
|
| 196 |
+
"num_warmup_steps": 1000,
|
| 197 |
+
"num_decay_steps": 30000,
|
| 198 |
+
"peak_lr": 0.0001,
|
| 199 |
+
"decay_lr": 2.5e-06
|
| 200 |
+
},
|
| 201 |
+
"eval": {
|
| 202 |
+
"n_episodes": 50,
|
| 203 |
+
"batch_size": 50,
|
| 204 |
+
"use_async_envs": false
|
| 205 |
+
},
|
| 206 |
+
"wandb": {
|
| 207 |
+
"enable": false,
|
| 208 |
+
"disable_artifact": false,
|
| 209 |
+
"project": "lerobot",
|
| 210 |
+
"entity": null,
|
| 211 |
+
"notes": null,
|
| 212 |
+
"run_id": null,
|
| 213 |
+
"mode": null
|
| 214 |
+
}
|
| 215 |
+
}
|