modified: README.md

new file: __init__.py
modified: config.json
deleted: config_surf.json
deleted: fingernet_surf.onnx
renamed: fingernet.onnx -> model.onnx
new file: model.safetensors
new file: modeling.py

README.md

@@ -1,6 +1,15 @@
 ---
 license: bsd-3-clause
 pipeline_tag: robotics
+tags:
+  - fingernet
+  - asfinger
+  - multimodal
+  - onnx
+  - pytorch
+library_name: transformers
+datasets:
+  - asRobotics/fingernet-100k
 ---
 
 # Model Card for FingerNet
@@ -16,10 +25,12 @@ pipeline_tag: robotics
 
 ## Model Description
 
-FingerNet is a dual-branch MLP model designed for the asFinger. It can predict both 6D force and 3D shape (mesh nodes) from the 6D motion of the asFinger.
+FingerNet is an MLP model designed for the asFinger. It can predict both 6D force and 3D shape (mesh nodes) from the 6D motion of the asFinger.
 
-- Developer: Xudong Han, Ning Guo, Xiaobo Liu, Tianyu Wu, Fang Wan, and Chaoyang Song. See [GitHub repo](https://github.com/AncoraSpring/metafinger).
-- Model type: Dual-branch MLP
+Try it out on the [Spaces Demo](https://huggingface.co/spaces/asRobotics/fingernet-demo)!
+
+- Developer: Xudong Han, Ning Guo, Xiaobo Liu, Tianyu Wu, Fang Wan, and Chaoyang Song.
+- Model type: MLP
 - License: BSD-3-Clause
 - Resources for more information:
   - [Project page](https://doc.ancoraspring.com/asfinger)
@@ -31,27 +42,39 @@ This model is intended for researchers and developers working in robotics and ta
 
 See the [project page](https://doc.ancoraspring.com/asfinger) for more details.
 
+To load the model:
+
+```python
+# Example code to load safetensors
+from transformers import AutoModel
+
+model = AutoModel.from_pretrained("asRobotics/fingernet", trust_remote_code=True)
+x = torch.zeros((1, 6))  # Example input: batch size of 1, 6D motion
+output = model(x)
+```
+
+Or to load the ONNX version:
+
+```python
+# Example code to load onnx
+import onnxruntime as ort
+import numpy as np
+from huggingface_hub import hf_hub_download
+
+onnx_model_path = hf_hub_download(repo_id="asRobotics/fingernet", filename="model.onnx")
+ort_session = ort.InferenceSession(onnx_model_path)
+x = np.zeros((1, 6)).astype(np.float32)  # Example input: batch size of 1, 6D motion
+outputs = ort_session.run(None, {"motion": x})
+```
+
 ## Training Data
 
-The model was trained on the asFinger Dataset, which includes a variety of motion, force, and shape data collected by simulation and real-world experiments.
+The model was trained on the [FingerNet-100K](https://huggingface.co/datasets/asRobotics/fingernet-100k), which includes a variety of motion, force, and shape data collected by finite element simulations.
 
 ## Citation
 
 If you use this model in your research, please cite the following papers:
 
-```bibtex
-@article{guo2024proprioceptive,
-  title={Proprioceptive State Estimation for Amphibious Tactile Sensing},
-  author={Guo, Ning and Han, Xudong and Zhong, Shuqiao and Zhou, Zhiyuan and Lin, Jian and Dai, Jian S and Wan, Fang and Song, Chaoyang},
-  journal={IEEE Transactions on Robotics},
-  volume={40},
-  pages={4684-4698},
-  year={2024},
-  publisher={IEEE},
-  doi={10.1109/TRO.2024.3463509}
-}
-```
-
 ```bibtex
 @article{liu2024proprioceptive,
   title={Proprioceptive learning with soft polyhedral networks},
@@ -66,14 +89,13 @@ If you use this model in your research, please cite the following papers:
 }
 ```
 
+[](https://arxiv.org/abs/2308.08538)
+
 ```bibtex
-@article{han2025anchoring,
-    title={Anchoring Morphological Representations Unlocks Latent Proprioception in Soft Robots},
-    author={Han, Xudong and Guo, Ning and Xu, Ronghan and Wan, Fang and Song, Chaoyang},
-    journal={Advanced Intelligent Systems},
-    pages={2500444},
-    year={2025},
-    publisher={Wiley Online Library},
-    doi={10.1002/aisy.202500444}
+@article{wu2025magiclaw,
+  title={MagiClaw: A Dual-Use, Vision-Based Soft Gripper for Bridging the Human Demonstration to Robotic Deployment Gap},
+  author={Wu, Tianyu and Han, Xudong and Sun, Haoran and Zhang, Zishang and Huang, Bangchao and Song, Chaoyang and Wan, Fang},
+  journal={arXiv preprint arXiv:2509.19169},
+  year={2025}
 }
 ```

__init__.py

@@ -0,0 +1 @@
+from .modeling import FingerNet, FingerNetConfig

config.json

@@ -1,17 +1,15 @@
 {
-  "model_name": "FingerNet",
-  "architecture": "dual-branch-mlp",
-  "framework": "onnx",
-  "input_motion": 6,  
-  "output_force_dim": 6,
-  "output_shape_dim": 1800,
-  "quantized": false,
-  "version": "v1.0",
-  "description": "Dual-branch MLP model for force (6D) and shape (1800D) prediction from motion of the asFinger.",
-  "author": "AncoraSpring",
-  "license": "BSD-3-Clause",
-  "training_dataset": "asFinger Dataset",
-  "dependencies": {
-    "onnxruntime": ">=1.16.0"
+  "_name_or_path": "asRobotics/fingernet",
+  "architectures": ["FingerNet"],
+  "model_type": "fingernet",
+  "x_dim": [6],
+  "y_dim": [6, 1800],
+  "h1_dim": [100, 1000],
+  "h2_dim": [100, 1000],
+  "torch_dtype": "float32",
+  "layer_norm": false,
+  "use_activation": "relu",
+  "auto_map": {
+    "AutoModel": "modeling.FingerNet"
   }
 }

config_surf.json

@@ -1,17 +0,0 @@
-{
-  "model_name": "FingerNet(Surface)",
-  "architecture": "dual-branch-mlp",
-  "framework": "onnx",
-  "input_motion": 6,  
-  "output_force_dim": 6,
-  "output_shape_dim": 1803,
-  "quantized": false,
-  "version": "v1.0",
-  "description": "Dual-branch MLP model for force (6D) and shape (1803D) prediction from motion of the asFinger with surface.",
-  "author": "AncoraSpring",
-  "license": "BSD-3-Clause",
-  "training_dataset": "asFinger Dataset",
-  "dependencies": {
-    "onnxruntime": ">=1.16.0"
-  }
-}

fingernet_surf.onnx → model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0f32a0bd34080898076b643e4be9bbf12b4f6b22f3e6736b5721911954656653
-size 11299042
+oid sha256:3209d7f93ef4059f4cf9108139e7a4f26cce35ff8b0cdc654f7a2abea7617a53
+size 11285832

modeling.py

@@ -0,0 +1,52 @@
+import torch
+import torch.nn as nn
+from transformers import PreTrainedModel, PretrainedConfig
+
+class FingerNetConfig(PretrainedConfig):
+    model_type = "fingernet"
+
+    def __init__(
+        self,
+        x_dim=[6],
+        y_dim=[6, 1800],
+        h1_dim=[100, 1000],
+        h2_dim=[100, 1000],
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+        self.x_dim = x_dim
+        self.y_dim = y_dim
+        self.h1_dim = h1_dim
+        self.h2_dim = h2_dim
+
+
+class FingerNet(PreTrainedModel):
+    config_class = FingerNetConfig
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.x_dim = config.x_dim
+        self.y_dim = config.y_dim
+        self.h1_dim = config.h1_dim
+        self.h2_dim = config.h2_dim
+
+        # build sub-networks
+        self.branches = nn.ModuleList()
+        for i in range(len(self.y_dim)):
+            net = nn.Sequential(
+                nn.Linear(self.x_dim[0], self.h1_dim[i]),
+                nn.ReLU(),
+                nn.Linear(self.h1_dim[i], self.h2_dim[i]),
+                nn.ReLU(),
+                nn.Linear(self.h2_dim[i], self.y_dim[i]),
+            )
+            self.branches.append(net)
+
+        # initialize weights
+        self.post_init()
+
+    def forward(self, x):
+        if isinstance(x, (list, tuple)):
+            x = torch.tensor(x, dtype=torch.float32)
+        outputs = [branch(x) for branch in self.branches]
+        return tuple(outputs)