Dex1B 0.0.2__py3-none-any.whl

Dex1B/Dex1B.py

@@ -0,0 +1,116 @@
+import torch
+from torch import cdist
+from torch.nn import Module, ModuleList
+
+from x_transformers import Encoder
+
+from x_mlps_pytorch import MLP
+
+import einx
+from einops import rearrange
+
+# helpers
+
+def exists(v):
+    return v is not None
+
+def default(v, d):
+    return v if exists(v) else d
+
+# losses
+
+def simple_sdf_loss(
+    surface_points,     # (b n 3)
+    hand_points,        # (b m 3)
+    hand_point_radius,  # (b m)
+    mask = None         # (b n) | none
+):
+    """
+    add their simple penetration loss in section IV
+    """
+
+    dist = cdist(hand_points, surface_points)
+
+    hand_to_all_surface = einx.subtract('b m, b m n', hand_point_radius, dist).relu() # max(0, radius - dist)
+
+    if exists(mask):
+        mask_value = torch.finfo(hand_to_all_surface.dtype).max
+        hand_to_all_surface = einx.where('b n, b m n,', mask, hand_to_all_surface, mask_value)
+
+    hand_to_closest_dist = hand_to_all_surface.amin(dim = -1)
+
+    return hand_to_closest_dist.sum()
+
+# classes
+
+class PointTransformer(Module):
+    """ https://arxiv.org/abs/2312.10035v1 """
+
+    def __init__(self):
+        super().__init__()
+
+class CVAE(Module):
+    def __init__(
+        self,
+        dim,
+        dim_hiddens = (256, 512, 256), # from Table 6. in paper
+        kl_loss_weight = 1e-4
+    ):
+        super().__init__()
+        assert len(dim_hiddens) > 0
+        dim_latent = default_layer_sizes[-1]
+
+        self.encode = MLP(dim, *default_layer_sizes)
+
+        self.to_mean_log_variance = nn.Linear(dim_latent, dim_latent * 2, bias = False)
+
+        self.decode = MLP(*default_layer_sizes, dim)
+
+        # loss weights
+
+        self.kl_loss_weight = kl_loss_weight
+
+    def forward(
+        self,
+        inp, # (b d)
+        return_loss = False
+    ):
+
+        encoded = self.encode(inp)
+
+        mean, log_variance = self.to_mean_log_variance(encoded).chunk(2, dim = -1)
+
+        std = (0.5 * log_variance).exp()
+
+        noise = torch.randn_like(mean)
+
+        reparamed = mean + std * noise
+
+        recon = self.decode(reparamed)
+
+        if not return_loss:
+            return recon
+
+        mse_loss = F.mse_loss(recon, inp)
+
+        kl_loss = 0.5 * (mean.square() + log_variance.exp() - log_variance - 1.).sum(dim = -1).mean()
+
+        total_loss = (
+            mse_loss +
+            kl_loss * self.kl_loss_weight
+        )
+
+        loss_breakdown = (mse_loss, kl_loss)
+
+        return total_loss, loss_breakdown
+
+class DexSimple(Module):
+    def __init__(
+        self,
+        point_transformer: PointTransformer,
+        cvae: CVAE
+    ):
+        super().__init__()
+
+        self.pointnet = pointnet
+        self.cvae = cvae

dex1b-0.0.2.dist-info/METADATA

@@ -0,0 +1,72 @@
+Metadata-Version: 2.4
+Name: Dex1B
+Version: 0.0.2
+Summary: MMDiT
+Project-URL: Homepage, https://pypi.org/project/Dex1B/
+Project-URL: Repository, https://github.com/lucidrains/Dex1B
+Author-email: Phil Wang <lucidrains@gmail.com>
+License: MIT License
+        
+        Copyright (c) 2025 Phil Wang
+        
+        Permission is hereby granted, free of charge, to any person obtaining a copy
+        of this software and associated documentation files (the "Software"), to deal
+        in the Software without restriction, including without limitation the rights
+        to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+        copies of the Software, and to permit persons to whom the Software is
+        furnished to do so, subject to the following conditions:
+        
+        The above copyright notice and this permission notice shall be included in all
+        copies or substantial portions of the Software.
+        
+        THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+        IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+        FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+        AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+        LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+        OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+        SOFTWARE.
+License-File: LICENSE
+Keywords: artificial intelligence,deep learning,dexterity,scaling,synthetic data
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3.9
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.9
+Requires-Dist: einops>=0.8.0
+Requires-Dist: gotennet-pytorch>=0.3.1
+Requires-Dist: torch>=2.0
+Requires-Dist: x-mlps-pytorch
+Requires-Dist: x-transformers>=2.5.3
+Provides-Extra: examples
+Provides-Extra: test
+Requires-Dist: pytest; extra == 'test'
+Description-Content-Type: text/markdown
+
+<img src="./fig3.png" width="400px"></img>
+
+## Dex1B (wip)
+
+## Citations
+
+```bibtex
+@inproceedings{ye2025dex1b,
+    title   = {Dex1B: Learning with 1B Demonstrations for Dexterous Manipulation},
+    author  = {Ye, Jianglong and Wang, Keyi and Yuan, Chengjing and Yang, Ruihan and Li, Yiquan and Zhu, Jiyue and Qin, Yuzhe and Zou, Xueyan and Wang, Xiaolong},
+    booktitle = {Robotics: Science and Systems (RSS)},
+    year    = {2025}
+}
+```
+
+```bibtex
+@misc{wu2024pointtransformerv3simpler,
+    title   = {Point Transformer V3: Simpler, Faster, Stronger}, 
+    author  = {Xiaoyang Wu and Li Jiang and Peng-Shuai Wang and Zhijian Liu and Xihui Liu and Yu Qiao and Wanli Ouyang and Tong He and Hengshuang Zhao},
+    year    = {2024},
+    eprint  = {2312.10035},
+    archivePrefix = {arXiv},
+    primaryClass = {cs.CV},
+    url     = {https://arxiv.org/abs/2312.10035}, 
+}
+```

dex1b-0.0.2.dist-info/RECORD

@@ -0,0 +1,6 @@
+Dex1B/Dex1B.py,sha256=qpX5BNxLs5LUoWJZRZOW4oK_bxnkJxoYiqPKNOSWunI,2633
+Dex1B/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+dex1b-0.0.2.dist-info/METADATA,sha256=QZ4MIlr2vflOZsHOAd7iLF7eou-NYNALDXF2qCZwR_8,3007
+dex1b-0.0.2.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+dex1b-0.0.2.dist-info/licenses/LICENSE,sha256=1yCiA9b5nhslTavxPjsQAO-wpOnwJR9-l8LTVi7GJuk,1066
+dex1b-0.0.2.dist-info/RECORD,,

dex1b-0.0.2.dist-info/WHEEL

@@ -0,0 +1,4 @@
+Wheel-Version: 1.0
+Generator: hatchling 1.27.0
+Root-Is-Purelib: true
+Tag: py3-none-any

dex1b-0.0.2.dist-info/licenses/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 Phil Wang
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.