pku-epic / unidexgrasp Goto Github PK

Hi, thanks for your contribution.

I have test the code of dexgrasp_generation and find due to lack of data the whole training part is unexecutable.
Will you update the data/preparation in the future?

代码是不是还没放全呀，小小催更꒰ᐢ⸝⸝•༝•⸝⸝ᐢ꒱

Hi,
I'm confused about some codes in "dexgrasp_generation/datasets/dex_dataset.py" and I hope to get some explanations.
First, in line 125, this appears to be a 1x4 vector. What does it correspond to?
plane = recorded_data["plane"]
Second, in line 152-157, pcs_table and pose_matrices appear to be 100x400x3 and 100x4x4 vectors. What 100 means?
obj_pc_path = pjoin(self.root_path, "DFCData", "mesh", category, instance_no, "pcs_table.npy") pose_path = pjoin(self.root_path, "DFCData", "mesh", category, instance_no, "poses.npy") pcs_table = torch.tensor(np.load(obj_pc_path, allow_pickle=True), dtype=torch.float) pose_matrices = torch.tensor(np.load(pose_path, allow_pickle=True), dtype=torch.float)
Third, in line 132, this seems to use the distance between two vectors to retrieve an index from the object's pose, in order to select an object pose for grasping. What is the reason for doing this?
index = (torch.tensor(plane[:3], dtype=torch.float) - pose_matrices[:, 2, :3]).norm(dim=1).argmin()
Finally, there are 100 object poses and about 200 grasp poses in the datasets, I am confused about the code that establishes their correspondence.
Best wishes.

Hi there,

I hope this message finds you well. I have been working on training the goal grasp generation stage but feel uncertain regarding to some details of this process, specifically in relation to joint training ipdf and glow with ContactNet.

Firstly, I would like to confirm whether my understanding is correct. Am I supposed to initially train these three networks independently and then proceed to fine-tune the glow network using the pretrained ipdf and ContactNet? I became a bit confused because in the provided config file, the checkpoint selection for ipdf and ContactNet is set as the initial one.

Furthermore, It would be so helpful if you could tell me the number of epochs you have used for the pretraining and joint training.

Thank you so much for presenting such a great work 👍

Hi, when I try to run the eval code using the checkpoints you provided, there are parameters that have mismach size. I was trying to figure out it by myself but none of my attempts worked... Could you please provide a little help ? Thanks a lot!

Here is the output for ipdf model:

RuntimeError: Error(s) in loading state_dict for IPDFModel:
size mismatch for net.implicit_model.feat_mlp.weight: copying a param with shape torch.Size([256, 64]) from checkpoint,
the shape in current model is torch.Size([256, 128]).

During run dexgrasp_ Generation, an error occurred while training GrassIPDF: FileNotFoundError: [Errno 2] No such file or directory: 'data/DFCData/poses/core/browl-530c2d8f55f9e9e8ce364511e87f52b0';FileNotFoundError: [Errno 2] No such file or directory: 'data/DFCData/meshes/core/bowl-d1addad5931dd337713f2e93cbeac35d However, there are no pose or meshes in the DFCData folder in the downloaded dataset. May I ask if there is an error in the code or if the dataset name is incorrect.

Hi, the work is truly inspiring! I have a question: Is the uploaded example model for the policy net the teacher model? If so, could you kindly share the distilled student model checkpoint for quick inference? Thank you so much.

Hello, thank you very much for open-sourcing this excellent framework and providing a large-scale dataset. While studying, I encountered an error message: ImportError: cannot import name 'get_one_shape' from 'algo.pn_utils.maniskill_learn.utils.data'. I've implemented some functions myself, but I'm unsure about how to implement certain functions. It seems that I might be missing some utility files. In order to ensure that the training process is not affected, would it be possible to provide those missing files? Thank you."

Hello, I have a question about training time. I'm using an RTX 3080 with 10GB of VRAM for training, and it takes 3 to 6 hours to complete one epoch. Is this correct? Thank you for your response.

Hello, I encountered error when installing CSDF: identifier "CHECK_EQ" is undefined
Error. I used pip install -e . and python setup.py install to install but still failed. Can you help me solve this problem? Or tell me the environment required for compilation. Thank you

Hi, I have tried to set the num_envs and nminibatches both to 1 and this error still happens. How large is your GPU memory? I am using 4GB currently. Or is this caused by another problem?

Thank you for your excellent work! Can you provide the code for the Dataset Generation?

Cannot execute the script file in policy
from algo.pn_utils.maniskill_learn.utils.data import is_tuple_of
ModuleNotFoundError: No module named 'algo.pn_utils.maniskill_learn.utils.data'

I can't download the data linked to https://mirrors.pku.edu.cn/dl-release/UniDexGrasp_CVPR2023/DFCData/ in grasp generation

Thanks for this amazing repo! I have a small question about the code related to the paper. For the generate_queries function in dexgrasp_generation/network/models/graspipdf/ipdf_network.py, there are two modes "random" and "grid". I think in the paper, GraspIPDF mentions generating rotation by "grid" mode but the code is using "random" mode. Is that because the "random" mode has better performance?

Also for the "random" mode, the rotation matrices are directly generated by pytorch3d.transforms.random_rotations, will these rotations be the uniform distribution over SO(3)?

Thanks a lot!

When I attempted training, I encountered an error similar to the one in the title. Below is the error message and the directory structure of the project. I suspect there might be an issue with data loading. I appreciate your valuable time; please assist me. Thank you very much.

你好，我运行你们代码dexgrasp_generation部分eval.py的时候出现了std：：bad allocate和RuntimeError: falseINTERNAL ASSERT FAILED at "..\aten\src\ATen\MapAllocator.cpp":135, please report a bug to PyTorch.cannot allocate memory（12）等问题，总的来说好像都是内存不足问题。不知道你们运行时是否出现过类似问题，是存在代码未释放内存的现象还是我们服务器的内存不足呢？麻烦您告知一下是否有相关内容。
图例如下

Hi, thank you for sharing such a great piece of work!
May I ask where I can download the pre-trained checkpoints?

PointCloud in observation is:

Above code gives obs as [.....| point_clouds_xyz(1024*6)| mask_hand(1024)| mask_object(1024)|...], for example,
[....|(x1,y1,z1, r1,g1,b1),(x2,y2,z2,r2,g2,b2)....|(hm1,hm2,....)|(om1,om2....)|.....]
But when reshaping point cloud:

Yet in student policy:

This is only right if pointcloud is dim-8, which requires the obs to be [.....| point_cloud_xyz_hm_om(1024*8)|...]
in which
point_cloud_xyz_hm_om is torch.cat([pointcloud_xyz, hand_mask.float().unsqueeze(-1), object_mask.float().unsqueeze(-1)], dim=-1), for example [....|(x1,y1,z1,r1,g1,b1,hm1,om1),(x2,y2,z2,r2,b2,g2,hm2,om2)|....]

Hi, I have two questions, could you help me to understand them?

1. what's the difference between assets/meshdatav3_pc_feat.zip and assets/meshdatav3_pc_fps.zip.
2. why there is no "apple" or "banana" in the cfg/train_set.yaml, cfg/test_set_seen_cat.yaml and cfg/test_set_unseen_cat.yaml?