search

xukechun / Efficient_goal-oriented_push-grasping_synergy

[RAL & IROS 2021] Efficient learning of goal-oriented push-grasping synergy in clutter
60 stars 10 forks source link

what the torch cuda and cudnn version are ? #5

Closed littlefiveRobot closed 2 years ago

littlefiveRobot commented 2 years ago

Dear Xu,

I want to know exactly what your torch ,cuda and cudnn version are. (in your readme file, you say cuda10.1 but in the other issues,you say cuda10.2 )Because I encountered some questions as follows.And I hope you can share your code of real-robot, becase I want to run it in my real UR5 robot.Thank you!

Hoping to hear from you! Xia Xiaowu

littlefiveRobot commented 2 years ago

(pytorch1.5) summer@xxw-OMEN:~/efficient_goal-oriented_push-grasping_synergy$ python main.py --stage grasp_only --num_obj 5 --grasp_goal_conditioned --goal_conditioned --goal_obj_idx 4 --experience_replay --explore_rate_decay --save_visualizations Connected to simulation. CUDA detected. Running with GPU acceleration. /home/summer/efficient_goal-oriented_push-grasping_synergy/models.py:202: UserWarning: nn.init.kaiming_normal is now deprecated in favor of nn.init.kaimingnormal. nn.init.kaiming_normal(m[1].weight.data) /home/summer/anaconda3/envs/pytorch1.5/lib/python3.7/site-packages/torch/nn/_reduction.py:43: UserWarning: size_average and reduce args will be deprecated, please use reduction='none' instead. warnings.warn(warning.format(ret)) False False False False False False Creating data logging session: /home/summer/efficient_goal-oriented_push-grasping_synergy/logs/2022-02-15.09:57:47

Training iteration: 0 Traceback (most recent call last): File "main.py", line 798, in main(args) File "main.py", line 577, in main push_predictions, grasp_predictions, state_feat = trainer.goal_forward(color_heightmap, valid_depth_heightmap, goal_mask_heightmap, is_volatile=True) File "/home/summer/efficient_goal-oriented_push-grasping_synergy/trainer.py", line 333, in goal_forward output_prob, state_feat= self.model.forward(input_color_data, input_depth_data, input_goal_mask_data, is_volatile, specific_rotation) File "/home/summer/efficient_goal-oriented_push-grasping_synergy/models.py", line 229, in forward flow_grid_before = F.affine_grid(Variable(affine_mat_before, requires_grad=False).cuda(), input_color_data.size(),True) File "/home/summer/anaconda3/envs/pytorch1.5/lib/python3.7/site-packages/torch/nn/functional.py", line 3324, in affine_grid return torch.affine_grid_generator(theta, size, align_corners) RuntimeError: cublas runtime error : the GPU program failed to execute at /pytorch/aten/src/THC/THCBlas.cu:359

xukechun commented 2 years ago

Hello Xiaowu:

  1. We run our repo mainly with torch 1.2.0, CUDA 10.1 and CUDNN 7.6.5. Also we have reproduced it with torch 1.2.0 CUDA 10.2 in that issue.
  2. Our lab has designed a set of platforms according to our firmware to realize the physical experiments. And the codes of real-world experiments are all based on those platforms. It might be unrealistic to release all the codes. We are so sorry for that><
  3. To realize the real-world experiment, you need to realize functions for grasp and push actions of real robot (which take the predicted discrete position and orientation as input, and output the position of end effector and the orientation of gripper), and get RGB-D images from your camera. Note: you might should cut the image to a proper size for stable mask extraction.
littlefiveRobot commented 2 years ago

Thank you for your careful answer!