encounter issues when I run other yml files

[henryliuliuliu]

Specifically, when I runpython fluidlab/run.py --cfg_file configs/exp_gathering_easy.yaml --record it reports: aceback (most recent call last): File "fluidlab/run.py", line 63, in <module> main() File "fluidlab/run.py", line 41, in main env = gym.make(cfg.EXP.env_name, seed=cfg.EXP.seed, loss=False, loss_type='diff', renderer_type=args.renderer_type) File "C:\Users\Shipe\miniconda3\envs\fluidlab\lib\site-packages\gym\envs\registration.py", line 200, in make return registry.make(id, **kwargs) File "C:\Users\Shipe\miniconda3\envs\fluidlab\lib\site-packages\gym\envs\registration.py", line 105, in make env = spec.make(**kwargs) File "C:\Users\Shipe\miniconda3\envs\fluidlab\lib\site-packages\gym\envs\registration.py", line 75, in make env = cls(**_kwargs) File "c:\users\shipe\dropbox\myworkspace\fluidlab\fluidlab\envs\gatheringeasy_env.py", line 35, in __init__ self.build_env() File "c:\users\shipe\dropbox\myworkspace\fluidlab\fluidlab\envs\fluid_env.py", line 38, in build_env self.setup_bodies() File "c:\users\shipe\dropbox\myworkspace\fluidlab\fluidlab\envs\gatheringeasy_env.py", line 69, in setup_bodies material=RIGID, File "c:\users\shipe\dropbox\myworkspace\fluidlab\fluidlab\fluidengine\taichi_env.py", line 93, in add_body self.particle_bodies.add_body(**kwargs) File "c:\users\shipe\dropbox\myworkspace\fluidlab\fluidlab\fluidengine\bodies\bodies.py", line 40, in add_body self.add_mesh(filling=filling, **kwargs) File "c:\users\shipe\dropbox\myworkspace\fluidlab\fluidlab\fluidengine\bodies\bodies.py", line 199, in add_mesh voxels = pkl.load(open(voxelized_file_path, 'rb')) AttributeError: 'TrackedArray' object has no attribute 'fast_hash'. When I run: python fluidlab/run.py --cfg_file configs/exp_mixing.yaml --record Traceback (most recent call last): File "fluidlab/run.py", line 63, in <module> main() File "fluidlab/run.py", line 44, in main record_target(env, path=args.path, user_input=args.user_input) File "c:\users\shipe\dropbox\myworkspace\fluidlab\fluidlab\optimizer\recorder.py", line 107, in record_target recorder.record(user_input) File "c:\users\shipe\dropbox\myworkspace\fluidlab\fluidlab\optimizer\recorder.py", line 16, in record policy = self.env.demo_policy(user_input) TypeError: demo_policy() takes 1 positional argument but 2 were given. There are some other issues with renderred_type, but I assume that following the latter_art.env, we could solve this~ Is there a plan to have documentation about how to generate the paper results? Best, Shipeng

[henryliuliuliu]

I am wondering if the algorithm used to compare with the DP is here. If I want to get the result of SAC, etc, what should I do to minimize the efforts? Thanks very much!

[henryliuliuliu]

AttributeError: 'TrackedArray' object has no attribute 'fast_hash' this error can be solved by deleting the file and regenerate

[zhouxian]

I tried exp_gathering_easy.yaml and it seems ok on my end. Maybe it's a package version mismatch issue? Did you create your env using environment.yml? Anyhow, glad you found a solution.

For baselines, we simply used stable baselines (https://github.com/DLR-RM/stable-baselines3), and we didn't include it in our released repo to avoid making it too messy.

[henryliuliuliu]

I am wondering if there is a way to simulate a real rigid robot with joint？ and did you calculate the interaction force between robot and endeffector，in other words， can I do force control？

Best， Shipeng

On Sat, Sep 9, 2023 at 12:14 PM Zhou Xian @.***> wrote:

I tried exp_gathering_easy.yaml and it seems ok on my end. Maybe it's a package version mismatch issue? Did you create your env using environment.yml? Anyhow, glad you found a solution.

For baselines, we simply used stable baselines ( https://github.com/DLR-RM/stable-baselines3 https://urldefense.com/v3/__https://github.com/DLR-RM/stable-baselines3__;!!LIr3w8kk_Xxm!pOkp2eEqPWCywlgzESR--ywABcgoyTR2q6bv2C39ikTQ6HmrcyfPHVTnbKRldGFAu_N-U0w7cmTA53jvtyiFk2LW$), and we didn't include it in our released repo to avoid making it too messy.

— Reply to this email directly, view it on GitHub https://urldefense.com/v3/__https://github.com/zhouxian/FluidLab/issues/7*issuecomment-1712581690__;Iw!!LIr3w8kk_Xxm!pOkp2eEqPWCywlgzESR--ywABcgoyTR2q6bv2C39ikTQ6HmrcyfPHVTnbKRldGFAu_N-U0w7cmTA53jvty5btzF_$, or unsubscribe https://urldefense.com/v3/__https://github.com/notifications/unsubscribe-auth/AK5LFKN4MCJYBNWNOOHYHLDXZS5Y3ANCNFSM6AAAAAA4OBK3ZY__;!!LIr3w8kk_Xxm!pOkp2eEqPWCywlgzESR--ywABcgoyTR2q6bv2C39ikTQ6HmrcyfPHVTnbKRldGFAu_N-U0w7cmTA53jvt7pZai9F$ . You are receiving this because you authored the thread.Message ID: @.***>

[zhouxian]

Interesting you asked - current fluid lab doesn't have serious rigid body dynamics, not to mention articulated body such as robot arm. However, we are indeed working on a revolutionary platform that will support an advanced rigid body physics engine, with also coupling with other materials. In addition, we are adding support for other physics solvers such as PBD, SPH, FEM and more (current fluidlab only has MPM). We also redesigned the whole infrastructure and now the gradient access is 100x easier than fluidlab. It's still under development and we will try to have a alpha release probably next month or so. Let me know if you want to know more. We can have a chat if you want and possibly add you to our private repo.

[henryliuliuliu]

The project you mentioned looks promising. My goal is to add a quadruped robot scenario， and run the robot in various fluid. If you could add me to the repo, that would be great!

Best, Shipeng

On Sat, Sep 9, 2023 at 12:38 PM Zhou Xian @.***> wrote:

Interesting you asked - current fluid lab doesn't have serious rigid body dynamics, not to mention articulated body such as robot arm. However, we are indeed working on a revolutionary platform that will support an advanced rigid body physics engine, with also coupling with other materials. In addition, we are adding support for other physics solvers such as PBD, SPH, FEM and more (current fluidlab only has MPM). We also redesigned the whole infrastructure and now the gradient access is 100x easier than fluidlab. It's still under development and we will try to have a alpha release probably next month or so. Let me know if you want to know more. We can have a chat if you want and possibly add you to our private repo.

— Reply to this email directly, view it on GitHub https://urldefense.com/v3/__https://github.com/zhouxian/FluidLab/issues/7*issuecomment-1712586151__;Iw!!LIr3w8kk_Xxm!sRTgYuSOFB1ExBYyge7vgLLb0Llg6A61-4dpzEUobqi9INbgZEv9LFp8u7W2fjuE5ywnpIr7mC7xwctPPRJrhcHE$, or unsubscribe https://urldefense.com/v3/__https://github.com/notifications/unsubscribe-auth/AK5LFKMGX6OVQAJK4A7I7J3XZTAUXANCNFSM6AAAAAA4OBK3ZY__;!!LIr3w8kk_Xxm!sRTgYuSOFB1ExBYyge7vgLLb0Llg6A61-4dpzEUobqi9INbgZEv9LFp8u7W2fjuE5ywnpIr7mC7xwctPPRaxgRpb$ . You are receiving this because you authored the thread.Message ID: @.***>