Comments (10)
Thanks for the perspective on the docker issue, I understand now. At the moment our bandwidth is a bit strained between me and @fbxiang so not sure when we can try and address this. It seems like it should be possible but i'm not an expert on this side of the benchmark.
Definitely we will strive to avoid having to use sudo rights to install dependencies. I'll keep the issue open for now.
from maniskill.
Actually, in the next sapien / maniskill, we should be able to run the environment code without vulkan, i will need to add code to automatically check if vulkan is there, and if it is not i will disable all code adding cameras. I'll update here when we have that done
from maniskill.
Hi there! I'm having the same issue as you @abliao , how did you solve the error?
from maniskill.
@gabrieletiboni @abliao take a look at https://haosulab.github.io/ManiSkill2/getting_started/installation.html#troubleshooting. Let me know if those instructions resolve the issue!
from maniskill.
Thank you for your quick reply @StoneT2000 .
I'm currently trying to have ManiSkill2 work on remote cluster nodes, where libvulkan is not present. Also, I'd prefer not to use docker at the moment. As no rendering is needed on the cluster, I was hoping there would be a way to bypass the vulkan dependency altogether.
Clearly, apt installing libvulkan is not an option when sudo rights are not given.
Please let me know if you happen to know any way how to do this :)
from maniskill.
@fbxiang might know.
But I recommend first trying to see if you can vulkan installed in any way possible. Docker is definitely a very good solution (we ourselves use docker for all our remote cluster nodes to run experiments) and if you can use it definitely do. Is there a reason why docker is not okay?
from maniskill.
To resolve this issue, you just need to install libvulkan1. I also recommend using Docker.
from maniskill.
Thank you all for your replies!
Here's my hot takes:
- Docker does introduce a bit of complexity in the process. I'm planning to add support to ManiSkill2 envs to a year-old active project, hence it would require changing the deployment / experimenting pipeline for the whole project.
- Besides the extra work in the first point, one non-trivial point that is often overlooked is that not many people have great familiarity with Docker to begin with. You can certainly make the point that it's not the "package's fault", and more people should simply learn how to containerize their projects. Nevertheless, my personal experience in multiple research groups tells me that ~50% of PhD students / Post Doc habitually make use of Docker containers. I'm part of the less familiarized half, so I might be biased.
- Docker isn't necessarily compatible with all compute clusters. I've worked with multiple clusters there were internally maintained at the research group level, and did not provide great compatibility with Docker containers. Bigger clusters may also not provide compatibility at all due to security reasons, but at least I heard
singularity
can help in those cases (again, extra work and knowledge = less accessibility). - Any app is containerizable, does this mean every app should be containerized? I don't think so, mainly for the three points above. Clearly, some packages must be containerized as no workaround exists, e.g., requirements of specific system dependencies on remote clusters. However, when this is possible, I believe making the effort to make it simply pip installable without sudo rights makes it that much more accessible to the community.
For example, do ManiSkill2 envs absolutely require libvulkan when no rendering is needed? If not, a "no-rendering" option would be great at runtime, which would disable all rendering functionalities and only make use of the CPU-based underlying physics engine.
I guess this would only work for state-based observations (which is nevertheless my use case), as rendering is likely needed to compute point-clouds and depth maps anyways.
That being said, I'll probably end up going for the Docker installation regardless. Just hot takes on the "is there a reason why docker is not okay?" question :)
from maniskill.
Hi, my remote cluster does not support docker command, how can i solve this problem? can i just download the libvulkan.so.1 file on the internet and link it to the environment variables?
from maniskill.
from maniskill.
Related Issues (20)
- Command in the Document not Executable HOT 5
- Upgrade fast-kinematics dependency version
- ModuleNotFoundError: No module named 'sapien.physx' HOT 2
- Documentation and code are inconsistent HOT 1
- Image resolution HOT 14
- Fix bug with get_picture_cuda function for RenderCamera struct
- Fix bug with dependencies not supported on windows systems and document the system/gpu support of ManiSkill
- Remove dependency on CV2 in favor of Imageio/plt
- Only 2 tasks available when downloading demonstrations HOT 3
- Question: Can I swap simulation robot by a real robot? HOT 1
- Expose Joint Force/Torque values
- Fix replay tool for converting pd_joint_pos to pd_ee_delta_pos/pose control HOT 3
- Add simple behavior cloning baseline
- Add SAC + Demos baseline
- Question about benchmarks HOT 4
- Improving env.step() latency/fps? HOT 7
- [Question/Bug] Missing file (and folder) `mani_skill2_ycb/info_pick_v0.json' HOT 2
- Auto print error / check assets are available for tasks
- Issue with running ppo HOT 1
- Multi-gpu simulation support
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from maniskill.