I am trying to train a DQN agent in an environment coded in JAX, but the initialization of the trainer fails when it first tries to reset the environment (with a not-valid JAX type error). Before getting into the debugging process, I thought of looking for example projects but I cannot find anything so i am wondering whether interfacing them is not possible.
Currently there is no example in RLlib right out of the box that supports jax envs. I think the main functionalities should be there, it was just never tested with Jax envs to figure out the feature requirements and the missing parts. I encourage you to give it a try and create a feature request for it on ray's issue tracker on github and if you have bandwidth give it a try and make a contribution. You can also let the RL team know about the missing features for enabling RLlib to work on JAX envs.
Related
Hi I am developing a runtime using Substrate-FRAME and I would like to know how can I write a system test for my runtime?
The main purpose of writing a system test is to ensure that the final build is fulfilling all of the required specifications and also to ensure nothing is compromised on a runtime upgrade.
The idea for me is something similar to point no. 2 mentioned in this thread.
Any documentation regarding this type of tests would be greatly helpful.
Update:
I ended up using py-substrate-interface to make test scenarios. Now I can automatically deploy nodes to form a network (thanks to Python) and run my custom system test scenarios. Very useful tool for developing runtimes in Substrate.
There is an overview here on the DevHub
And there are examples throughout substrate that include tests.rs and mock.rs files to use as reference.
If you have not already, checkout the create a pallet tutorial and the recipes all have some tasty examples to look at for these as well.
I hope you are doing good and i really appreciate your help here for my query.
We have our system T3000 written in C++ (http://www.temcocontrols.com/ftp/software/9TstatSoftware.zip and codes are available here https://github.com/temcocontrols/T3000_Building_Automation_System).
I am trying to integrate 'BIRT reporting tool' in my C++ application. I want to create report based on the data available in our T3000 system. I think BIRT is embeddable (??). We don't need to compile and change the project, just need to be able to call it from T3000.exe mainly.
My thinking is we may put one menu label in existing T3000 and try to display report in user single click.
Can you please help me to solve my issue with 'BIRT' ? I really appreciate your answer.
Regards
Raju
Well, the answer depends on what your definition of "embeddable" is.
BIRT is written in pure Java.
I could think of 3 different ways:
Of course it is possible to integrate Java code into an existing C/C++ program (see Embed Java into a C++ application?).
You could just use the BIRT runtime engine and generate the report as PDF or HTML from the command line (that means, basically you call the java executable from your program with several arguments). See Birt - How to run report engine on the console? and http://eclipser-blog.blogspot.de/2008/02/automatic-generation-of-birt-reports.html for more information.
You could run a Java web server like Tomcat in a second process and then start your report by calling a http URL (e.g. you could use the included Servlet example). See http://www.eclipse.org/birt/documentation/integrating/viewer-usage.php
Similar to 3. (see below)
Some notes:
The second option is slow, due to the Java and BIRT engine startup overhead (this may take several seconds). With the first and third option, the startup overhead is or can be minimized to only once (and for each report).
For the second and third option it may be necessary to modify the existing code of the example programs to suit your needs.
The first option is probably the best for an industry-quality solution, but it is also the most difficult to develop.
Anyway, Java skills are necessary IMHO.
If you plan to run this on a SOC instead of a PC, take performance into account.
Is a Java-based solution well-suited for this kind of hardware? BIRT needs quite a lot of RAM and CPU (for a SOC). Hardware like the Raspi 3 should handle this quite easily, I reckon.
I integrated the BIRT runtime into an existing Python application (all this running on an application server) in a fourth way: I wrote a listener program that listens on a TCP socket for BIRT tasks. It uses a pool of worker processes (written in Java) which in turn use the BIRT report engine to generate the output. The client program (here: written in Python) opens a TCP connection to the listener and uses this socket to tell it which report to generate (including report parameters and destination file name). The listener program then in turn chooses a worker process for the task and gives the task to the worker process.
So, basically, this fourth option is similar to the third one, with two differences:
The communication is socket-based (instead of http), allowing bi-di communication.
The architecture is multi-processes instead of multi-threading. We choose this because very large reports could cause out-of-memory errors for otherwise unrelated reports that just happen to run at the same time. It's the same basic architecture Oracle chose for their reports server.
However, developing the programs took months.
HVB: I have to give you more than a simple thanks for the explanation above, this info will save us time I am sure. Raju will be sharing our experience after we get into the project a little deeper so others can benefit.
I am aware that we can make templates of domains in weblogic very easily using config_builder script. Is there a similar thing in websphere?
I know nothing about WebLogic, but fix pack 9 for WebSphere added something you may find useful. The wsadmin command AdminTask.extractConfigProperties with GenerateTemplates and PortablePropertiesFile options set to true will generate a portable, editable file transferable to another cell. AdminTask.applyConfigProperties is used to read your edited output and apply the properties to a new cell, server, etc. I haven't tried this yet outside of a controlled sandbox environment; so, I'm not sure what pitfalls may await you. But if you have a ton of servers to build, it may be worth your time to do some experimentation.
Here's the IBM doc on the topic.
As far as i know i don't think there is such a capability.
You can use the default product shipped profiles to start with and create the servers and configure them the way you want.
These servers can then be used as a template to build other servers.
I am not sure if this helps you but i thought i would point this out.
Manglu
I am new to GAE (cloud based development in general actually) and I am looking for some advice.
I am looking to use GAE only as a service (REST + JSON) with my client-side in Sproutcore. I am looking to find a light-weight service-based framework that will provide me with strong security (authentication, protection against XSS etc).
Java or Python is fine, the priority being speed and security.
Does anybody have any ideas on this? Any links to relevant information. I am going through an information overload phase. Any help in this regard would be much appreciated!
Thank you.
How about tipfy, a open source lightweight python-based framework made for GAE?
Its core can be enhanced through usage of extensions.
Combining some of those extensions
would allow you to perform auth
against different systems (Google,
Facebook, Twitter...) or against your
own baked one. An example of this can
be found here, source code is
available here.
A really simple sample of JSON
rendering is available here.
Of course, other python frameworks run on top of AppEngine. You'll find in the AppEngine-Python google group, a thread discussing pros/cons of most of them.
Considering speed, those links may shed some light on the subject
Choosing Java vs Python on Google App Engine
Google AppEngine Language Performance Comparison Followup
proceed with caution if you want to use a framework. since GAE is still evolving, frameworks tend to break. i have tried both tipfy and django on large projects.
http://blog.bygsoft.com/2011/04/27/why-we-should-avoid-tipfy-a-gae-framework/
in many places tipfy tends to just write meaningless (read no value add) python wrappers around what the core gae frameworks provide just causing code bloat.
last i checked, the map reduce package did not work well with tipfy. sorry, i may sound very biased but my experience with tipfy has been average.
So looks like I can't comment on your answer but the link was very helpful! Looks very much like something I am looking for:
Server-side sessions
Custom user authentication
JSON based service with no View layer
Fast and secure
Any other python framework that you might possibly know of along the lines of Tipfy? Just so I have something to compare with?
Does anybody know of any Java libraries along the same lines of Tipfy?
Thanks again for the link!
Edit 1:
Hmm...I have no idea how I ended up with 2 accounts. I'll try to iron it out..
Your links were very helpful..Thanks!
I am inching towards Tipfy at this point, but I think I am going to wait a little longer on this thread to see if I get some more opinions..
Thanks again...
Let me first say I am aware of this faq for Mach-II, which discusses using application specific mappings as a third option when:
locating the framework in the server root is not possible and
creating a server wide mapping to the Mach-II framework directory is impossible
Using application specific mappings would also work for other ColdFusion frameworks with similar requirements (ColdSpring). Here is my issue however: my (I should say "their") production servers are all running ColdFusion MX7, and application specific mappings were introduced in ColdFusion 8. I most likely will be unable to do option 1 or 2 because they involve creating server wide changes that could conflict with other applications (I don't have a final word on this but I am preparing for that to be the case).
That said, is there anybody out there who was in similar bind and has done an option 4, in any ColdFusion version, or with any similar framework? The only option 4 I can think of is modifying the entire framework to change this hardcoded path, and even if that worked it would be time consuming and risky. I'm fairly sure that if there was a simple modification or other simple solution it would already be included in the framework (maybe it's included in version 1.8 of Mach-II and I don't know about it yet).
Any thoughts on solving this problem or even unorthodox setups with libraries that have specific path requirements would be appreciated. Any thoughts from Team Mach-II would especially appreciated...we're on the same team here Matt! ;-)
EDIT
Apparently, the ColdBox framework includes a refactor.xml ANT task which includes a target that refactors the ColdBox code to use a different absolute path as a base along with several other useful refactoring targets. So problem solved for ColdBox users.
Looking at the build.xml for Mach-II (1.6 and 1.8) I don't see any target in there that would allow me to refactor the code. I thought about creating a feature request ticket for such a task for Mach-II but frankly I don't think creating such an ANT task is a big priority for the MachII team since the need really only relates to either
a) users of ColdFusion versions below 8
b) someone who wants to use multiple Mach-II versions in the same application, a use I doubt they want to support
The ColdSpring code I have doesn't come with any ANT tasks at all, although I do have unit tests, and I bet if I poked around the SVN I'd find a few build scripts.
Using Ant tasks to refactor and retest the code, or the simpler (and sort of cop out) solution of creating a separate ColdFusion instance for the application are the best answers I've been able to come up with. I don't need this application to exist in the shared scope of other applications, so my first solution is going to be to try and get a dedicated CF instance for this application.
I'm also going to look at the ColdBox refactor.xml ANT task however and see if I can modify it to work generically to recognize and refactor CFC references with modified absolute paths. If I complete this task I'll be sure to post the code somewhere and edit create an answer to link to it. If anybody else wants to take a crack at that or help me out with it feel free.
Until then I'll leave this question open and see if someone comes up with a better solution.
Fusebox is not so strict, I think.
In XML mode (maybe I call this not 100% correcly, just mean using the Application.cfm) it's just proper include in index.cfm, something like:
<cfinclude template="fusebox5/fusebox5.cfm" />
In non-XML mode it will need proper extending in the root Application.cfc:
<cfcomponent extends="path.to.fusebox5.Application" output="false">
All you need is to know the path.
Perhaps you could create a symbolic link and let the operating system resolve the issue for you?
I've been playing with FW/1 lately, and while it may look like you need to add a mapping and extend org.corfield.framework, you can actually move the framework.cfc file into your web root and just extend="framework". It's dead simple, and gets you straight into a great framework with no mess and very little overhead.
It should be as simple as dropping the 'MachII' folder at the root of your domain (i.e. example.com/MachII). No mappings are required to use Mach-II if you just deploy at the root of the domain of your website.
Also:
Please file a ticket for the ANT task you mentioned in your question. Team Mach-II would love to have this issue logged:
Enter a new ticket on the Mach-II Trac
If you want to tackle an ANT task for us, we can get stuff like this incorporated into the builds faster than waiting to for a Team member to work on the ticket. Code submissions from the community are welcome and appreciated.
We don't keep an eye on Stack Overflow very often so we invite you to join our official community group at called "Mach-II for ColdFusion" at Google Groups. The Google Group is the best place to ask questions or comments like this if you want feedback from the Team.