How between-graph replication with asynchronous updates in Tensorflow runs? - asynchronous

We focus on this situation: between-graph replication with asynchronous updates.
The following figure shows how it works. And I known that each worker will send gradients to every ps and receive its updated parameters from every ps.(Is it correct? Is it path 1 and path 2?)
But can anyone explain in detail what does the figure 1,2,3,4,5,6 mean respectively?
Thank you in advance!

Related

What exactly is timekey_wait parameter in Fluentd?

I'm new in EFK. I have a problem with showing logs in Kibana. I already resolved, but I'm not sure of my approach.
Problem: Kibana shows log after 10 minutes when Elasticsearch is restarted.
I studied in https://docs.fluentd.org/configuration/buffer-section
And I found out that if I configure timekey_wait parameter in buffer section is 0s, Kibana shows logs without delay.
The problem is resolved, but I have still concerned about timekey_wait parameter.
Are there others impacted by the change?
Why is timekey_wait needed? Please give me an example of the necessity of it.
Thank you for your time!
1 & 2 ) According to the documentation by using timekey_wait parameter fluentd waits the specified amount of time, before writing chunks. By this way delayed log lines that needs to be in the same log chunk won't be missed.
If your timekey is 60m and timekey_wait is 10m, now the chunks will be written after 70m not 60m.
If you don't have delayed log lines to be come it becomes less important.In one of my implementations, I use flush_interval parameter. That way timekey is not needed. (buffer chunks will be flushed after this time)

How to clear message error that occur from the Corda flow test

I try to run the test that I created from Corda 3.3 in Corda 4.1
I have 2 test case for test the flow
in the first test I expected fail that came from contract
and the result from first test is also correct as I expected to
but I error that I got from first test was send to hospital flow
and the error have been shown in the second test
actually the error that come for the first test not effect to the second test but it make the second test to slow
I really don't know how to clear the the error message before sun the second test
If someone have any idea please let me know thank you.
Note: If you have the way that not stop the nodes and re-create mock node again before run new test, it will be the solution that I looking for.
==============================
I have 6 tests in one file
first I try to create the network and use that net work for all of 6 test in this way I can reduce the time for initiate the network
but I need to clear the database after each test finish for avoid create duplicate data.
everything work until I change to Corda 4.1
in the 4.1 I don't know why the way that I use for clear database in Corda 3.3 not work like before (In 4.1 take long times for tuncate the table)
so I need to change the way to create the network and stop after finish each test.
In this way take more time for initiate the network (around 20-30 seconds per test)
and the point that surprised me is when I finish 5 tests in the 6th test take the long time (the log show house keeper clean) they use 6 minutes for finished
but when I run only that test, they use 1 minute for finished.
my question are
1. How I clear everything after finish each test
2. Have another way for initiate the network and use for every test? and how to clear the database and message after finish each test
It is not visible actual cause of exception.
But be aware that for 4.x corda you have to put
subFlow(ReceiveFinalityFlow(otherPartySession))
As last operation.
Dunno if this helps
It sounds like you are sharing state between tests, which is generally bad. Consider creating a MockNetwork in JUnit's #Before method or use the DriverDSL to create an isolated test for each test case.

Custom code on the updateHandler from HKWorkoutSession

My question is plain and simple. Can i run custom code in the updateHandler when I´ve executed a HKWorkoutSession and is listening for HeartRate samples? (Even when the Watch is locked from "wrist down" movement)
If this i possible what are my limitations?
I´m interested in processing the HeartRate data when my code receives them. I don´t have a device yet so I haven't been able to test it yet.
Would love your thoughts on this if anyone have experimented with an actual device.
Yes you can do this, I've had it append every HKSample that came back from my query onto an array, so when I resume the array is much larger. However UI won't update this way, on resume you need to update to the values you've received from the updateHandler.
Whether I should be doing this, or how far it can be pushed, I'm not sure.
Update
In the latest Xcode 7 beta you can get simulated workout data, so you won't need to install the Watch OS 2 beta on your device.

steady state initialization in Modelica

For example, I have a multibody vehicle model with an initial height of, say 0.1 meter (all wheel vertical loads = 0), as the sim runs, the vehicle will drop onto the ground, after 10 seconds, it reaches its steady state.
I wonder if it is possible to initialize the model exactly at the steady state? I read something about the homotopy command, but I was not even sure if it is something that I was looking for due to lack of examples, so I am not able to implement it to my model. I wonder if there are any other solutions to this kind of initialization problem?
Thanks in advance!
Thanks for Matths comments.
The web page matth has provided is very helpful, and if anyone wants to start your simulation from steady state, you should take a look.
I found some details on simulation continuation and more commands from User Manual 1, "Simulator API" section.
Here's one more additional question based on this one,
Is there an equivalent C function in the Dymola/source folder of ImportInitial(), Or ImportInitialResult()? Thanks.

IMediaControl::Run followed by IMediaControl::Stop followed by IMeidaControl::Run doesn't switch on certain Onboard cameras

I have a DirectShow webcam application. I make use of Sample Grabber to get the buffer callbacks and IVideoWindow to control the display co-ordinates for the Preview. I have Preview and Capture Streams which I run as below.
g_pBuild->RenderStream(&PIN_CATEGORY_CAPTURE, &MEDIATYPE_Video,cam,g_pGrabberF,pNullRenderer2); g_pBuild->RenderStream(&PIN_CATEGORY_PREVIEW, &MEDIATYPE_Video,cam,NULL,NULL);
On certain On board cameras, IMediaControl::Run followed by IMediaControl::Stop followed by IMediaCOntrol::Run doesn't switch on the camera.
Extenal USB cameras work properly here. How can I diagnose more on this? Any pointers, please help.
Maybe its specific to a certain hardware issue in the unit.
Do a quick test by adding sleep of 1 sec between calls.
If it does help than you need to find a way to know when to unit state in idle or not.
There are two important parts of the question which you did not provide:
Filter graph topologies
HRESULTs of the method calls
A problem you might be having is that one of the filters in the topology does not handle well state transitions and fails somewhere between states. Supposedly your second Run meets it still trying to complete Stop. You might get a HRESULT there which indicates the issue (better for you) or the filter fails silently.
The filter graph's is the unlikely source of the bug itself. Chances are high that it does everything flawlessly, however since internally it distributes the calls between filters, one of the filter is letting you down.

Resources