Millisecond Forums

Experiment freezing halfway through final block

https://forums.millisecond.com/Topic27380.aspx

By Adam.Rodriguez3 - 7/3/2019

Hi Inquisit,

I am running an experiment requiring participants to respond to a succession of videos. There are six blocks in total. For some reason, the experiment lags heavily or freezes roughly one third of the way through the final block each time I run it.

This appears to be the case when I run the script from Inquisit, when I run the monkey, when I run the experiment from the weblink (I'm using a Mac, IOS Mojave 10.4), and when I run the experiment from another computer via the weblink (OSX Yosemite 10.10.5). I am running the most current version of Inquisit.

When I run each block individually via Inquisit, there are no problems. It seems to happen only when the full experiment is run. The errors do not appear to be occurring at consistent events. Blocks are presented randomly, and so the problematic sixth block is different every time. Further, the trial number wherein the freeze occurs is different each time.

Could this be a problem with cumulative information build-up? Does freezing typically occur with experiments with a high number of videos?

I have attached the link so you can access the script:

https://mili2nd.co/hvnb

Many thanks,

Adam
By Dave - 7/3/2019

Adam.Rodriguez3 - 7/4/2019
Hi Inquisit,

I am running an experiment requiring participants to respond to a succession of videos. There are six blocks in total. For some reason, the experiment lags heavily or freezes roughly one third of the way through the final block each time I run it.

This appears to be the case when I run the script from Inquisit, when I run the monkey, when I run the experiment from the weblink (I'm using a Mac, IOS Mojave 10.4), and when I run the experiment from another computer via the weblink (OSX Yosemite 10.10.5). I am running the most current version of Inquisit.

When I run each block individually via Inquisit, there are no problems. It seems to happen only when the full experiment is run. The errors do not appear to be occurring at consistent events. Blocks are presented randomly, and so the problematic sixth block is different every time. Further, the trial number wherein the freeze occurs is different each time.

Could this be a problem with cumulative information build-up? Does freezing typically occur with experiments with a high number of videos?

I have attached the link so you can access the script:

https://mili2nd.co/hvnb

Many thanks,

Adam

> Could this be a problem with cumulative information build-up?

Yes. What you describe very much sounds like memory build-up in the Mac's media subsystem until the system starts to run out of resources. Things you can try:
- Different video format, e.g. MP4. MOV isn't necessarily the most efficient format these days.
- Cutting down the number of trials. It seems there are currently 7 repetitions of each video in a given block. Perhaps 5 or 6 repetitions would be sufficient, too? If the data files in your web account are a vaild indication of when the lagging starts, it most commonly seems to occur in the 5th block (not the 6th), after a total of around 190-200 trials have been completed.
- You can also try introducing a /posttrialpause to give Inquisit some specified time after a trial to free up memory. I'm not convinced this will help in this particular case because I suspect the memory build-up to occur in an OS component, not Inquisit directly, but it's worth a try nonetheless.
- Lastly, you could split the script into several separate ones (e.g. one per block) and then run those in random order via <batch>. That might prevent the build-up from reaching critical mass.

<batch 1>
/ file = "practice.iqx"
</batch>

<batch 2>
/ file = "mlothi.iqx"
/ file = "mhithi.iqx"
/ file = "mhitlo.iqx"
/ file = "mlotlo.iqx"
/ selectionmode = random
</batch>

By Adam.Rodriguez3 - 7/7/2019

Dave - 7/4/2019
Adam.Rodriguez3 - 7/4/2019
Hi Inquisit,

I am running an experiment requiring participants to respond to a succession of videos. There are six blocks in total. For some reason, the experiment lags heavily or freezes roughly one third of the way through the final block each time I run it.

This appears to be the case when I run the script from Inquisit, when I run the monkey, when I run the experiment from the weblink (I'm using a Mac, IOS Mojave 10.4), and when I run the experiment from another computer via the weblink (OSX Yosemite 10.10.5). I am running the most current version of Inquisit.

When I run each block individually via Inquisit, there are no problems. It seems to happen only when the full experiment is run. The errors do not appear to be occurring at consistent events. Blocks are presented randomly, and so the problematic sixth block is different every time. Further, the trial number wherein the freeze occurs is different each time.

Could this be a problem with cumulative information build-up? Does freezing typically occur with experiments with a high number of videos?

I have attached the link so you can access the script:

https://mili2nd.co/hvnb

Many thanks,

Adam

> Could this be a problem with cumulative information build-up?

Yes. What you describe very much sounds like memory build-up in the Mac's media subsystem until the system starts to run out of resources. Things you can try:
- Different video format, e.g. MP4. MOV isn't necessarily the most efficient format these days.
- Cutting down the number of trials. It seems there are currently 7 repetitions of each video in a given block. Perhaps 5 or 6 repetitions would be sufficient, too? If the data files in your web account are a vaild indication of when the lagging starts, it most commonly seems to occur in the 5th block (not the 6th), after a total of around 190-200 trials have been completed.
- You can also try introducing a /posttrialpause to give Inquisit some specified time after a trial to free up memory. I'm not convinced this will help in this particular case because I suspect the memory build-up to occur in an OS component, not Inquisit directly, but it's worth a try nonetheless.
- Lastly, you could split the script into several separate ones (e.g. one per block) and then run those in random order via <batch>. That might prevent the build-up from reaching critical mass.

<batch 1>
/ file = "practice.iqx"
</batch>

<batch 2>
/ file = "mlothi.iqx"
/ file = "mhithi.iqx"
/ file = "mhitlo.iqx"
/ file = "mlotlo.iqx"
/ selectionmode = random
</batch>


Thanks for this prompt and helpful response Dave. I will try what you have suggested.

Adam