-
I was trying to degrain / dejitter an old BW movie (Emile), and had a confusing set of issues:
a) Difficult to find the complete name of the movie or file in the UI, hard to manage a half dozen movies if I wanted to. My file names are 40 to 120 characters long. Ok, I finally found … on the left. Maybe a file name mode vs an icon mode? Found mouseover to get whole name. I didn’t find that for the Output window though. Maybe do the same there?
b) Preview started processing over when I didn’t intend it to, sometimes confusing about which buttons to press in the lower left half of the screen.
c) Estimated 10 hours for 2 passes (45m+9h), ok, not optimal, switched to single pass, now 45mins estimated. Noted that the 10h run didn’t complete overnight (>12h) on the status bar. Possibly having to do with running / starting / stopping other GPU tasks in the background? I didn’t really need that much bob fixing, so waiting for 1-pass now. Maybe unloading and reencoding the data from the GPU is too intensive? Hard to tell. I tried to stop and start the FoldingAtHome GPU usage while single pass was operating, no issues.
d) I see the commandline has a number of color tweak options, which may not be correct for mono BW movie. Not sure which ones to remove, if I was going to run without GUI.
e) Some mouseover popups would be nice to explain the options, or link to a web page.
f) If I click on Open Folder for a processing video, the only smaller file sitting there is from yesterday, despite having stopped and closed previous tasks. Which path/drive is cwd, and how is that controlled? I would prefer bouncing across my M.2 modules, unless the intermediate is tiny.
g) ffmpeg is using the second slower GPU, not sure why. Less memory, lower chip type.
h) I noticed that clicking into the Export options causes the timer to show some other state (not 15 mins of pass 1/1), see screenshot “weird timer state”. Estimate was 45 mins, ffmpeg showing 2h wall time. Update: 6 wall hours later, it seems that it’s doing 2 passes despite asking for 1 and cancelling the 2-pass job. GPU and CPU >65%.
i) Ran for about 8 hours, crashed with error message asking to add logs, but the tar only seems to contain json files and not the 230MB log file. Adding it here, and starting over from a closed app.
logsForSupport.tar.gz (67.6 KB)
2023-07-25-17-27-30-Main.tzlog - Google Drive 7z of log. Lots of errors, the last one looks like a math error, possibly generated by FFMPEG to itself? -
32GB 3700X, 1x2080 TI, 1x2080, 1TB M.2, 512GB M.2, 3x10TB.
-
logsForSupport.tar.gz (67.6 KB)
-
Any screenshots:
It is best to break up the issues in different posts instead of lumping issues together. This makes it easier for other users to comment if they are having the same issue or to offer advice/solutions.
- We will be adding the input filename to the export queue in a coming update which will allow users to hover over the export/preview to see the full file name.
b) I am not clear on what you are describing here.
c) The processing speed is almost entirely dependent on the models being used. I am unsure why the longer render did not complete, however, it sounds like your machine may have went to sleep.
d) It would be best if you would open a support ticket for this.
e) We have in-app tooltips for the majority of models and settings and more information can be found on our docs page. https://docs.topazlabs.com/
f) I am also unsure what you are describing here.
g) What GPU do you have selected for the main AI Processor?
h) Could you capture a screen recording of this issue? It is difficult to see what is happening based on this screenshot.
i)Yes, I see the processing failure and am awaiting feedback from our team.
a) Difficult to find the complete name of the movie or file in the UI, hard to manage a half dozen movies if I wanted to. My file names are 40 to 120 characters long. Ok, I finally found … on the left. Maybe a file name mode vs an icon mode? Found mouseover to get whole name. I didn’t find that for the Output window though. Maybe do the same there?
b) Preview started processing over when I didn’t intend it to, sometimes confusing about which buttons to press in the lower left half of the screen.
c) Estimated 10 hours for 2 passes (45m+9h), ok, not optimal, switched to single pass, now 45mins estimated. Noted that the 10h run didn’t complete overnight (>12h) on the status bar. Possibly having to do with running / starting / stopping other GPU tasks in the background? I didn’t really need that much bob fixing, so waiting for 1-pass now. Maybe unloading and reencoding the data from the GPU is too intensive? Hard to tell. I tried to stop and start the FoldingAtHome GPU usage while single pass was operating, no issues.
d) I see the commandline has a number of color tweak options, which may not be correct for mono BW movie. Not sure which ones to remove, if I was going to run without GUI.
e) Some mouseover popups would be nice to explain the options, or link to a web page.
f) If I click on Open Folder for a processing video, the only smaller file sitting there is from yesterday, despite having stopped and closed previous tasks. Which path/drive is cwd, and how is that controlled? I would prefer bouncing across my M.2 modules, unless the intermediate is tiny.
g) ffmpeg is using the second slower GPU, not sure why. Less memory, lower chip type.
h) I noticed that clicking into the Export options causes the timer to show some other state (not 15 mins of pass 1/1), see screenshot “weird timer state”. Estimate was 45 mins, ffmpeg showing 2h wall time. Update: 6 wall hours later, it seems that it’s doing 2 passes despite asking for 1 and cancelling the 2-pass job. GPU and CPU >65%.
i) Ran for about 8 hours, crashed with error message asking to add logs, but the tar only seems to contain json files and not the 230MB log file. Adding it here, and starting over from a closed app.
Just a quick update about your log; we suspect this is an Out-of-Memory issue. Please use one filter at a time. And, if the video is very long, we suggest trimming it into smaller sections and processing them individually.
720V B&W 90min video, 32GB of CPU, 16GB of GPU memory, out of memory? This isn’t a lap top.
Which step(s) show memory overflow in the log?
Could you propose a formula how you calculated that, and that it wasn’t a heap overflow or memory leak?
It finally did run with the last settings, took 13 hours on 2080std GPU. It did seem to start over every time I turned the monitor back on.
I am not clear on your response. I do not believe that I made any comments about the PC being a laptop or not.
This is from your logs;
As you can see the app only has 6GB of RAM to use which is being exceeded quickly. Are you using any other apps besides tvai? It certainly could be a memory leak, however, I would rather not jump to conclusions unless we can verify this.
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.
