Kate

same issue on wsl2 fastertransformer works fine but can only run 350M model whereas i could run 2B one with python backend ```bash $ uname -a Linux puter 5.15.90.1-microsoft-standard-WSL2 #1...

The first link you provide has little to do with the problem because we are talking about any non-root user which by default in the linux kernel cannot bind to...

While waiting for this issue to get solved : The localStorage allowed size per origin varies browser to browser. It is 2.5 MB on Chrome, 5 MB on Firefox by...

heres another log with a mixintrace ``` ---- Minecraft Crash Report ---- // I feel sad now :( Time: 1/26/21, 1:42 PM Description: Watching Server java.lang.Error: Watchdog at [email protected]/jdk.internal.misc.Unsafe.park(Native Method)...

so just to check that its not actually the disk write speed throttling the operation somehow i ran a quick benchmark ``` dd if=/dev/random of=/tmp/output bs=8k count=1000k; rm -f /tmp/output...

The more I look into this the more it looks like a problem with the vanilla chunk saving module not being able to keep up with chunks being generated at...

I wrote a quick mixin to hook on the chunk save queue method and the actual io write return result. The result is what was expected. The queue saves just...

I just replicated this on an almost vanilla server, so I'm pretty much sure that the issue is just vanilla TACS at this point, from what I gathered snooping around...

I did benchmarks (see above) and write speed is not the limiting factor. Typical Java IO should be around 2.5 to 3 times slower than native code so I would...

So I did some more testing to figure out exactly what happens once the amount of chunks in memory gets close to or over the amount of free memory (maybe...