Installation error on Ubuntu Server 24.04.4 LTS

[eumel1968]

Hello,

My operating system:

Distributor ID: Ubuntu
Description: Ubuntu 24.04.4 LTS
Release: 24.04
Codename: noble

First i do:

sudo apt-get update

and

sudo apt-get upgrade

I used the command

curl -fsSL https://lmstudio.ai/install.sh | bash

Then I get the following message:

Installation finished successfully! llmster is ready to launch.
To start the daemon, run:

    lms daemon up

main: line 603:  2317 Illegal instruction     (core dumped) "$_lms_installed_path" bootstrap -y > /dev/null 2>&1

Could not add lms to your PATH
lms install location: /root/.lmstudio/bin/LMS

If I Verify the installation:

lms --help

I get the following message:

Command 'lms' not found, did you mean:
  command 'wms' from snap wms (0.0.13)
  command 'lsm' from deb lsm (1.0.4-2)
  command 'ls' from deb coreutils (9.4-3ubuntu6.2)
  command 'ams' from deb ams (2.2.1-1)
  command 'dms' from deb anacrolix-dms (1.5.0-2ubuntu0.24.04.3)
  command 'lmms' from deb lmms (1.2.2+dfsg1-6build2)
  command 'les' from deb atm-tools (1:2.5.1-5)
  command 'fms' from deb openafs-client (1.8.10-2.1ubuntu3.4)
  command 'lmp' from deb lammps (20231121+dfsg-4)
  command 'pms' from deb pms (0.42-1.1)
  command 'lvs' from deb lvm2 (2.03.16-3ubuntu3.2)
  command 'lrs' from deb lrslib (0.71b-2)
See 'snap info <snapname>' for additional versions.

The problem also occurs when I use: curl -fsSL https://lmstudio.ai/install.sh | sudo bash

What am I doing wrong?

What do I need to do to use LM Studio on Ubuntu Server 24.04.4 LTS?

[sherrellbc]

Related: lmstudio-ai/lmstudio-bug-tracker#1885

Your CPU does not support AVX -512 extensions.

[sherrellbc]

Looks like it's probably related to the way the tool is bundled with make-bin-linux.sh. Consider trying the x86-baseline package from bun. It looks like other packages make some host assumptions.

[eumel1968]

My CPU supports:

AVX-512 Galois Fields New Instructions
AVX-512 Vector AES
AVX-512 Carry-Less Multiplication Quadword (VPLCLMULQDQ)

Extensions.

Is that not enough?

And what does the x86-baseline package from bun refer to?

I'm new to Linux, so please explain it in a way that even a complete beginner can understand.

[sherrellbc]

Perhaps your issue is different, then. Which CLI version are you using? Also, try this for me and report back the results:

$ strace ~/.lmstudio/bin/lms

You should see something like this at the end of the output.

--- SIGILL {si_signo=SIGILL, si_code=ILL_ILLOPN, si_addr=0x407e878} ---
+++ killed by SIGILL +++
Illegal instruction

[eumel1968]

I'm using the version I installed with the command curl -fsSL https://lmstudio.ai/install.sh | bash. llmster 0.0.15-2 Linux x86_64

The Results:

fredl@playground:~$ strace ~/.lmstudio/bin/lms
execve("/home/fredl/.lmstudio/bin/lms", ["/home/fredl/.lmstudio/bin/lms"], 0x7ffe914e8830 /* 22 vars */) = 0
brk(NULL)                               = 0x22ea4000
mmap(NULL, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7a5afe08b000
access("/etc/ld.so.preload", R_OK)      = -1 ENOENT (No such file or directory)
openat(AT_FDCWD, "/etc/ld.so.cache", O_RDONLY|O_CLOEXEC) = 3
fstat(3, {st_mode=S_IFREG|0644, st_size=23675, ...}) = 0
mmap(NULL, 23675, PROT_READ, MAP_PRIVATE, 3, 0) = 0x7a5afe085000
close(3)                                = 0
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libc.so.6", O_RDONLY|O_CLOEXEC) = 3
read(3, "\177ELF\2\1\1\3\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0\220\243\2\0\0\0\0\0"..., 832) = 832
pread64(3, "\6\0\0\0\4\0\0\0@\0\0\0\0\0\0\0@\0\0\0\0\0\0\0@\0\0\0\0\0\0\0"..., 784, 64) = 784
fstat(3, {st_mode=S_IFREG|0755, st_size=2125328, ...}) = 0
pread64(3, "\6\0\0\0\4\0\0\0@\0\0\0\0\0\0\0@\0\0\0\0\0\0\0@\0\0\0\0\0\0\0"..., 784, 64) = 784
mmap(NULL, 2170256, PROT_READ, MAP_PRIVATE|MAP_DENYWRITE, 3, 0) = 0x7a5afde00000
mmap(0x7a5afde28000, 1605632, PROT_READ|PROT_EXEC, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x28000) = 0x7a5afde28000
mmap(0x7a5afdfb0000, 323584, PROT_READ, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x1b0000) = 0x7a5afdfb0000
mmap(0x7a5afdfff000, 24576, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x1fe000) = 0x7a5afdfff000
mmap(0x7a5afe005000, 52624, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS, -1, 0) = 0x7a5afe005000
close(3)                                = 0
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libpthread.so.0", O_RDONLY|O_CLOEXEC) = 3
read(3, "\177ELF\2\1\1\0\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0\0\0\0\0\0\0\0\0"..., 832) = 832
fstat(3, {st_mode=S_IFREG|0644, st_size=14408, ...}) = 0
mmap(NULL, 16400, PROT_READ, MAP_PRIVATE|MAP_DENYWRITE, 3, 0) = 0x7a5afe080000
mmap(0x7a5afe081000, 4096, PROT_READ|PROT_EXEC, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x1000) = 0x7a5afe081000
mmap(0x7a5afe082000, 4096, PROT_READ, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x2000) = 0x7a5afe082000
mmap(0x7a5afe083000, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x2000) = 0x7a5afe083000
close(3)                                = 0
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libdl.so.2", O_RDONLY|O_CLOEXEC) = 3
read(3, "\177ELF\2\1\1\0\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0\0\0\0\0\0\0\0\0"..., 832) = 832
fstat(3, {st_mode=S_IFREG|0644, st_size=14408, ...}) = 0
mmap(NULL, 16400, PROT_READ, MAP_PRIVATE|MAP_DENYWRITE, 3, 0) = 0x7a5afe07b000
mmap(0x7a5afe07c000, 4096, PROT_READ|PROT_EXEC, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x1000) = 0x7a5afe07c000
mmap(0x7a5afe07d000, 4096, PROT_READ, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x2000) = 0x7a5afe07d000
mmap(0x7a5afe07e000, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x2000) = 0x7a5afe07e000
close(3)                                = 0
openat(AT_FDCWD, "/lib/x86_64-linux-gnu/libm.so.6", O_RDONLY|O_CLOEXEC) = 3
read(3, "\177ELF\2\1\1\3\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0\0\0\0\0\0\0\0\0"..., 832) = 832
fstat(3, {st_mode=S_IFREG|0644, st_size=952616, ...}) = 0
mmap(NULL, 950296, PROT_READ, MAP_PRIVATE|MAP_DENYWRITE, 3, 0) = 0x7a5afdd17000
mmap(0x7a5afdd27000, 520192, PROT_READ|PROT_EXEC, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x10000) = 0x7a5afdd27000
mmap(0x7a5afdda6000, 360448, PROT_READ, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x8f000) = 0x7a5afdda6000
mmap(0x7a5afddfe000, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0xe7000) = 0x7a5afddfe000
close(3)                                = 0
mmap(NULL, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7a5afe079000
mmap(NULL, 331776, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7a5afe028000
arch_prctl(ARCH_SET_FS, 0x7a5afe0772c0) = 0
set_tid_address(0x7a5afe077590)         = 2403
set_robust_list(0x7a5afe0775a0, 24)     = 0
rseq(0x7a5afe077be0, 0x20, 0, 0x53053053) = 0
mprotect(0x7a5afdfff000, 16384, PROT_READ) = 0
mprotect(0x7a5afddfe000, 4096, PROT_READ) = 0
mprotect(0x7a5afe07e000, 4096, PROT_READ) = 0
mprotect(0x7a5afe083000, 4096, PROT_READ) = 0
mprotect(0x7a5afe0c3000, 8192, PROT_READ) = 0
prlimit64(0, RLIMIT_STACK, NULL, {rlim_cur=8192*1024, rlim_max=RLIM64_INFINITY}) = 0
munmap(0x7a5afe085000, 23675)           = 0
futex(0x66efbbc, FUTEX_WAKE_PRIVATE, 2147483647) = 0
futex(0x66efbc8, FUTEX_WAKE_PRIVATE, 2147483647) = 0
getrandom("\x8a\x95\xb4\x76\x6e\xd4\x6b\x0a\xc6\x7f\x2a\xe1\x59\xfd\xbb\xc1\x21\x66\x7d\x07\xeb\xdd\x8c\x0d\x53\x85\x1f\xb4\x9b\xc3\xfb\x2e", 32, GRND_NONBLOCK) = 32
--- SIGILL {si_signo=SIGILL, si_code=ILL_ILLOPN, si_addr=0x407e878} ---
+++ killed by SIGILL (core dumped) +++
Illegal instruction (core dumped)

[sherrellbc]

Hm, so are you certain your CPU supports AVX? This is the opcode at that address, which is using AVX extensions. (It's just regular AVX/2, not -512).

.text:000000000407E878                 vmovdqu ymm0, [rbp+var_50]
.text:000000000407E87D                 jmp     loc_407EB74

What does your CPU info say? For instance:

$ cat /proc/cpuinfo | grep flags | sort -u
flags		: fpu ... avx avx2 ...

Are you running lms on the host CPU or are you running it inside a VM?

[eumel1968]

I'm running lms in a VM with VirtualBox 7.2.8 r173730 (Qt 6.8.0 on Windows) on Windows 11 Pro 25H2 Build 26200.8655.

Hwinfo (https://www.hwinfo.com/download/) shows a green checkmark for the three AVX-512 extensions I mentioned above.

fredl@playground:~$ cat /proc/cpuinfo | grep flags | sort -u
flags           : fpu ... avx avx2 ...
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq ssse3 cx16 pcid sse4_1 sse4_2 movbe popcnt aes rdrand hypervisor lahf_lm abm 3dnowprefetch ibrs_enhanced fsgsbase bmi1 bmi2 invpcid rdseed adx clflushopt sha_ni arat md_clear flush_l1d arch_capabilities
Command 'flags' not found, did you mean:
  command 'flagd' from snap flagd (v0.2.5)
  command 'pflags' from deb python3-linux-procfs (0.6.3-1.1build1)
See 'snap info <snapname>' for additional versions.

My CPU: Intel Core i9-13900KF

[sherrellbc]

Yeah so your host CPU (what Windows is running on) supports AVX, but your virtual CPUs presented inside the VM do not. You should be able to find the appropriate config setting in VirtualBox to enable CPU features (or make them mirror the host CPUs).

That's your problem here. If you were to run LMStudio natively in Windows it would work fine, but you need to pass through the appropriate CPU options into your virtual machine.

[eumel1968]

I installed the guest extension as described here:

https://gist.github.com/otienog1/2aefec1a2c75951e325020c377053510

I’m going to try the following next: https://stackoverflow.com/questions/54251855/virtualbox-enable-nested-vtx-amd-v-greyed-out

I’ll let you know as soon as I’ve tried it.

[eumel1968]

Sorry for the late reply.

After many unsuccessful attempts, I got frustrated and installed VirtualBox on my second computer—which is about 10 years old and runs Windows 10 Pro—and then set up an Ubuntu Server VM.

LMS installs here without any error messages and runs smoothly.

So it’s almost certainly not an Ubuntu guest or VirtualBox/host configuration issue. Rather, it’s likely a Windows 11 problem.

Who knows what Microsoft has managed to mess up again!

[eumel1968]

So I decided to ask Microsoft Q&A:

https://learn.microsoft.com/de-de/answers/questions/5919799/die-avx-und-avx2-befehlss-tze-meiner-cpu-werden-ni

Since German is my native language, I submitted the request in German..

The text translated into English:

Hello Microsoft,

I have installed Ubuntu Server 24.04.4 LTS in a VM running VirtualBox version 7.2.8 r173730 (Qt6.8.0 on Windows).

My host operating system: Windows 11 Pro; Version: 25H2; Build: 26200.8655; Windows Feature Experience Pack 1000.26100.315.0

Host hardware: CPU: i9-13900KF @ 5 GHz; GPU: Nvidia RTX 4090 OC; RAM: 64 GB DDR5 @6000; 3 x 2 TB NVMe drives; 3 x 4 TB SATA SSDs

Guest hardware: Processors: 8; Main memory: 8192 MB; Graphics memory: 256 MB; Graphics controller: VMSVGA; 3D acceleration: enabled

Started the VM, logged in. And installed LMS with `curl -fsSL https://lmstudio.ai/install.sh | bash`.

Error messages appeared. After extensive research, I figured out what the problem was. The AVX and AVX2 instruction sets of my CPU are not being passed on to the VM.

See, for example, here:

https://github.com/lmstudio-ai/lms/issues/582

https://github.com/lmstudio-ai/lmstudio-bug-tracker/issues/1885

https://github.com/lmstudio-ai/lms/issues/584

https://www.wer-weiss-was.de/t/wie-aktiviere-ich-die-avx-und-avx2-befehlssaetze-in-einer-gast-vm/9531871

On my second computer, which is about 10 years old and runs Windows 10 Pro,

I installed VirtualBox and then set up an Ubuntu Server VM.

I then installed LMS using `curl -fsSL https://lmstudio.ai/install.sh | bash`.

LMS installed without any error messages and works without any issues.

So it is almost certainly not an Ubuntu guest or VirtualBox/host configuration issue. Rather, it is likely a Windows 11 issue.

Please restore this functionality.

Translated with Deepl (https://www.deepl.com/)