Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Bug: Running with ngl option on A10 GPU leads to SIGSEGV with [0.9.0] release #679

Open
yingdachen opened this issue Jan 17, 2025 · 0 comments

Comments

@yingdachen
Copy link

Contact Details

No response

What happened?

Converted llamafile with 0.9.0 release, and run with ngl option on A10 GPU leads to SIGSEGV

launching command:
./qwen2.5-3b-instruct-q4_k_m.llamafile -ngl 999

result:

Image

Things run just fine when no ngl option is given (which runs on CPU):

Image

Things also runs fine when llamafile was converted with previous 0.8.17 release:

Image

Version

llamafile v0.9.0

What operating system are you seeing the problem on?

Linux

Relevant log output

error: Uncaught SIGSEGV (SEGV_MAPERR) at 0x328 on dsw-816703-767cf48f4b-zxn64 pid 483 tid 530
  ./qwen2.5-3b-instruct-q4_k_m.llamafile
  No error information
  Linux Cosmopolitan 4.0.2 MODE=x86_64; #1 SMP Wed Sep 15 17:27:09 CST 2021 dsw-816703-767cf48f4b-zxn64 4.19.91-012.ali4000.alios7.x86_64

RAX 0000000000000320 RBX 00007f71e001682c RDI 0000000000000001
RCX 00007f71ebc10c20 RDX 0000000000000001 RSI 00007f71e838c760
RBP 00007f71e0016860 RSP 00007f71e0016758 RIP 00007f71ebcac439
 R8 0000000012000000  R9 00007f71e0016cf0 R10 0000000000200000
R11 0000000000000001 R12 00007f71e0016974 R13 0000000080ee8920
R14 00007f71e0016cf0 R15 0000000000000000
TLS 00007f7148478c00

XMM0  00000000004000000000000000400000 XMM8  00000000000000000000000000000000
XMM1  00000000000000000000000000400000 XMM9  ffffffffffffffffffffffffffffffff
XMM2  00000000000000000000000000000190 XMM10 ffffffffffffffffffffffffffffffff
XMM3  00007f71e82d3e4000007f71e82d4e10 XMM11 00000000000000000000000000000000
XMM4  000000000000000000007f71e82d4b70 XMM12 00000000000000000000000000000000
XMM5  00000000000000000000000000000000 XMM13 ffffffffffffffffffffffffffffffff
XMM6  203d2032585641207c2030203d20494e XMM14 00000000000000000000000000000000
XMM7  4e565f585641207c2031203d20585641 XMM15 00000000000000000000000000000000

cosmoaddr2line /mnt/workspace/llamafile-0.9.0/qwen2.5-3b-instruct-q4_k_m.llamafile 7f71ebcac439 7f71e83aba88 7f71e82d3205  12000000

7f71e0013580 7f71ebcac439 NULL+0
7f71e0016860 7f71e83aba88 NULL+0
7f71e0016960 7f71e82d3205 NULL+0
<dangerous frame>

000000400000-000000ae21e0 r-xi- 7048kb
000000ae3000-000003251000 rw-i- 39mb
000003251000-0006fe000000       28gb
0006fe000000-0006fe001000 rw-pa 4096b
0006fe001000-7f71442c5000       127tb
7f71442c5000-7f71444c5000 rw-pa 2048kb
7f71444c5000-7f7147a94000       54mb
7f7147a94000-7f714a894000 rw-pa 46mb
7f714a894000-7f71c7fff9a0 r--s- 2007mb
7f71c8000000-7f71cc17c000       65mb
7f71cc17c000-7f71ccd7c000 rw-pa 12mb
7f71ccd7c000-7f71e0005000       307mb
7f71e0005000-7f71e0006000 ---pa 4096b
7f71e0006000-7f71e001a000 rw-pa 80kb
7f71e001a000-7f71ebe4c000       190mb
7f71ebe4c000-7f71ebe4c980 rw-pa 2432b
7f71ebe4d000-7f71ebea1000       336kb
7f71ebea1000-7f71ebfdb600 rw-pa 1258kb
7f71ebfdc000-7f727720e0d1 r--s- 2226mb
7f727720f000-7f727728f000 rw-pa 512kb
7f727728f000-7ffced811000       554gb
7ffced811000-7ffced911000 ---pa 1024kb
7ffced911000-7ffcee111000 rw-pa 8192kb
# 4'562'100'224 bytes in 15 mappings
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

1 participant