core dumped on riscv
I'm trying to run Llama-3.2-1B-Instruct-Q8_0.gguf on riscv by using llama.cpp, it gives illegal instruction core dumped error.
which instruction? which riscv processor?
Hifive unmatched hardware board
find the instruction that caused this, you can use gdb for this
gdb doesn't work
gdb doesn't work
what is the error?
gdb doesn't work for the command ./llama-cli
gdb doesn't work for the command ./llama-cli
very strange
- try compile with debug
- use gdb with '--args' argument
- if not working - google 'gdb connect to pid'
- also may try use gdb server runned on riscv & connect from "standart" pc (x86_64 on example) via gdb
- also you may try use strace (may be, i don't know)
- also you may run gui-terminal or bash under gdb)) & run llama in it
- hope everything works out for you before this line)))
Can you tell me how you compiled the RISC-V version? I'm currently encountering some issues and can't use static compilation. When I try to use static compilation, I get some warnings:
/home/share/ToolChain/riscv64-unknown-linux-toolsuite-ubuntu-18/bin/../lib/gcc/riscv64-unknown-linux-gnu/13.2.0/../../../../riscv64-unknown-linux-gnu/bin/ld: ../../ggml/src/libggml.a(ggml-backend-reg.cpp.o): in function `.L0 ':
/home//llama.cpp/ggml/src/ggml-backend-reg.cpp:133:(.text+0x286e): warning: Using 'dlopen' in statically linked applications requires at runtime the shared libraries from the glibc version used for linking
Then, when running, it directly results in illegal instructions, occurring in _dl_aux_init().
Thanks!
This issue was closed because it has been inactive for 14 days since being marked as stale.