tjake
The qwen2 model downloaded from Huggingface reports an error during runtime
Qwen2.5-Coder-3B-Instruct Qwen2-0.5B jdk 21 macOS
A fatal error has been detected by the Java Runtime Environment:
SIGSEGV (0xb) at pc=0x0000000102ea3430, pid=31439, tid=88067
JRE version: OpenJDK Runtime Environment Microsoft-11369939 (21.0.7+6) (build 21.0.7+6-LTS)
Java VM: OpenJDK 64-Bit Server VM Microsoft-11369939 (21.0.7+6-LTS, mixed mode, sharing, tiered, compressed oops, compressed class ptrs, g1 gc, bsd-aarch64)
Problematic frame:
[thread 87555 also had an error]
C [libjlama.dylib+0x3430] gemm_bf16_128_arm+0x290
No core dump will be written. Core dumps have been disabled. To enable core dumping, try "ulimit -c unlimited" before starting Java again
[thread 128771 also had an error] [thread 129539 also had an error]
JFR recording file will be written. Location: /Users/xx/aiProjects/xxx/hs_err_pid31439.jfr
An error report file with more information is saved as:
/Users/peng/aiProjects/ictriskplatform/hs_err_pid31439.log
[29.399s][warning][os] Loading hsdis library failed
If you would like to submit a bug report, please visit:
https://github.com/microsoft/openjdk/issues
The crash happened outside the Java Virtual Machine in native code.
See problematic frame for where to report the bug.
Disconnected from the target VM, address: '127.0.0.1:56073', transport: 'socket'
Can't I run the model I downloaded myself, only you can run it https://huggingface.co/tjake Issued models
hello, can you give some more details about how you are running jlama? Can you also provide some details about the computer you're running it on?
It's Spring introducing jlama's jar through Maven, and then running the model with the case code you provided. The problem now is that you https://huggingface.co/tjake The Qwen model can be executed by publishing it, but the Qwen model I downloaded myself cannot run and reports an error.
My computer parameters are
MacBook Air
Chip: Apple M2
Memory: 16GB
macos:15.5
How did you download the models? Can you post your main code as well? Maybe you're not referencing your downloaded models correctly in the model cache path?
Looks like an issue in the native code. Can you tell me the exact model you are using (with link) and how you are loading it in the code?
