llmodel.llmodel_mpt_create() causes ntdll.dll crash

[TheItCrOw]

System Info

• Windows11
• Using the python bindings
• Followed the instructions, built all dependencies and I am trying to run the example program to get a simple prompt.
• Gpt4All desktop application works
• Python 3.9.0

Information

• [X] The official example notebooks/scripts
• [ ] My own modified scripts

Related Components

• [ ] backend
• [X] bindings
• [ ] python-bindings
• [ ] chat-ui
• [ ] models
• [ ] circleci
• [ ] docker
• [ ] api

Reproduction

When using the example script:

gptj = GPT4All("ggml-mpt-7b-chat") messages = [{"role": "user", "content": "Name 3 colors"}] print(gptj.chat_completion(messages))

The script starts, loads for a while and then just ends. After debugging, I find the problem to be in the pyllmodel.py script in the MPTModel class (and prolly in all the other model classes as well). Specificially line:

self.model = llmodel.llmodel_mpt_create()

causes the crash. The win event logs state:

_Name der fehlerhaften Anwendung: python.exe, Version: 3.9.150.1013, Zeitstempel: 0x5f7b3dcb Name des fehlerhaften Moduls: ntdll.dll, Version: 10.0.22000.1880, Zeitstempel: 0x40303075 Ausnahmecode: 0xc0000005 Fehleroffset: 0x000000000003599d ID des fehlerhaften Prozesses: 0x5e34 Startzeit der fehlerhaften Anwendung: 0x01d98bce297c3696 Pfad der fehlerhaften Anwendung: C:\Users\Nutzer\AppData\Local\Programs\Python\Python39\python.exe Pfad des fehlerhaften Moduls: C:\WINDOWS\SYSTEM32\ntdll.dll Berichtskennung: bb0b5d97-2daf-47ec-a377-952222fdf774 Vollständiger Name des fehlerhaften Pakets: Anwendungs-ID, die relativ zum fehlerhaften Paket ist: _

Expected behavior

Should print out the prompt.

[mvenditto]

Was the build made with MinGW or msvc?

[TheItCrOw]

With MinGW. I explicitly stated:

cmake -G "MinGW Makefiles" ..

since otherwise I wouldn't have even gotten this far.

[dgiunchi]

same here, in windows 10, my crash happens when loading the model gptj = GPT4All("ggml-mpt-7b-chat") I built with MSYS2, 64 bit, 13

I was able to debug the code with gdb and the error, depending on the model selected (mpt or gptj) is at this line:

LLModelWrapper *wrapper = new LLModelWrapper

Thread 1 hit Breakpoint 2, llmodel_gptj_create () at E:\Experiments\server_VR\vocalMessages\gpt4all\gpt4all\gpt4all-backend\llmodel_c.cpp:14 14 LLModelWrapper *wrapper = new LLModelWrapper; (gdb) s

Thread 1 received signal SIGSEGV, Segmentation fault. 0x00000000000d41fc in ?? ()

[niansa]

Stale, please open a new, updated issue if this is still relevant to you. You're encouraged to open new issues or even PRs if there's anything you need!