Getting a traceback when generating helptext

[jherrman]

I'm getting the following error message when running logdetective --help :

---

Traceback (most recent call last):
File "/usr/bin/logdetective", line 5, in
from logdetective.logdetective import main
File "/usr/lib/python3.12/site-packages/logdetective/logdetective.py", line 6, in
from logdetective.utils import process_log, initialize_model, retrieve_log_content, format_snippets
File "/usr/lib/python3.12/site-packages/logdetective/utils.py", line 7, in
from llama_cpp import Llama
File "/usr/lib64/python3.12/site-packages/llama_cpp/init.py", line 1, in
from .llama_cpp import *
File "/usr/lib64/python3.12/site-packages/llama_cpp/llama_cpp.py", line 1434, in
@ctypes_function(
^^^^^^^^^^^^^^^^
File "/usr/lib64/python3.12/site-packages/llama_cpp/llama_cpp.py", line 122, in decorator
func = getattr(lib, name)
^^^^^^^^^^^^^^^^^^
File "/usr/lib64/python3.12/ctypes/init.py", line 392, in getattr
func = self.getitem(name)
^^^^^^^^^^^^^^^^^^^^^^
File "/usr/lib64/python3.12/ctypes/init.py", line 397, in getitem
func = self._FuncPtr((name_or_ordinal, self))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: /usr/lib64/libllama.so: undefined symbol: llama_model_apply_lora_from_file

---

System: Fedora 40
Hardware: ThinkPad T14

[TomasTomecek]

Thanks for the report!

The issue is very simple :) python-llama-cpp library cannot find certain symbols inside llama-cpp library in F40 - these 2 libraries are out of sync. So I though I'd just update python-llama-cpp and we'd be good.

Unfortunately, it's not that easy.

llama-cpp was recently updated in f40:

commit 00ddbdbfead3eadaf58d849604254b1b6a2668f3 (HEAD -> f40, origin/f40)
Author: Tom Rix <[email protected]>
Date:   Sat Oct 26 09:00:20 2024 -0700

    Update to b3561

Even though when I update python-llama-cpp to "0.3.1", the latest upstream release, logdetective will still pick the older version due to our dependency setting:

[tool.poetry.dependencies]
llama-cpp-python = "^0.2.56,!=0.2.86"
...

Which causes:

Requirement already satisfied: llama-cpp-python!=0.2.86,<0.3.0,>=0.2.56 in /usr/lib64/python3.13/site-packages (from logdetective==0.2.7) (0.2.75)

Downgrading:
 python3-llama-cpp-python     x86_64     0.2.75-5.fc40   updates   280 k

So in order to fix F40, we need to:

1. Update the llama-cpp-python dependency requirement here
2. Check if we need to update llama-cpp-python itself in F40

[jpodivin]

Should be helped by #87

[TomasTomecek]

Finally figured it out, thanks everyone for help!

https://src.fedoraproject.org/rpms/python-llama-cpp-python/pull-request/10

Will merge, build and do a bodhi update soon.

Smoke-tested in a f40 container and all worked well:

[root@1f5dc2461be6 src]# logdetective --help
usage: logdetective [-h] [-M MODEL] [-F FILENAME_SUFFIX] [-S SUMMARIZER] [-N N_LINES] [-C N_CLUSTERS] [-v] [-q] file

positional arguments:
  file                  The URL or path to the log file to be analyzed.

options:
  -h, --help            show this help message and exit
  -M MODEL, --model MODEL
                        The path or Hugging Face name of the language model for analysis.
  -F FILENAME_SUFFIX, --filename_suffix FILENAME_SUFFIX
                        Suffix of the model file name to be retrieved from Hugging Face. Makes sense only if the model is specified with Hugging Face name.
  -S SUMMARIZER, --summarizer SUMMARIZER
                        Choose between LLM and Drain template miner as the log summarizer. LLM must be specified as path to a model, URL or local file.
  -N N_LINES, --n_lines N_LINES
                        The number of lines per chunk for LLM analysis. This only makes sense when you are summarizing with LLM.
  -C N_CLUSTERS, --n_clusters N_CLUSTERS
                        Number of clusters for Drain to organize log chunks into. This only makes sense when you are summarizing with Drain
  -v, --verbose
  -q, --quiet

  Installing       : llama-cpp-b3561-1.fc40.x86_64
  Installing       : python3-llama-cpp-python-0.2.87-1.fc40.x86_64
  Installing       : python3-logdetective-0.2.5-1.fc40.noarch

[TomasTomecek]

https://bodhi.fedoraproject.org/updates/FEDORA-2024-25db690b63

please provide karma if the update fixes this problem

@jherrman thanks again for reporting!