ModuleNotFoundError: No module named 'extensions_built_in.llm_models' when initializing job
Hello, I am encountering a ModuleNotFoundError when trying to run a training job. The error occurs during the initialization phase when the toolkit tries to load all built-in extensions, specifically failing inside the AceStepCaptioner. It seems that the extensions_built_in.llm_models module is either missing from the repository, not properly initialized as a submodule, or there is a pathing issue. Steps to Reproduce: Clone/download ai-toolkit (version v0.13.12). Set up a standard training job configuration. Run the job using the provided run script / run.py. The process crashes immediately before starting the actual training steps. Expected Behavior: The job should initialize successfully, load the extensions without errors, and proceed to the training steps. Actual Behavior: The process exits unexpectedly with a ModuleNotFoundError. Error Log / Traceback:
Job process exited unexpectedly. Last log line: ModuleNotFoundError: No module named 'extensions_built_in.llm_models'
Error running job: No module named 'extensions_built_in.llm_models'
Error running on_error: cannot access local variable 'job' where it is not associated with a value
========================================
Result:
- 0 completed jobs
- 1 failure
========================================
Traceback (most recent call last):
File "H:\AI_Toolkit_010826\ai-toolkit\run.py", line 147, in <module>
main()
File "H:\AI_Toolkit_010826\ai-toolkit\run.py", line 132, in main
raise e
File "H:\AI_Toolkit_010826\ai-toolkit\run.py", line 119, in main
job = get_job(config_file, args.name)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "H:\AI_Toolkit_010826\ai-toolkit\toolkit\job.py", line 29, in get_job
return ExtensionJob(config)
^^^^^^^^^^^^^^^^^^^^
File "H:\AI_Toolkit_010826\ai-toolkit\jobs\ExtensionJob.py", line 12, in __init__
self.process_dict = get_all_extensions_process_dict()
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "H:\AI_Toolkit_010826\ai-toolkit\toolkit\extension.py", line 56, in get_all_extensions_process_dict
process_dict[extension.uid] = extension.get_process()
^^^^^^^^^^^^^^^^^^^^^^^
File "H:\AI_Toolkit_010826\ai-toolkit\extensions_built_in\captioner\__init__.py", line 11, in get_process
from .AceStepCaptioner import AceStepCaptioner
File "H:\AI_Toolkit_010826\ai-toolkit\extensions_built_in\captioner\AceStepCaptioner.py", line 120, in <module>
from extensions_built_in.llm_models.src.thinker import ( # noqa: E402 shared fast paths
ModuleNotFoundError: No module named 'extensions_built_in.llm_models'Environment: OS: Windows 10/11 AI-Toolkit Version: v0.13.12 Python Version: 3.12 Installation Method: Git clone Additional Context: I have verified the extensions_built_in directory, and the llm_models folder seems to be missing or not correctly linked. I tried updating via git pull and git submodule update --init --recursive, but the issue persists. Thank you for your help!
Source: ostris/ai-toolkit