LATO.2 Setup Check:
LATO2SetupCheck, displayed in ComfyUI as “LATO.2 Setup Check” under the “3d/LATO.2” category, helps you check whether the supporting files for LATO.2 low-poly remeshing are in place before you attempt to process a model. It reports setup information about LATO.2’s separate Python environment, source code, and model weights, including where the extension looks for these resources and how to install them. This is useful when setting up the extension for the first time, moving model files to another drive, or investigating a remeshing workflow that cannot start. The runtime’s setup checks look for configured paths and LATO.2 weight files rather than performing a complete remeshing test, so a successful setup report should not be treated as proof that every dependency or GPU operation will work. The node is diagnostic: it does not install missing components or modify your mesh.
LATO.2 Setup Check Input Parameters:
This node has no input parameters. You do not need to connect a mesh, enter a path, or adjust any controls to run it. Its setup information comes from the extension’s default installation layout, any settings in lato2_config.json, and applicable LATO2_ environment variables; these are external configuration settings, not node inputs.
LATO.2 Setup Check Output Parameters:
status
The status output is a STRING containing the setup report returned by the extension’s status-reporting function. Its purpose is to help you understand whether LATO.2’s environment, code, and weights are in place and where the extension expects to find them, with installation guidance as described by the node. The same text is also sent to ComfyUI for display, allowing you to inspect the report directly or connect the string to a compatible text-display or text-processing node. Treat it as readable diagnostic information rather than a numerical score or a guaranteed success flag: the supplied source does not define a fixed report format, and the underlying path checks do not verify a complete GPU inference run.
LATO.2 Setup Check Usage Tips:
- Run this node before using “LATO.2 Low-Poly Remesh,” especially after installation or after moving files. It is marked as an output node, so it can run without a downstream connection.
- Run the workflow again after correcting a setup issue. The node’s change-detection method returns
NaN, which makes ComfyUI reevaluate it rather than reuse a previously cached result when it is executed. - If setup components are missing, run
python install_lato2.pyfrom theComfyUI/custom_nodes/ComfyUI-LATO2directory, then restart ComfyUI. The supplied installer targets Windows or Linux with an NVIDIA GPU and requires Git and internet access; it creates a separate environment without installing LATO.2’s pinned dependencies into ComfyUI’s own Python. - Check the expected default locations before changing configuration: the environment is in
envs/lato2inside the extension, the source code is invendor/LATO.2, and weights normally reside in ComfyUI’smodels/lato2directory. The environment’s Python executable isScripts/python.exeon Windows orbin/pythonon Linux. - To install weights on another drive, use
python install_lato2.py --models-dir "PATH_TO_MODELS". The installer supports writinglato2_config.jsonwhen you select a non-default weights location, helping the extension locate those files afterward. - When troubleshooting an unexpected path, check configuration precedence: defaults are overridden by
lato2_config.json, and matching environment variables override those settings. Relevant variables includeLATO2_PYTHON,LATO2_REPO,LATO2_CKPT, andLATO2_DINO_HUB_DIR. An old environment-variable override can keep directing the extension to a folder you no longer use. - Use the installer’s
--onlyoption to repair selected components when appropriate, for examplepython install_lato2.py --only envorpython install_lato2.py --only models. There are no node-level performance controls; the runtime’sLATO2_TIMEOUTsetting controls remeshing execution time, not the thoroughness of this setup check.
LATO.2 Setup Check Common Errors and Solutions:
python is not set / repo is not set / ckpt is not set
- Explanation: These are diagnostic strings defined by the shared setup checker. They mean a required configuration value is empty or missing:
pythonidentifies LATO.2’s Python executable,repoidentifies its source-code folder, andckptidentifies its weights folder. - Solution: Check
lato2_config.jsonand the correspondingLATO2_PYTHON,LATO2_REPO, orLATO2_CKPTenvironment variable for empty values. Correct the setting or remove an unnecessary override to use the default installation location.
python does not exist: PATH
- Explanation: The configured Python executable cannot be found. The separate LATO.2 environment may not have been installed, may have been deleted, or may still be referenced at an old location.
- Solution: Run
python install_lato2.py --only envfrom the extension directory. If you intentionally use another compatible environment, pointpythonin the configuration orLATO2_PYTHONto its actual executable, not just its containing folder.
repo does not exist: PATH
- Explanation: The configured LATO.2 source-code path does not exist. Installing the ComfyUI extension alone does not necessarily mean that the separate LATO.2 repository has been downloaded.
- Solution: Run
python install_lato2.py --only repowith Git installed and internet access available. If the code already exists elsewhere, correct thereposetting orLATO2_REPOoverride.
ckpt does not exist: PATH
- Explanation: The configured weights path does not exist, often because the models were not downloaded, were moved, or are stored on a drive that is currently unavailable.
- Solution: Restore access to the intended drive or folder, correct
ckptorLATO2_CKPT, or run the installer’s models step. If you are choosing a new location, include--models-dir "PATH_TO_MODELS"so the installer uses that destination.
no LATO.2 weights (*.pt) in PATH
- Explanation: The required paths exist, but the checker found no files matching
*.ptdirectly inside the configured weights directory. This can happen after an incomplete download or when the configured path points to a parent folder instead of the folder containing the weights. - Solution: Check that the actual LATO.2
.ptfiles are directly inside the configuredckptdirectory. Correct the path or runpython install_lato2.py --only modelsto obtain the weights; an empty folder or DINOv2 files alone will not satisfy this check.
JSONDecodeError while reading lato2_config.json
- Explanation: The configuration loader reads
lato2_config.jsonas JSON without catching parsing errors. Invalid syntax, such as a trailing comma, a missing quotation mark, or improperly escaped Windows backslashes, can therefore stop the setup check instead of producing a normal report. - Solution: Repair the file using a JSON-aware editor. Use double-quoted keys and values, avoid trailing commas, and use forward slashes or escaped backslashes in Windows paths. If you do not need custom settings, back up and remove the file to return to defaults, while also checking for environment-variable overrides.
LATO.2 is not set up: ... Run python ".../install_lato2.py" and restart ComfyUI.
- Explanation: This message is raised by the remeshing runtime, rather than directly by the setup-check node, when the shared setup checker finds missing resources. The text between the introduction and installation instruction lists the detected problems.
- Solution: Use the setup report to identify the missing paths or weights, repair those components or run the full installer, and restart ComfyUI before retrying remeshing.
PyTorch does not see a CUDA GPU
- Explanation: This is an assertion message from the installer’s environment verification, not a check performed by this node. It means that PyTorch in LATO.2’s separate environment cannot access a CUDA GPU; existing folders and weights do not rule out this problem.
- Solution: Confirm that you have a supported NVIDIA GPU and a working NVIDIA driver, then rerun the environment installation step. Verify GPU access in the separate LATO.2 environment rather than assuming that ComfyUI’s own GPU access guarantees the same result.
