SenseNova U1.5 8-Step LoRA (Final only):
The SenseNova U1.5 8-Step LoRA (Final only) node applies the verified official eight-step LoRA adaptation to a supported SenseNova U1.5 Final model, preparing it for the corresponding short sampling workflow. A LoRA is a set of targeted adjustments to an existing model rather than a replacement for the entire model. Here, those adjustments support an eight-step generation setup, reducing the number of sampling steps needed in the intended workflow. This is a specialized loader, not a general-purpose style LoRA loader: it checks the model’s identity, verifies the exact supported LoRA file and its internal structure, and confirms that all expected adjustments are applied. These safeguards help prevent incompatible combinations, modified files, incomplete loading, and accidental application of the same eight-step adaptation twice. It is intended only for supported official Final checkpoints, not the SFT variant, and it does not configure the sampler automatically.
SenseNova U1.5 8-Step LoRA (Final only) Input Parameters:
model
A required MODEL connection containing a supported SenseNova U1.5 Final model loaded through the SenseNova U1.5 loader. The node checks the model’s attached checkpoint information to confirm that its variant is Final, its source repository matches the expected official repository, and its revision is among the supported Final revisions. A model from another loader without that information, an SFT model, or an unsupported checkpoint revision is rejected. The node also checks whether this eight-step LoRA has already been applied, preventing duplicate application. These checks run even when the adaptation strength is zero. This connection has no numeric range or default value; you must supply a compatible model.
lora_name
A required COMBO selection identifying the LoRA file to load from ComfyUI’s registered loras locations, typically including ComfyUI/models/loras. The available options are generated from the files registered in those locations; the node does not declare a fixed list or an explicit default selection. The provided example uses SenseNova-U1.5-8B-MoT-LoRA-8step-ComfyUI.safetensors. Although other files may appear in the dropdown, the node accepts only the verified official ComfyUI conversion: it checks the file’s size and SHA256 fingerprint, its metadata, and its tensor structure. Renaming another LoRA to match the example filename will not make it compatible. With a nonzero strength, a missing, altered, or incompatible file stops execution rather than being partially applied; with a strength of zero, file loading and LoRA validation are skipped after the model checks pass.
strength_model
A required FLOAT controlling how strongly the LoRA adjustments are applied to the model, with a default of 1.0, a minimum of 0.0, a maximum of 2.0, and an adjustment step of 0.01. A value of 1.0 applies the adjustments at their stored scale and matches the supplied eight-step workflow example. Values between 0.0 and 1.0 reduce their contribution, while values above 1.0 amplify it; neither direction is guaranteed to improve image quality or preserve the intended eight-step behavior. At exactly 0.0, the node returns the incoming model unchanged without reading or validating the selected LoRA file, although model compatibility and duplicate-application checks still run. Treat this as the strength of the specialized eight-step adaptation, not as a general image-quality or detail control.
SenseNova U1.5 8-Step LoRA (Final only) Output Parameters:
MODEL
The MODEL output contains the model with the verified eight-step LoRA adjustments applied at the selected strength. For a nonzero strength, the node clones the incoming model, applies the adjustments to that clone, and confirms that every expected model patch was accepted, leaving the incoming model object unmodified by this operation. It also attaches the LoRA metadata and an eight-step identification record to the returned model, allowing subsequent checks to recognize the adaptation and prevent it from being stacked twice. Connect this output to the appropriate downstream model-processing or sampling nodes; it is a model object, not a generated image or latent image. When the strength is exactly zero and model validation succeeds, the output is the original incoming model without the adaptation.
SenseNova U1.5 8-Step LoRA (Final only) Usage Tips:
- Start with a supported official SenseNova U1.5 Final checkpoint loaded through the SenseNova U1.5 loader. Do not substitute the SFT variant: the validation message explicitly identifies SFT as using 50 steps.
- Place the verified ComfyUI-converted LoRA in a registered LoRA folder and select it by name. The example filename is
SenseNova-U1.5-8B-MoT-LoRA-8step-ComfyUI.safetensors; matching the filename alone is not sufficient because the contents are verified. - Use
strength_model= 1.0 as your starting point. This matches the supplied example and avoids weakening or exaggerating the intended adaptation before you have established a working baseline. - Configure sampling separately. The supplied example uses KSampler with 8 steps, CFG 1, the
eulersampler, thenormalscheduler, and denoise 1. Applying this node does not automatically change those settings. - Follow the supplied workflow’s model path: SenseNova U1.5 loader → this LoRA node → SenseNovaSamplingOptions → KSampler. The example also uses a 2048 × 2048 latent image with batch size 1; treat these as example settings rather than restrictions enforced by this node.
- Apply the eight-step LoRA only once per model path. For independent strength comparisons, branch from the original compatible Final model and apply the node separately on each branch rather than chaining multiple copies.
- Use a strength of 0.0 to pass through a compatible, unpatched Final model without loading the LoRA. This is useful for checking workflow connections, but it neither verifies the selected LoRA file nor establishes that eight-step sampling is appropriate for the unpatched model.
- Keep the verified file unchanged. Re-saving it, changing its precision, editing its metadata, or using a different conversion can trigger validation failures even if it originated from the official LoRA.
SenseNova U1.5 8-Step LoRA (Final only) Common Errors and Solutions:
SenseNova U1.5 8-step LoRA requires a model from the SenseNova U1.5 loader
- Explanation: The connected model does not carry the checkpoint information required to verify its origin. A MODEL connection alone does not establish compatibility.
- Solution: Load the supported official Final checkpoint through the SenseNova U1.5 loader and connect its model output to this node. If another node sits between them, check whether connecting the loader directly resolves the problem.
SenseNova U1.5 8-step LoRA requires its fixed official Final checkpoint; SFT uses 50 steps
- Explanation: The checkpoint information identifies an unsupported variant, source repository, or revision. This node is restricted to specific official Final checkpoints.
- Solution: Select a supported official Final checkpoint using the SenseNova U1.5 loader. If you intend to use SFT, remove this eight-step LoRA node and use the appropriate SFT workflow instead.
SenseNova U1.5 8-step LoRA is already applied; do not stack it twice
- Explanation: The incoming model already has the eight-step adaptation recorded on it. This check also runs when the current node’s strength is zero.
- Solution: Remove the duplicate node or reconnect it to the original unpatched Final model. To change strength, adjust the first application rather than applying another copy afterward.
LoRA file cannot be found or opened
- Explanation: The selected file cannot be resolved in the registered LoRA locations, or the file cannot be accessed. The exact error wording depends on ComfyUI and the operating system.
- Solution: Confirm that the file exists in a registered LoRA folder, refresh the file list or restart ComfyUI if necessary, and select it again. Check read permissions if the file is visible but cannot be opened.
LoRA file size/SHA256 does not match the verified SenseNova U1.5 8-step conversion
- Explanation: The selected file does not exactly match the expected verified conversion. Possible causes include an incomplete download, a different LoRA, an unconverted source file, or a file that has been edited or re-saved.
- Solution: Obtain the verified official ComfyUI conversion expected by this extension and select it without modifying it. Renaming a different file or changing strength will not resolve a fingerprint mismatch.
LoRA is not the verified official SenseNova U1.5 8-step ComfyUI conversion
- Explanation: The file’s metadata is missing or differs from the required conversion information, such as its source revision, format identifier, or conversion marker.
- Solution: Replace the file with the verified ComfyUI conversion. Avoid metadata editors or conversion tools that rewrite the file, and keep the extension and its supported model assets aligned.
SenseNova U1.5 8-step LoRA tensor count mismatch: N
- Explanation: The file contains a different number of stored tensors than expected. A tensor is a block of numerical data used to describe the adaptation; an incorrect count indicates that the file does not have the required layout.
- Solution: Replace the file with the verified conversion rather than attempting to add or remove tensors manually. Here,
Nrepresents the actual count reported by the node.
SenseNova U1.5 8-step LoRA key has no ComfyUI prefix: NAME
- Explanation: An internal tensor name does not start with the prefix required by this ComfyUI conversion. This indicates an incompatible naming layout.
- Solution: Use the verified ComfyUI-converted file rather than the raw source LoRA or a conversion intended for another loader.
NAMEidentifies the offending tensor.
Unexpected SenseNova U1.5 8-step LoRA tensor: NAME
- Explanation: The file contains an internal tensor name with an unsupported ending. The validator expects the defined alpha, down-weight, and up-weight components.
- Solution: Replace the file with the verified conversion. Do not manually rename internal tensors, because naming is only one of several compatibility checks.
SenseNova U1.5 8-step LoRA tensor must be BF16: NAME
- Explanation: A stored tensor is not in the required BF16 numerical format. A precision-converted or re-saved LoRA may fail this check.
- Solution: Use the original verified BF16 conversion. Changing the base model’s precision will not repair an incompatible LoRA file.
SenseNova U1.5 8-step LoRA module count mismatch: N
- Explanation: After grouping the tensor components, the node found a different number of LoRA-adjusted model modules than expected.
- Solution: Obtain a fresh copy of the verified conversion and confirm that the selected file is the intended eight-step adaptation, not a partial or alternative LoRA.
Incomplete SenseNova U1.5 8-step LoRA module: NAME
- Explanation: A group of LoRA adjustments is missing one or more required components, or does not contain the expected component set.
- Solution: Replace the file with the complete verified conversion. Do not copy components from another LoRA, as they may not match the model or the remaining data.
SenseNova U1.5 8-step LoRA alpha mismatch: NAME
- Explanation: A module’s alpha value, an internal scaling component of the LoRA, is not the required scalar value or has the wrong shape. This is separate from the adjustable
strength_modelsetting. - Solution: Restore the verified conversion. Adjusting
strength_modelcannot fix an invalid internal alpha value.
SenseNova U1.5 8-step LoRA rank mismatch: NAME
- Explanation: A pair of LoRA weight matrices has an unexpected shape or internal rank, so it does not match the supported adaptation structure.
- Solution: Use the exact supported conversion rather than a resized, compressed, retrained, or otherwise modified LoRA. Strength adjustments cannot correct incompatible matrix dimensions.
SenseNova U1.5 8-step LoRA matched N of M model modules
- Explanation: The LoRA passed earlier checks, but its adjustments could not all be matched to the connected model.
Nis the number matched andMis the expected total. - Solution: Confirm that you are using the supported Final checkpoint and verified LoRA together. If both files are correct, check that your ComfyUI installation and SenseNova extension are compatible, and retry with a direct connection from the loader.
SenseNova U1.5 8-step LoRA applied N of M model patches
- Explanation: The model did not accept every expected adjustment after the LoRA components were matched. The node stops instead of returning a partially adapted model.
- Solution: Retry from a freshly loaded supported Final model without other model-modifying nodes before this LoRA. If the problem persists, check ComfyUI and extension compatibility and include the complete error log when reporting the issue.
