ComfyUI Node: KSampler (SPD LoRA / auto-schedule)

KSampler (SPD LoRA / auto-schedule):

The SpectrumSPDLoRAKSampler node is designed to streamline the process of using SPD-trained LoRAs (Low-Rank Adaptations) by automatically configuring the SPEED (SPD + Spectrum) sampler. This node reads the resolution schedule embedded in the metadata of SPD-trained LoRAs, allowing it to align the inference geometry with the training setup without requiring manual tuning of scale or sigma values. This ensures that the model's performance is optimized and consistent with the conditions it was trained under. The node is particularly useful for AI artists who want to leverage the benefits of SPD-trained LoRAs without delving into the complexities of manual configuration. By automatically applying the appropriate settings, the node simplifies the workflow and enhances the quality of the generated outputs.

KSampler (SPD LoRA / auto-schedule) Input Parameters:

lora_name

The lora_name parameter specifies the name of the SPD-trained LoRA to be loaded. This parameter is crucial as it determines which LoRA's resolution schedule will be applied to the model. The node reads the ss_spd_stages and ss_spd_transition_sigmas metadata from the specified LoRA file to configure the sampler automatically. This ensures that the inference process aligns with the training conditions of the LoRA, providing optimal results.

lora_strength

The lora_strength parameter is a floating-point value that acts as a weight multiplier for the LoRA applied to the model. It allows you to adjust the influence of the LoRA on the model's output. The default value is 1.0, with a range from -10.0 to 10.0, allowing for both amplification and attenuation of the LoRA's effect. This flexibility enables fine-tuning of the model's behavior to achieve the desired artistic effect.

adaptive_smc_alpha

The adaptive_smc_alpha parameter is used to control the adaptive SMC (Sequential Monte Carlo) alpha value. This parameter influences the sampling process, potentially affecting the diversity and quality of the generated outputs. The default value is set to _SMC_CFG_ALPHA_DEFAULT, which is a predefined constant. Adjusting this parameter can help in achieving different artistic styles or effects.

KSampler (SPD LoRA / auto-schedule) Output Parameters:

LATENT

The LATENT output parameter represents the latent image generated by the node after applying the SPD-trained LoRA and running the SPEED sampler. This output is crucial as it forms the basis for further processing or rendering into a final image. The latent image encapsulates the model's understanding and interpretation of the input parameters, influenced by the LoRA and the resolution schedule.

KSampler (SPD LoRA / auto-schedule) Usage Tips:

• Ensure that the LoRA you select has the appropriate ss_spd_stages and ss_spd_transition_sigmas metadata to fully utilize the node's automatic configuration capabilities.
• Experiment with the lora_strength parameter to find the right balance for your artistic needs, as it can significantly alter the model's output.
• Use the default adaptive_smc_alpha value initially, and adjust it only if you are seeking specific stylistic changes or effects.

KSampler (SPD LoRA / auto-schedule) Common Errors and Solutions:

SPD LoRA '<lora_name>' has no ss_spd_stages metadata; falling back to the validated 0.50 / σ0.70 single handoff.

• Explanation: This error occurs when the selected LoRA does not contain the necessary ss_spd_stages metadata for automatic configuration.
• Solution: Verify that the LoRA file is correctly formatted and contains the required metadata. If not, consider using a different LoRA or manually configuring the sampler settings.

SPEED/SPD re-spaces σ mid-loop and is Euler-only; ignoring requested sampler '<sampler_name>' and using Euler.

• Explanation: The SPEED sampler is designed to work with the Euler method, and this warning indicates that a different sampler was requested but will be ignored.
• Solution: Ensure that you are using the Euler sampler when working with the SPEED configuration to avoid this warning and ensure optimal performance.