You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Description:
While using the WD14Tagger node, I encountered a severe issue: even when a workflow branch does not require any downstream output, the presence of an image node connected to the WD14Tagger forces the entire branch to execute. This situation arises in the following scenario:
A branch passes through the WD14Tagger node without forwarding its result to any subsequent nodes or final output.
Despite no actual need for the WD14Tagger output downstream, the structure and connection of the workflow trigger the execution of all nodes in that branch.
Impact:
Since the WD14Tagger node involves loading models and performing inference, its unnecessary execution wastes significant computational resources. This overhead becomes particularly impactful in large or complex workflows. Given that WD14Tagger is a commonly used and computationally intensive node, this issue becomes even more pressing.
Expected Behavior:
The system should optimize node execution logic and workflow scheduling to avoid the following:
When a branch does not require any actual output or there are no downstream dependencies on a node's output, the system should intelligently skip unnecessary computations, thus reducing resource consumption.
Specifically for the WD14Tagger node, if there is no real downstream dependency, the inference process should be bypassed, or there should be a mechanism allowing users to disable the node's computation in such cases.
Additional Notes:
This issue is not about UI display, but rather concerns the node execution mechanism and workflow scheduling logic.
It is hoped that developers can investigate the workflow scheduling logic to determine how to avoid executing unnecessary nodes—especially high-computation—when no downstream output is actually required.
Thank you for your time and contribution to this project. I look forward to potential improvements or solutions regarding this matter.
The text was updated successfully, but these errors were encountered:
Description:
While using the WD14Tagger node, I encountered a severe issue: even when a workflow branch does not require any downstream output, the presence of an image node connected to the WD14Tagger forces the entire branch to execute. This situation arises in the following scenario:
Impact:
Since the WD14Tagger node involves loading models and performing inference, its unnecessary execution wastes significant computational resources. This overhead becomes particularly impactful in large or complex workflows. Given that WD14Tagger is a commonly used and computationally intensive node, this issue becomes even more pressing.
Expected Behavior:
The system should optimize node execution logic and workflow scheduling to avoid the following:
Additional Notes:
Thank you for your time and contribution to this project. I look forward to potential improvements or solutions regarding this matter.
The text was updated successfully, but these errors were encountered: