24h GitHub Repo Summary

GitHub Tracked repos · Mar 28, 2026, 10:03 p.m.

24h commit roundup with 15 commits across 4 active repos in the tracked set.

fc1fdf3: Fixes a shader bug so color curves apply correctly on WebGL2 setups that mis-handle nested sampler calls.
b353a7c: Integrates RAM cache behavior with model RAM management to improve memory handling consistency.
3696c5b: Adds a has_intermediate_output flag so interactive UI nodes can expose mid-pipeline outputs cleanly.
3a56201: Allows Flux conditioning flows to run even when pooled output is unavailable.

6a1fc54: Adds a t0 loss target option for finer training objective control.
8302b21: Bumps project version to publish the latest training and UX updates.
20929b9: Fixes EMA Decay input onChange behavior so UI config updates apply reliably.
4ef5cbe: Corrects inverted width/height handling during control image resizing.
700c4b5: Pins timm==1.0.22 to stabilize dependency behavior across environments.
ca72eb1: Adds native 1328 resolution support for Qwen Image training setups.
5ce87fa: Applies a version bump to package recent merged fixes/features.

7186b11: Merges PR #15 to bring in repository agent-governance rules under .moltfounders/.
6eb4378: Adds a .moltfounders/ ruleset to standardize AI-agent triage, review, and maintenance practices.