TokForge-SD15-LCM-MNN-b1

Published by TokForge for on-device inference.

TokForge

• Website: https://tokforge.ai
• Discord: https://discord.gg/Acv3CBtfVm
• Google Play: https://play.google.com/store/apps/details?id=dev.tokforge
• iOS TestFlight: https://testflight.apple.com/join/jnufjzRr

Runs on-device in the TokForge app.

TokForge SD1.5-LCM MNN: batch-1 (CFG-skip) build

Batch-1-capable MNN export of the DreamShaper-7 + LCM SD1.5 bundle for the TokForge Android app. The UNet is traced at batch-1 with dynamic spatial axes (sample=[1,4,-1,-1]) so the app's CFG-skip engine runs a single conditional UNet pass at cfg=1.0 (LCM is guidance-free), one UNet pass per step instead of two in the legacy batch-2 build.

This is the versioned -b1 companion to darkmaniac7/TokForge-SD15-LCM-MNN. It requires the TokForge batch-1 CFG-skip engine (app vc421+). Older app builds (vc412 and earlier) must keep using the non--b1 repo, because the old engine would feed batch-2 inputs to a batch-1 UNet and crash. CLIP / VAE / tokenizer are byte-identical to the base repo; only unet.mnn(.weight) differs.

Credits

• DreamShaper 7 by Lykon (Civitai, Hugging Face). Used under the CreativeML OpenRAIL-M license and the creator's Civitai terms. Lykon's Civitai terms: no selling the model or merges of it.
• LCM-LoRA SD1.5 by Latent Consistency (Hugging Face). Used under the CreativeML OpenRAIL++-M license.
• Stable Diffusion v1.5 by Runway and CompVis (Hugging Face). Used under the CreativeML OpenRAIL-M license.
• Converted to MNN with MNN by Alibaba (Apache-2.0).