Downloads · 30 days
0
lcm-ssd1b-npu-8elite
latentdivergence/lcm-ssd1b-npu-8elite
lcm-ssd1b-npu-8elite is a text-to-image model from latentdivergence. Use it when you need an image from a text prompt. The card lists the license as apache-2.0.
LCM-SSD-1B compiled for the Qualcomm Hexagon v79 NPU (Snapdragon 8 Elite), as used by Latent Studio. UNet and VAE decoder only - the text encoders and tokenizer are shared with sdxl-lightning-4step-int8.
Access
Public
Updated Sep 24, 2026
Repo size
2.9 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.bin2.9 GB · 100%
At a glance
- Task
- Text-to-Image
- License
- apache-2.0
- Access
- Public
- Created
- Sep 24, 2026
- Updated
- Sep 24, 2026
- SHA
- 41a255a9
Base models
Model card
From the Hugging Face model README
lcm-ssd1b-npu-8elite
LCM-SSD-1B compiled for the Qualcomm Hexagon v79 NPU (Snapdragon 8 Elite), as used by Latent Studio. UNet and VAE decoder only - the text encoders and tokenizer are shared with sdxl-lightning-4step-int8.
Measured on a Snapdragon 8 Elite: 733 ms per UNet step, 1.53 s VAE decode, ~27 s end to end for a 1024px image (versus 5 m 40 s for SDXL-Lightning on CPU).
unet/- 6 inputs:sample [1,4,128,128],timestep [1],encoder_hidden_states [1,77,2048],time_ids [1,6],text_embeds [1,1280],timestep_cond [1,256](LCM guidance embedding). Outputoutput_0.vae_decoder/- built from madebyollin/sdxl-vae-fp16-fix. The stock SDXL VAE overflows in fp16 (two convs peak at 640,538 and 106,287 against fp16's 65,504 ceiling) and renders pure black; the fix weights peak at 2,301.
Both are EPContext wrappers: model.onnx names ./model.bin beside it. The
architecture is fixed - a v79 binary will not load on v75.
- Task
- Text-to-Image
- License
- apache-2.0
- Created
- Sep 24, 2026
- Updated
- Sep 24, 2026