Git commit
Latest version
Operating System & Version
Termux native no Ubuntu installed
GGML backends
CPU
Command-line arguments used
./bin/sd-cli --diffusion-model ~/stable-diffusion.cpp/build/models/diffusion_pytorch_model.safetensors --t5xxl ~/stable-diffusion.cpp/build/models/flan_t5_xxl_TE-only_FP8.safetensors -H 256 -W 256 -p "'The scene explodes in a hyper-detailed IMAX still, a breathtaking collision of chrome, neon, and impossible architecture set against the perpetually misty, towering backdrop of a futuristic San Francisco. The air is thick with the scent of ozone and synthetic fuel, illuminated by the chaotic, saturated glow of holographic advertisements that cascade down the impossibly slender, vertical structures of the city. In the foreground, a sleek, matte-black NASCAR-style vehicle, its chassis interwoven with glowing cyan circuitry and sharp, angular plating, tears through the dense, rain-slicked thoroughfare. The driver is a cybernetic entity, their face a mask of polished obsidian and integrated optical sensors that flicker with an intense, electric blue light, their movements fluid and impossibly precise, a testament to augmented reality and raw, kinetic power. The environment is a symphony of light and shadow: deep, saturated shadows are violently contrasted by the blinding, prismatic reflections bouncing off the wet pavement and the towering, impossibly complex bridges that span the bay, their structural lines sharp and impossibly elegant. Shader effects are palpable, showing the metallic sheen of the car juxtaposed against the soft, diffused glow of the ambient neon—a vibrant, almost painful magenta and electric lime that bleeds into the deep, atmospheric blues of the fog. Every surface, from the pitted texture of the road to the intricate wiring of the vehicle, is rendered with microscopic fidelity, creating a sensory overload where the speed of the race is captured not just by motion, but by the very vibration of the light and the structural tension of the scene, a perfect, high-octane tableau of technological sublime." --cfg-scale 3.3 --sampling-method euler -v --threads 8 --rng cpu --seed 0
Steps to reproduce
Put in cli minit2i commands
What you expected to happen
Start steps and generate image
What actually happened
I'm trying sd3 medium same problem it worked on first install and minit2i but I try reinstall and clear termux but same error.
Logs / error messages / stack trace
[INFO ] stable-diffusion.cpp:5321 - generating image: 1/1 - seed 42 /data/data/com.termux/files/home/stable-diffusion.cpp/ggml/src/ggml.c:3243: GGML_ASSERT(ggml_can_mul_mat(a, b)) failed 0: 0x630d741ff0 1: 0x630d741fb0 2: 0x630d755858 3: 0x630d747d3c 4: 0x630d4c2720 5: 0x630d4fe904 6: 0x630d5b6540 7: 0x630d5b5dbc 8: 0x630d42c9a4 9: 0x630d429db4 10: 0x630d429954 11: 0x630d5b5b40 12: 0x630d658284 13: 0x630d6546bc 14: 0x630d4d4a48 15: 0x630d4eacf4 16: 0x630d4b4394 17: 0x630d3e7ee4 18: 0x70950fc148 __libc_init Aborted
Additional context / environment details
Dimensity 7025 16gb ram Outikel wp55
Git commit
Latest version
Operating System & Version
Termux native no Ubuntu installed
GGML backends
CPU
Command-line arguments used
./bin/sd-cli --diffusion-model ~/stable-diffusion.cpp/build/models/diffusion_pytorch_model.safetensors --t5xxl ~/stable-diffusion.cpp/build/models/flan_t5_xxl_TE-only_FP8.safetensors -H 256 -W 256 -p "'The scene explodes in a hyper-detailed IMAX still, a breathtaking collision of chrome, neon, and impossible architecture set against the perpetually misty, towering backdrop of a futuristic San Francisco. The air is thick with the scent of ozone and synthetic fuel, illuminated by the chaotic, saturated glow of holographic advertisements that cascade down the impossibly slender, vertical structures of the city. In the foreground, a sleek, matte-black NASCAR-style vehicle, its chassis interwoven with glowing cyan circuitry and sharp, angular plating, tears through the dense, rain-slicked thoroughfare. The driver is a cybernetic entity, their face a mask of polished obsidian and integrated optical sensors that flicker with an intense, electric blue light, their movements fluid and impossibly precise, a testament to augmented reality and raw, kinetic power. The environment is a symphony of light and shadow: deep, saturated shadows are violently contrasted by the blinding, prismatic reflections bouncing off the wet pavement and the towering, impossibly complex bridges that span the bay, their structural lines sharp and impossibly elegant. Shader effects are palpable, showing the metallic sheen of the car juxtaposed against the soft, diffused glow of the ambient neon—a vibrant, almost painful magenta and electric lime that bleeds into the deep, atmospheric blues of the fog. Every surface, from the pitted texture of the road to the intricate wiring of the vehicle, is rendered with microscopic fidelity, creating a sensory overload where the speed of the race is captured not just by motion, but by the very vibration of the light and the structural tension of the scene, a perfect, high-octane tableau of technological sublime." --cfg-scale 3.3 --sampling-method euler -v --threads 8 --rng cpu --seed 0
Steps to reproduce
Put in cli minit2i commands
What you expected to happen
Start steps and generate image
What actually happened
I'm trying sd3 medium same problem it worked on first install and minit2i but I try reinstall and clear termux but same error.
Logs / error messages / stack trace
[INFO ] stable-diffusion.cpp:5321 - generating image: 1/1 - seed 42 /data/data/com.termux/files/home/stable-diffusion.cpp/ggml/src/ggml.c:3243: GGML_ASSERT(ggml_can_mul_mat(a, b)) failed 0: 0x630d741ff0 1: 0x630d741fb0 2: 0x630d755858 3: 0x630d747d3c 4: 0x630d4c2720 5: 0x630d4fe904 6: 0x630d5b6540 7: 0x630d5b5dbc 8: 0x630d42c9a4 9: 0x630d429db4 10: 0x630d429954 11: 0x630d5b5b40 12: 0x630d658284 13: 0x630d6546bc 14: 0x630d4d4a48 15: 0x630d4eacf4 16: 0x630d4b4394 17: 0x630d3e7ee4 18: 0x70950fc148 __libc_init Aborted
Additional context / environment details
Dimensity 7025 16gb ram Outikel wp55