nihui

春雨为观，香月为邻 | 喜欢玩游戏时看风景 | 想成为优秀的程序员 | 看上去是女装大佬实际并不是（ | 上海老阿姨 | qwqwqwq

[nihui @nihui]

ncnn telegram chat group 技术交流电报组
ncnn community
https://t.co/msawZ6S1da

[nihui @nihui]

qwenimage-ncnn-vulkan 已支持 qwen-image-2.1-turbo
8步出图

https://t.co/Egfq4F3X6m

[nihui @nihui]

spacemit k3 pico itx

[nihui @nihui]

@Kunana_XP N卡cuda还是更快，vulkan主要是兼容性好。后面vulkan还能再优化优化的

[nihui @nihui]

用 ncnn 跑 Qwen-Image-2.1：完整 BF16、2GB ��存和全平台 Vulkan
https://t.co/lPLdyaiC4i
https://t.co/RtbbDs72jR

[nihui @nihui]

@seesunlight 我想的是 8g m1也能玩起来

[nihui @nihui]

预告：4G内存 + 2G显存 跑BF16精度 qwen-image-2.1 1024x1024 出图并且速度不变慢

[nihui @nihui]

用 ncnn 跑 Qwen-Image-2.1：完整 BF16、2GB 显存和全平台 Vulkan
https://t.co/lPLdyaiC4i
https://t.co/RtbbDs72jR

[并非日落果 @purofle]

Qwen-Image-2.1 running on Intel B580!

Thanks to @nihui for qwenimage-ncnn-vulkan

[nihui @nihui]

@izutorishima tested on amd rx9060xt 😇

[nihui @nihui]

其实没有我你分不出那些差别
结局还能多明显
别说nihui难过
别说你想改变
AI的人不用道歉😐

[nihui @nihui]

qwenimage-ncnn-vulkan added support for Qwen-Image-2.1 4-step LoRA and ControlNet

It runs faster than pytorch-rocm or sd.cpp on my RX9060XT machine

Hope my tool proves helpful to you :)

https://t.co/RtbbDs72jR

[Lei Zhang @LeiLMx]

Visualizing to learn model internals in the agentic era can be surprisingly low-friction.

I had Codex one-shot architecture notes earlier for several recent OSS models—DeepSeek V4/V4.1, GLM 5.3, and Kimi K3—specifically from a kernel-development perspective: tensor contracts, exact shapes/layouts, attention variants, KV-cache representation, precision choices, dataflow, and operator boundaries.

Sharing in case they’re useful to others:
https://t.co/HViLxaaJdj

[要优雅不要污的零 @ling0322]

好耶，libwaifu支持Qwen-Image-2.1成功🎉
https://t.co/RqquRszGIN

[nihui @nihui]

@XShude 和gpu和分辨率无关，是prefix cache机制节约了输入图过dit loop

[nihui @nihui]

文生图性能优化 10%~20% (cooperative matrix)
图编辑性能优化 40%~85% (prefix cache)
调整低显存的 tile ���略，KV cache host swap，进一步降低显存占用，防止多图输入的显存溢出

https://t.co/yiDGnSwemZ

[nihui @nihui]

@uezochan 2G vram if you try qwenimage-ncnn-vulkan

[nihui @nihui]

预告：controlnet 和 lora safetensor 😋

[nihui @nihui]

@fujiyoshiten 更新了下，可以再试试看

[nihui @nihui]

Qwen-Image-2.1 full BF16 running with 2GB VRAM
ncnn/Vulkan for NVIDIA/AMD/Intel/Apple

No CUDA/PyTorch/Python
Portable executable
Text-to-image & Image editing
Up to 10 reference images
Transparent RGBA generation
Dynamic output resolution
Batch generation
https://t.co/mTiok6Qija

[nihui @nihui]

@fujiyoshiten 在修了😂

[nihui @nihui]

洗衣机里有 LVGL

[nihui @nihui]

@ditangongyuan qwenimage-ncnn-vulkan 试试看👀