Andrew Zhu

on Personal AI | open source contributor | startup founder member | ex-Microsoft
blog:

[Andrew Zhu @xhinker]

Today, book "Using Stable Diffusion with Python" is published. Hope it is useful for SD developers.
#stablediffusion #diffusion #ai

https://t.co/xMXUdv1B0k

[Andrew Zhu @xhinker]

@LottoLabs how is the quality compare with Qwen3.6-27B?

[Andrew Zhu @xhinker]

@TheAhmadOsman why not RTX Pro 6000?

[Andrew Zhu @xhinker]

This llama.cpp feature makes you run ONE LLM model across different machines

https://t.co/0XsO6wOimX

[Andrew Zhu @xhinker]

So far, the only local LLM model worth your time is Qwen3.6-27B_K_Q6

[Andrew Zhu @xhinker]

Running Qwen3.6–27B on Dual RTX 3090s

https://t.co/Ixg3hfyMBA

[Andrew Zhu @xhinker]

@LottoLabs Anyone successful with it? it can't run Qwen3.6-27B

[Andrew Zhu @xhinker]

And showing OpenClaw with Windows? how bad idea it can be. #MS, you should build your own agent, if you can't, I can help you build one

[Andrew Zhu @xhinker]

Watched MS build events, and I know why MS stock hit so much today, MS really don't know how to wow the world anymore , just showing small apps here and there are no longer working, those feature improvements are great, great for 3 or 4 years ago, Now? now people expect more than this, people want something big

[Andrew Zhu @xhinker]

What is your Aphrodite Inference version? the release version or self-build? I got error:
pydantic_core._pydantic_core.ValidationError: 1 validation error for ModelConfig
Value error, Model architectures ['Qwen3_5ForConditionalGeneration'] failed to be inspected. Please check the logs for more details. [type=value_error, input_value=ArgsKwargs((), {'model': ...nderer_num_workers': 1}), input_type=ArgsKwargs]
For further information visit https://t.co/2K5Xz2M35t

[Andrew Zhu @xhinker]

The one silly and arrogant PM who decided to take the vertical task bar away from Windows 11 cost M$ lost a Trillion, now M$ have to add it back, why did it years ago? So, there are lots of people with good looking titles, they are the same

[Andrew Zhu @xhinker]

@LottoLabs Me too!

[Andrew Zhu @xhinker]

@XAMTO_AI wow 这太不可思议了

[Andrew Zhu @xhinker]

https://t.co/HWJPiv0pB0

[Andrew Zhu @xhinker]

https://t.co/ac6OX3oVnk

[Andrew Zhu @xhinker]

@immortal_00994 只能说老黄的确比你懂

[Andrew Zhu @xhinker]

NVFP4 with llama.cpp is not faster at all, the truth is it is even slower than the normal Q4-K-M gguf version

[Andrew Zhu @xhinker]

@LottoLabs What makes it "soon"?

[Andrew Zhu @xhinker]

@LottoLabs Understand how 27b works, and build true open source version

[Andrew Zhu @xhinker]

@LottoLabs Sad sad

[Andrew Zhu @xhinker]

@jianshuo 别忘了KV cache, 这已经很节能了