
Senior Principal Architect at @VMwareTanzu / @Broadcom | Ex @Pivotal | Bichon Frise 🍋🐩
Tokyo, Japan
Joined April 2007
- Tweets36.6K
- Following201
- Followers3.7K
- Likes6.1K
10年契約すごい
Broadcom、VMware Cloud FoundationでNTTドコモのプライベートクラウド高度化を推進 prtimes.jp/main/html/rd/p/00…
やはりもう一台目追加かな...
GLM5.3-fullをDGX Sparkで
「とにかく大きい KV キャッシュでできるだけ高精度で使いたい」
人向けにレシピを作りました。
ctx:1M / prefill:1520tok/s / decode:53.2tok/s
parallelism:TP=4 / TP=8 / DCP=2
GitHub:
github.com/fiore-labo/fiore-…
model:
huggingface.co/fiore-lab/GLM…
huggingface.co/fiore-lab/GLM…
TP=3版を待ってます!
GLM 5.3 Flash now runs on TensorFold with ZIG 🔥
2x DGX Sparks only for now, 3-4 Sparks coming next.
The work on this started a while ago. Big thanks to @hecisaza for taking this over the finish line and helping out!
Currently it sits on its own branch, separate from main.
Get it here:
github.com/MiaAI-Lab/GLM-5.3…
DGX Spark 3台以内で動いて欲しい
Replying to @TheINAOG
The open weights for Muse Spark are coming soon, Muse Wonder models are large models that are too big for many people to use, in addition to potential safety concerns.
キタ!
GLM 5.3 Flash on 2x DGX Sparks v1.10 is out ⚡️
Lots of PRs merged and bugs squashed!
Here are some highlights:
- Works with Claude Code & Anthropic SDKs directly
- Much faster at very long chats (500k+ tokens)
- Opt-in disk cache: long chats resume in 0.2 s instead of 11 s, even after a restart
- Sampling is now as fast as greedy decoding
Full changelog here:
mia-ai.net/models/GLM-5.3-Fl…
Toshiaki Maki retweeted
GLM 5.3 Flash on 2x DGX Sparks v1.10 is out ⚡️
Lots of PRs merged and bugs squashed!
Here are some highlights:
- Works with Claude Code & Anthropic SDKs directly
- Much faster at very long chats (500k+ tokens)
- Opt-in disk cache: long chats resume in 0.2 s instead of 11 s, even after a restart
- Sampling is now as fast as greedy decoding
Full changelog here:
mia-ai.net/models/GLM-5.3-Fl…
自宅用にself-hostedなngrok/cloudflaredのようなものをSpring gRPCで作っているけど、全部Local LLMにかいてもらっている。これだけ書ければ十分じゃない?
github.com/making/sluice
引用元のポストを見て、「完全同意で、自分はフルテストも数タスクに一回、テストが失敗しても後段に直させる、をやっている」とポストしようとしたら、やはり同じことをされていた。


















