Signal
Advertise
Signal
Advertise
Sign in
Submit
Discover trends that matter
Trending repositories
Daily
Weekly
Monthly
Yearly
Live mentions
Topics
GitHub trending
Repositories
Developers
Insights
Stats
Log in
blazux/qwen3.8-Flash-DGX — GitHub trending stats & insights | Trendshift
Omnigraph
blazux/qwen3.8-Flash-DGX
#
Local LLM
recipe for running Qwen3.8-Flash-Next on a single DGX Spark
Visit GitHub
Like blazux/qwen3.8-Flash-DGX, 0 likes
0
Bookmark blazux/qwen3.8-Flash-DGX, 0 bookmarks
0
Python
311
36
1 contributors
Custom license
Social mentions
Recent discussions about this repository across the web
I run local LLMs with pride. Yesterday, I argued that all LLMs should be open-weight. And I love running open models on my DELL Pro Max with GB10. The same machine can run increasingly intelligent…
@DaigoTanaka · x.com
为什么 176B 模型能跑进单台 DGX Spark?关键不是继续压权重,而是给 KV 缓存抢回内存。 它把 48GiB 的 n-gram PLE 查表改为 NVMe mmap,单个 token 只按需取少量行,权重占用降至约75GiB,同时保留 GB10 原生内核。更难得的是,前缀缓存、确定性 top-k、混合精度和 MTP 都被明确做成质量、速度、上下文之间可审计的工程取舍。
@vintcessun · x.com
The checkpoint occupies ~126 GiB on disk, plus ~19 GiB for the Docker/vLLM image: approximately 145 GiB total. The checkpoint includes a ~47.7 GiB FP8 n-gram/PLE table.
@FreedEngineer · x.com
Repository activities
repository's daily and monthly activities across stars, forks, merged PRs, issues, and closed issues