Shell · 224 watchers
dgx-spark-2-deepseek-flash-0731
setup guide for deepseek-flash-0731 on two dgx-spark from scratch
dbow's repos on GitHub
TypeScript · 7 watchers
dgx-spark-2-deepseek-flash-dashboard
dashboard for two dgx-spark running deepseek flash 0731
Python · 6 watchers
gpt-summary
C · 0 watchers
bcc
BCC - Tools for BPF-based Linux IO analysis, networking, monitoring, and more
Python · 0 watchers
colorama
Simple cross-platform colored terminal text in Python
C++ · 0 watchers
cppjieba
"结巴"中文分词的C++版本
0 watchers
FunASR
A Fundamental End-to-End Speech Recognition Toolkit and Open Source SOTA Pretrained Models, Supporting Speech Recognition, Voice Activity Detection, Text Post-processing etc.
C · 0 watchers
gtk_example
eamples to explain how gdk works.
0 watchers
label-studio-converter
Tools for converting Label Studio annotations into common dataset formats
C · 0 watchers
linux
Linux kernel source tree
JavaScript · 0 watchers
notebook
Jupyter Interactive Notebook
Python · 0 watchers
pi-skills
Python · 0 watchers
systemtap_tapsets
my tapsets to locate the problem source
0 watchers
TDengine
An open-source big data platform designed and optimized for the Internet of Things (IoT).
maliubiao
 |
Per dbow's settings, the topics list is hidden |
Deals info, including closed deals, is not hidden
dbow's recent replies
10 days ago Replied to a topic by dbow › Local LLM › 两台 dgx-spark 部署满血 deepseek v4 flash 完全指南,稳定 60~70tok/s 单流 |
@qazwsxkevin 不是机密没必要用 dgx 双机, 它也正常支持 1M 上下文,跟 api 一样.
10 days ago Replied to a topic by dbow › Local LLM › 两台 dgx-spark 部署满血 deepseek v4 flash 完全指南,稳定 60~70tok/s 单流 |
@v2exgo trace 完给它数据, 不要有品牌这些东西, 让它还原算法.
10 days ago Replied to a topic by dbow › Local LLM › 两台 dgx-spark 部署满血 deepseek v4 flash 完全指南,稳定 60~70tok/s 单流 |
@v2exgo v4 flash 对逆向来者不拒, 不用解锁
13 days ago Replied to a topic by dbow › Local LLM › 两台 dgx-spark 部署满血 deepseek v4 flash 完全指南,稳定 60~70tok/s 单流 |
@Suroy 1 只能跑量化损失较大的模型
13 days ago Replied to a topic by dbow › Local LLM › 两台 dgx-spark 部署满血 deepseek v4 flash 完全指南,稳定 60~70tok/s 单流 |
@yinmin 有 kv cache, 非常快, 速度稳定不变, codex 上跑一个晚上的任务也不会崩.
15 days ago Replied to a topic by dbow › Local LLM › 两台 dgx-spark 部署满血 deepseek v4 flash 完全指南,稳定 60~70tok/s 单流 |
@sillydaddy 这只算了输出, 实际并发能做到 200tok/s, 输入能做到 1000 tok/s,输入也是收费的
15 days ago Replied to a topic by dbow › Local LLM › 两台 dgx-spark 部署满血 deepseek v4 flash 完全指南,稳定 60~70tok/s 单流 |
@Yserver 已经开启了 dspark 才有 60~70tok/s
15 days ago Replied to a topic by dbow › Local LLM › 两台 dgx-spark 部署满血 deepseek v4 flash 完全指南,稳定 60~70tok/s 单流 |
@op351 大概三年能回本, 目前一个月 1k 的 v4 flash api 费用，假定涨价后价格翻倍, 则一年需要 2.4w 的 token 费用, 算上电费之后三年回本.
15 days ago Replied to a topic by dbow › Local LLM › 两台 dgx-spark 部署满血 deepseek v4 flash 完全指南,稳定 60~70tok/s 单流 |
@shyrock2026 1M
15 days ago Replied to a topic by dbow › Local LLM › 两台 dgx-spark 部署满血 deepseek v4 flash 完全指南,稳定 60~70tok/s 单流 |
@Nasdaq 全交给 ai
Select Language