superhack's recent timeline updates superhack's recent timeline updates | superhackV2EX member #4949, joined on 2011-01-08 19:10:20 +08:00Today's activity rank 15271 |
| 求一份 sing-box binary 版客户端在 macOS 上的可用配置 程序员 superhack Sep 25, 2025 Lastly replied by zaihuilvcha | 22 |
| 完全没必要折腾各种 web clipper，最简组合 singlefile + alist 能覆盖 90% 使用场景 分享发现 superhack Jul 21, 2023 Lastly replied by SenLief | 1 |
| 虚拟信用卡的运行原理是什么？ 问与答 superhack May 8, 2023 Lastly replied by 8355 | 3 |
| 有全功能 typec 接口的盒子或者小设备有哪些 问与答 superhack Oct 17, 2022 Lastly replied by wangshushu | 11 |
| 有做过用 developer ID 证书发布带 network extension 的 macOS 应用的吗，达到实际 release 的 iDev superhack Mar 3, 2022 Lastly replied by superhack | 2 |
| 求一个使用了 network extension 的 macOS 开源项目 macOS superhack Aug 1, 2022 Lastly replied by superhack | 6 |
| 自建 tailscale 和 derp 服务器，有部署在大厂云环境的吗？ 宽带症候群 superhack Jan 18, 2022 Lastly replied by yangchuansheng33 | 4 |
| 油管 Premium 印区车位，要求印区 id，还剩 4 位 无要点 superhack Mar 11, 2020 |
| rmbp 15 配个 惠科 t7000 pro 这个方案怎么样？ MacBook Pro superhack Nov 11, 2014 Lastly replied by Showfom | 24 |
| 求OpenPhoto邀请码，谢谢！ 分享邀请码 superhack Dec 7, 2011 Lastly replied by superhack | 6 |
superhack's recent replies
Apr 29 Replied to a topic by zsj1029 Local LLM qwen3.6 27b 本地编码测试 |
@zsj1029 是投机推理
Apr 29 Replied to a topic by zsj1029 Local LLM qwen3.6 27b 本地编码测试 |
@zsj1029
```
services:
vllm-qwen36-27b:
image: vllm/vllm-openai:v0.20.0
container_name: vllm-qwen36-27b
restart: "unless-stopped"
ports:
- "8085:8000"
volumes:
- models/huggingface/qwen3.6-27b-autoround-int4:/model
environment:
- CUDA_VISIBLE_DEVICES=0
- VLLM_WORKER_MULTIPROC_METHOD=spawn
- PYTORCH_CUDA_ALLOC_COnF=expandable_segments:True
shm_size: "16gb"
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]
command:
- --model
- /model
- --served-model-name
- qwen3.6-27b
- --quantization
- auto-round
- --dtype
- float16
- --gpu-memory-utilization
- "0.92"
- --max-model-len
- "160000"
- --kv-cache-dtype
- fp8_e5m2
- --max-num-seqs
- "1"
- --trust-remote-code
- --reasoning-parser
- qwen3
- --enable-auto-tool-choice
- --tool-call-parser
- qwen3_xml
- --enable-prefix-caching
- --speculative-config
- '{"method":"mtp","num_speculative_tokens":2}'
- --host
- "0.0.0.0"
- --port
- "8000"
```
```
services:
vllm-qwen36-27b:
image: vllm/vllm-openai:v0.20.0
container_name: vllm-qwen36-27b
restart: "unless-stopped"
ports:
- "8085:8000"
volumes:
- models/huggingface/qwen3.6-27b-autoround-int4:/model
environment:
- CUDA_VISIBLE_DEVICES=0
- VLLM_WORKER_MULTIPROC_METHOD=spawn
- PYTORCH_CUDA_ALLOC_COnF=expandable_segments:True
shm_size: "16gb"
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: all
capabilities: [gpu]
command:
- --model
- /model
- --served-model-name
- qwen3.6-27b
- --quantization
- auto-round
- --dtype
- float16
- --gpu-memory-utilization
- "0.92"
- --max-model-len
- "160000"
- --kv-cache-dtype
- fp8_e5m2
- --max-num-seqs
- "1"
- --trust-remote-code
- --reasoning-parser
- qwen3
- --enable-auto-tool-choice
- --tool-call-parser
- qwen3_xml
- --enable-prefix-caching
- --speculative-config
- '{"method":"mtp","num_speculative_tokens":2}'
- --host
- "0.0.0.0"
- --port
- "8000"
```
Apr 28 Replied to a topic by zsj1029 Local LLM qwen3.6 27b 本地编码测试 |
5090 27B ，vllm 峰值 120 tok/s, 平均 85 tok/s 上下
Mar 27 Replied to a topic by shineonme 程序员 有什么写入全局 AGENTS.md 文件的规则分享吗？ |
如无必要，勿增实体
Mar 21 Replied to a topic by Nerv 分享创造 用 Claude Skill 解决 sing-box 配置难题 |
很不错，代理客户端的 skill ，你好像是第一个
Mar 20 Replied to a topic by Frank570 北京 出租酒仙桥-恒通园区车位 |
不写价格，也得写点别的信息啊
Oct 16, 2025 Replied to a topic bychaoyebugao DevOps 内部做一个新的环境，在更新生产环境前先把生产服务等先在这个服务上过一遍，确认没问题后再上生产，这个环境叫什么环境？这种流程叫什么流程？ |
staging
Oct 14, 2025 Replied to a topic by wi11iamZ 宽带症候群 外出时家宽回连方案讨论 |
wireguard 已被运营商搞残
Oct 13, 2025 Replied to a topic by dark495 宽带症候群 [后续 2] 泉州联通案件的近期情况，以及 10 月 13 日（15:00）开庭｜地点：泉州鲤城法院 |
支持！
Oct 11, 2025 Replied to a topic by zhjh0521 Apple macos 上的小火箭场景功能不工作 |
真想告诉作者，system_profiler 这个命令可以获取到 ssid ，就是解析比较麻烦
Select Language