(NO COMMS) NSFW Blender artist. Aspiring animator. Size, futa and WoW stuff~ Support me at: https://xitter.cf/t.co/KFI0WXMZ0g https://xitter.cf/t.co/6lD8dcSdoM
Up here~
Joined November 2018
- Tweets20.9K
- Following856
- Followers57.5K
- Likes75.2K
Vana3D🔞 retweeted
Premium Breeding Stock (pt.1)~
Featuring @Movin_targ3t's Abigale
4k version at links in bio
Vana3D🔞 retweeted
Premium Breeding Stock (pt.2)~
Featuring @Movin_targ3t's Abigale
4k versions at the usual places
Vana3D🔞 retweeted
Something that I'm actually kinda proud about with my Marina animation becoming weirdly mainstream is that no one says she looks off or weird besides the ass, even though she's clearly not the official model.
All that time re-creating the look in Honey Select really paid off
Vana3D🔞 retweeted
At the end of the day Midnight is basically "Someone played ff14 and saw the Light aspect being bad and shadow aspect being not evil" and decided to try and do that idea Really Badly without understanding why it worked in FF14 like at all
Vana3D🔞 retweeted
💥卧槽！6年前的3090被“破解”了！Tensor算力直接翻倍，官方锁了一半的性能被开源一键释放！
NVIDIA长期把30系显卡的FP16/BF16 Tensor指令发射速率限制在专业卡一半左右。
开源项目 gpu-unlock 用一个Linux内核模块，在驱动加载前改寄存器解锁，不刷BIOS、不改驱动、更新无忧。
牛逼之处：
🔹3090 Ti：BF16 87.5→157 TFLOPS，TF32 42.6→82.6 TFLOPS
🔹3080：FP16/BF16 约65→110–113 TFLOPS
🔹vLLM长上下文prefill快33–36%，首字延迟降25%
🔹结果比特一致，社区已在3070/3080/3090复现
关键步骤（仅Linux）：
1️⃣git clone github.com/Zagitta/gpu-unloc… && cd gpu-unlock && make install
2️⃣内核参数加 gpu_unlock.targets=all，关Secure Boot
3️⃣更新引导并重启
4️⃣journalctl -k | grep gpu_unlock 看到 unlocked 即成功
GitHub：github.com/Zagitta/gpu-unloc…
#RTX3090 #NVIDIA #AI算力 #开源 #GPU解锁 #本地AI #vLLM #黄仁勋
💥不容小觑！英伟达悄悄开源Flare！纯Rust本地推理服务器，专治DGX Spark和Blackwell长对话+Agent！DeepSeek也能直接上！
亮点直接拉满：
🔹运行时纯Rust，零Python开销
🔹内存提前规划 + 默认FP8 KV Cache
🔹CUDA Graph预捕获，启动后极速响应
🔹专为长上下文、工具调用、多轮Agent优化
🔹已内置DeepSeek配方
以DeepSeek V4.1 Flash 1017为例，关键操作步骤：
1️⃣克隆仓库并准备环境（Rust + uv + 依赖）
2️⃣用ignition导出运行时包：
python -m ignition export deepseek-v4.1-flash --gpus 0,1,2,3 --max-concurrency 4 --mtp-draft-tokens 5 --output dsv41.safetensors
3️⃣Rust引擎直接服务：
./flare run dsv41.safetensors --gpus 0,1,2,3 --port 8001
4️⃣（多机）4台Spark TP4部署，用orbit做集群服务，KV池可拉满超长上下文
这波NVIDIA明显在抢本地Agent赛道，和DeepSeek几乎同时出手，有意思。
仓库：github.com/NVIDIA/flare-loca…
#NVIDIA #Flare #DeepSeek #V41Flash #本地推理 #DGXSpark #Blackwell #Agent #Rust #开源 #黄仁勋 #梁文峰
