Deepseek v4 flash Q2 on a single 4090 😅
英文标题：Deepseek v4 flash Q2 on a single 4090 😅
It freaking worked lol🔥 Deepseek-v4-flash-0731 @UnslothAI 's IXQ2/Q3 checkpoint on one single RTX4090 with just 64 GB of RAM at usable token rate without dspark. All kernels running on Blaze (my custom developed ML compiler + inference engine) - no llama.cpp or vllm in the picture. The setup k
打开原文相关 2 · 机会 0
中文
It freaking worked lol🔥 Deepseek-v4-flash-0731 @UnslothAI 's IXQ2/Q3 checkpoint on one single RTX4090 with just 64 GB of RAM at usable token rate without dspark. All kernels running on Blaze (my custom developed ML compiler + inference engine) - no llama.cpp or vllm in the picture. The setup k
中文全文翻译正在补齐，可先查看下方英文原文。
English / 原文
It freaking worked lol🔥 Deepseek-v4-flash-0731 @UnslothAI 's IXQ2/Q3 checkpoint on one single RTX4090 with just 64 GB of RAM at usable token rate without dspark. All kernels running on Blaze (my custom developed ML compiler + inference engine) - no llama.cpp or vllm in the picture. The setup k