Could you update this for the new DeepSeek-V4-Flash-0731 checkpoint?
Could you update this for the new DeepSeek-V4-Flash-0731 checkpoint? Would be much appreciated! Thanks!!
Could you update this for the new DeepSeek-V4-Flash-0731 checkpoint? Would be much appreciated! Thanks!!
i'm on it already, but i'm still not 100% sure it's gonna go smooth, they switched from 1 mtp block to 3 dspark blocks (+7.3 GB), i'm preparing everything, then rent a 2x 6000 PRO and try to recreate the cache files (can't do it with 1x)
probably i can just ignore dspark blocks, they won't fit anyway, we were already on the very edge with mtp
Could you update this for the new DeepSeek-V4-Flash-0731 checkpoint? Would be much appreciated! Thanks!!
have a look, i will let it iterate for a few days to see if i can improve quality/speed, and also run more benchs to compare vs laguna (https://huggingface.co/anoane/Laguna-S-2.1-NVFP4-RTX6000PRO-vLLM) and qwen3.6 27b
https://huggingface.co/anoane/DeepSeek-V4-Flash-0731-vllm-moet-sm120-cache