Create README.md

README.md

@@ -0,0 +1,15 @@
+---
+base_model:
+- deepseek-ai/DeepSeek-R1-0528
+---
+
+This is a custom quant of [deepseek-ai/DeepSeek-R1-0528](https://huggingface.co/deepseek-ai/DeepSeek-R1-0528) that has the following:
+- Q8_0 for the default quantization type (attention, shared experts, etc.)
+- Q4_K for the FFN_UP and FFN_GATE tensors
+- Q5_K for the FFN_DOWN tensors
+
+The idea being that given the huge size of the FFN tensors compared to the rest of the tensors in the model, it should be possible to achieve a better quality while keeping the overall size of the entire model smaller compared to a similar naive quantization.
+
+This is derived from the quant provided by [AesSedai](https://huggingface.co/AesSedai/DeepSeek-R1-0528-GGUF) with additional keys added for expert group selection.
+
+Model is additionally split with `--no-tensor-first-split` to enable easier editing of metadata.