Update README.md

README.md

@@ -1,7 +1,86 @@
 ---
-language: en
+license: other
+license_name: modified-mit
 library_name: mlx
-pipeline_tag: text-generation
 tags:
 - mlx
+- transformers
+pipeline_tag: text-generation
+base_model: moonshotai/Kimi-K2.6
+---
+
+# mlx-community/Kimi-K2.6-mlx-DQ3_K_M-q8
+
+This model [mlx-community/Kimi-K2.6-mlx-DQ3_K_M-q8](https://huggingface.co/mlx-community/Kimi-K2.6-mlx-DQ3_K_M-q8) was converted to MLX format from [moonshotai/Kimi-K2.6](https://huggingface.co/moonshotai/Kimi-K2.6)
+using mlx-lm version **0.31.2**.
+
+After the success of [the first Kimi "DQ3_K_M" model](https://huggingface.co/mlx-community/Kimi-K2-Instruct-0905-mlx-DQ3_K_M) and the K2.5, this is a new update for Kimi-K2.6!
+
+This is created for people using a single Apple Mac Studio M3 Ultra with 512 GB. The 4-bit version of Kimi K2 does not fit. Using research results, we aim to get 4-bit performance from a slightly smaller and smarter quantization. It should also not be so large that it leaves no memory for a useful context window.
+
+You can find more similar MLX model quants for Apple Mac Studio with 512 GB at https://huggingface.co/bibproj
+
+```bash
+pip install mlx-lm
+
+mlx_lm.generate --model mlx-community/Kimi-K2.6-mlx-DQ3_K_M-q8--temp 0.6 --min-p 0.01 --max-tokens 4096 --trust-remote-code --prompt "Hallo"
+```
+
+---
+
+## What is this DQ3_K_M?
+
+In the Arxiv paper [Quantitative Analysis of Performance Drop in DeepSeek Model Quantization](https://arxiv.org/abs/2505.02390) the authors write,
+
+> We further propose `DQ3_K_M`, a dynamic 3-bit quantization method that significantly outperforms traditional `Q3_K_M` variant on various benchmarks, which is also comparable with 4-bit quantization (`Q4_K_M`) approach in most tasks.
+
+and
+
+> dynamic 3-bit quantization method (`DQ3_K_M`) that outperforms the 3-bit quantization implementation in `llama.cpp` and achieves performance comparable to 4-bit quantization across multiple benchmarks.
+
+The resulting multi-bitwidth quantization has been well tested and documented.
+
+---
+
+## How can you create your own DQ3_K_M quants?
+
+The recipe is the same as that for the K2.5 model. Both are a bit different from that of [the first Kimi "DQ3_K_M" model](https://huggingface.co/mlx-community/Kimi-K2-Instruct-0905-mlx-DQ3_K_M), which was described there. To make to the quant perform better under stress, only the expert tensors are quantized to a mix of 3-bit and 4-bit. All the other tensors are kept at 8-bit. You could say that this quant has an 8-bit "brain" and 3-bit/4-bit experts. The sizes of all three these quants are roughly the same. The 8-bit routing does reduce the tokens/second by a few %. You get a slightly slower TG, but better quality results.
+
+In the `convert.py` file of mlx-lm on your system ( [you can see the original code here](https://github.com/ml-explore/mlx-lm/blob/main/mlx_lm/convert.py) ), replace the code inside `def mixed_quant_predicate()` with something like
+
+```python
+        index = (
+            int(path.split(".")[layer_location])
+            if len(path.split(".")) > layer_location
+            else 0
+        )
+        # Build a mixed quant like "DQ3" similar to the "DQ3" of Arxiv paper https://arxiv.org/abs/2505.02390
+        #    Quantitative Analysis of Performance Drop in DeepSeek Model Quantization
+        q_bits = 8
+        if "switch_mlp.up_proj" in path:
+           q_bits = 3
+        if "switch_mlp.gate_proj" in path:
+           q_bits = 3
+        if "switch_mlp.down_proj" in path:
+           q_bits = 3
+           # Layers up to 5 are higher quality
+           if index < 5:
+              q_bits = 5
+           # Every 5th layer is "medium" quality
+           if (index % 5) == 0:
+              q_bits = 4
+        print("path:", path, "index:", index, "q_bits:", q_bits)
+        return {"group_size": group_size, "bits": q_bits, "mode": mode}
+```
+
+Then create your DQ3_K_M quant with
+
+```bash
+mlx_lm.convert --hf-path moonshotai/Kimi-K2.6 --mlx-path your-model-DQ3_K_M -q --quant-predicate mixed_3_4 --trust-remote-code
+```
+
+**NOTE***: With Kimi-K2.5 and Kimi-K2.6 you need to first dequantize the model before you can create the MLX quant. This step  requires just over 2TB of additional disk space.
+
 ---
+
+Enjoy!