update readme

README.md

@@ -97,7 +97,7 @@ numactl -N 0 -m 0 \
 
 These are probably the **best quants available in this size class** for `V3-0324`!
 
-[!][Benchmarks showing these quants are smaller in size yet similar in performance to the `Q8_0`](benchmarks-01.png "Benchmarks showing these quants are smaller in size yet similar in performance to the `Q8_0`")
+![Benchmarks showing these quants are smaller in size yet similar in performance to the `Q8_0`](benchmarks-01.png "Benchmarks showing these quants are smaller in size yet similar in performance to the `Q8_0`")
 
 ubergarm made no sacrifices for token embedding, attention, dense
 layers, or shared experts. This is possible because `ik_llama.cpp` MLA
@@ -220,6 +220,10 @@ Final estimate: PPL = 3.4755 +/- 0.03305
 
 #### Quant Cookers Secret Recipe
 
+<details>
+
+<summary>Secret Recipe</summary>
+
 ```bash
 #!/usr/bin/env bash
 
@@ -284,8 +288,14 @@ custom=$(
     24
 ```
 
+</details>
+
 #### Perplexity
 
+<details>
+
+<summary>Perplexity Logs</summary>
+
 ```bash
 $ CUDA_VISIBLE_DEVICES="0," \
 ./build/bin/llama-perplexity \
@@ -701,8 +711,16 @@ llama_print_timings:       total time = 2841519.57 ms / 287233 tokens
 Final estimate: PPL = 3.5614 +/- 0.02001
 ```
 
+</details>
+
 #### Split
 
+<details>
+
+<summary>Split GGUF</summary>
+
+*TODO*: Add key value metadata information before publishing.
+
 ```bash
 $ ./build/bin/llama-gguf-split \
     --dry-run \