Update README.md
Browse files
README.md
CHANGED
|
@@ -24,10 +24,13 @@ This is an experimental quant with mixed precision selective layers, some compre
|
|
| 24 |
|
| 25 |
Side effects of quanting with the qx86-hi formula
|
| 26 |
```bash
|
| 27 |
-
I needed Haskell code.
|
|
|
|
| 28 |
The q6 starts with Haskell, 10k tokens down the road writes Python, and finishes with React at 30k tokens
|
| 29 |
The q6-hi, encoded with group size 32, writes some Haskell, and stops around 24k tokens
|
| 30 |
The qx86-hi worked for 43k tokens, reasoning around the Haskell solution without skipping a beat
|
|
|
|
|
|
|
| 31 |
```
|
| 32 |
|
| 33 |
From the original model card:
|
|
|
|
| 24 |
|
| 25 |
Side effects of quanting with the qx86-hi formula
|
| 26 |
```bash
|
| 27 |
+
I needed Haskell code.
|
| 28 |
+
|
| 29 |
The q6 starts with Haskell, 10k tokens down the road writes Python, and finishes with React at 30k tokens
|
| 30 |
The q6-hi, encoded with group size 32, writes some Haskell, and stops around 24k tokens
|
| 31 |
The qx86-hi worked for 43k tokens, reasoning around the Haskell solution without skipping a beat
|
| 32 |
+
|
| 33 |
+
It didn't forget Python, it's just a bit more open-minded about other languages
|
| 34 |
```
|
| 35 |
|
| 36 |
From the original model card:
|