File size: 429 Bytes
a844567
 
 
 
 
 
 
 
 
 
 
 
 
 
aa1ed19
3a28e73
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

---
license: mit
language:
- en
---

My own (ZeroWw) quantizations.
output and embed tensors quantized to f16.
all other tensors quantized to q5_k or q6_k.

Result:
both f16.q6 and f16.q5 are smaller than q8_0 standard quantization
and they perform as well as the pure f16.

Note:  
as of now, to run this model you must use: https://github.com/mnlife/llama.cpp  
Later on the PR will be added to the main branch of llama.cpp