DavidAU commited on
Commit
7bb0b56
·
verified ·
1 Parent(s): c7f32e9

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +5 -2
README.md CHANGED
@@ -17,9 +17,10 @@ Mishima Imatrix Quants of new "Qwen 3 - 4B" model with MAX "output tensor" at BF
17
 
18
  Mishima Imatrix dataset was generated using some of the public domain works of author (in English):
19
 
20
- YUKIO MISHiMA (Japanese author and poet, 1925-1970)
21
 
22
- This is an experiment to determine prose changes / changes to the model using a specific, but long and detailed Imatrix dataset.
 
23
 
24
  To test against "Qwen 3 4B" regular, "Horror" and "NEO" versions:
25
 
@@ -57,6 +58,8 @@ Each one is the same dataset, however dataset #1 is "raw" text format, whereas d
57
 
58
  Both affect the model differently.
59
 
 
 
60
  Context Length: 32 K + 8K output generation. (can be extended to 128k)
61
 
62
  <B>NOTE - Jinja Template / Template to Use with this Model:</B>
 
17
 
18
  Mishima Imatrix dataset was generated using some of the public domain works of author (in English):
19
 
20
+ YUKIO MISHiMA (Japanese author and poet, 1925-1970) ; entire work of "SUN AND STEEL" (HIS PERSONAL TESTAMENT ON ART, ACTION, AND RITUAL DEATH)
21
 
22
+ This is an experiment to determine prose changes / changes to the model using a specific, but long and detailed Imatrix dataset
23
+ on the newest Qwen 3 model type.
24
 
25
  To test against "Qwen 3 4B" regular, "Horror" and "NEO" versions:
26
 
 
58
 
59
  Both affect the model differently.
60
 
61
+ Also, each quant has the output tensor at BF16 (16 bit precision) to improve reasoning and output generation.
62
+
63
  Context Length: 32 K + 8K output generation. (can be extended to 128k)
64
 
65
  <B>NOTE - Jinja Template / Template to Use with this Model:</B>