Text Generation
GGUF
English
reasoning
thinking
uncensored
gated
mixture of experts
expert gate controls
expert named controls
Mixture of Experts
8x3B
Llama 3.2 MOE
NEO Imatrix
128k context
creative
creative writing
fiction writing
plot generation
sub-plot generation
story generation
scene continue
storytelling
fiction story
science fiction
romance
all genres
story
writing
vivid prosing
vivid writing
fiction
roleplaying
float32
swearing
rp
horror
mergekit
llama-3
llama-3.2
imatrix
Update README.md
Browse files
README.md
CHANGED
|
@@ -61,8 +61,7 @@ models into one massive powerhouse at 18.4B parameters (equal to 24B - 8 X 3B).
|
|
| 61 |
|
| 62 |
This model's instruction following, and output generation for creative writing, prose, fiction and role play are exceptional.
|
| 63 |
|
| 64 |
-
This model is also "gated", contains a master reasoning model (this can be turned on/off), was built at float32 (32 bit) precision
|
| 65 |
-
and quants have the output tensor at Q8_0, with a few choice quants at f16 (16 bit) and a Q8_0 with f32 (32 bit).
|
| 66 |
|
| 67 |
These quants are also mastered using the HORROR Imatrix dataset, including new methods to "imatrix" both the output tensor and tokens too.
|
| 68 |
Horror quants have output tensor at Q8_0 up to Q3KL, and then F16 for IQ4/Q4 on up. Horror dataset was created using "Grand Horror 16.5B".
|
|
|
|
| 61 |
|
| 62 |
This model's instruction following, and output generation for creative writing, prose, fiction and role play are exceptional.
|
| 63 |
|
| 64 |
+
This model is also "gated", contains a master reasoning model (this can be turned on/off), was built at float32 (32 bit) precision.
|
|
|
|
| 65 |
|
| 66 |
These quants are also mastered using the HORROR Imatrix dataset, including new methods to "imatrix" both the output tensor and tokens too.
|
| 67 |
Horror quants have output tensor at Q8_0 up to Q3KL, and then F16 for IQ4/Q4 on up. Horror dataset was created using "Grand Horror 16.5B".
|