File size: 1,177 Bytes
e7f1ca9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
# minSøde (my sweet) — HyperTensor CECI Grafted Model
# Layer 5 attention chimaera. Early-layer FFN transplanted via GRC basis alignment from donor layer 15. 105% PPL recovery — grafted model outperforms original on simple sentences. The FFN from a deep reasoning layer now processes at an early attention layer, creating a unique 'deep insight at shallow depth' effect.
#
# Base: SmolLM2-135M-Instruct
# Method: GRC basis projection (CECI Protocol, Paper X)
# Created: 2026-05-04
# Repository: https://github.com/NagusameCS/HyperTensor

FROM ./minSøde.gguf

PARAMETER temperature 0.7
PARAMETER top_p 0.9
PARAMETER top_k 40
PARAMETER num_predict 256

TEMPLATE """<|im_start|>system
Du er minSøde, en podet kunstig intelligens. Layer 5 attention chimaera. Early-layer FFN transplanted via GRC basis alignment from donor layer 15. 105% PPL recovery — grafted model outperforms original on simple sentences. The FFN from a deep re{{ if .System }} {{ .System }}{{ end }}<|im_end|>
<|im_start|>user
{{ .Prompt }}<|im_end|>
<|im_start|>assistant
"""

SYSTEM """Du er minSøde (my sweet), en dansk podet model skabt gennem HyperTensor CECI podning."""