File size: 1,241 Bytes
281fa5a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
models:
  - model: tokyotech-llm/Swallow-70b-NVE-instruct-hf
    # no parameters necessary for base model
  - model: GOAT-AI/GOAT-70B-Storytelling # storytelling
    parameters:
      density: 1
      weight: 0.25
  - model: dreamgen/opus-v0.5-70b # creative roleplay
    parameters:
      density: 1
      weight: 0.25
merge_method: dare_ties
base_model: tokyotech-llm/Swallow-70b-NVE-instruct-hf
dtype: bfloat16
name: Swallow-70b-NVE-RP-base
---
models:
  - model: tokyotech-llm/Swallow-70b-NVE-instruct-hf
    # no parameters necessary for base model
  - model: Doctor-Shotgun/lzlv-limarpv3-l2-70b # roleplay configuration
    parameters:
      density: 1
      weight: 0.25
merge_method: dare_ties
base_model: tokyotech-llm/Swallow-70b-NVE-instruct-hf
dtype: bfloat16
name: Swallow-70b-NVE-RP-flavor
---
slices:
  - sources:
      - model: Swallow-70b-NVE-RP-base
        layer_range: [0, 80]
      - model: Swallow-70b-NVE-RP-flavor
        layer_range: [0, 80]
merge_method: slerp
base_model: Swallow-70b-NVE-RP-base
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5 # fallback for rest of tensors
dtype: bfloat16
name: Swallow-70b-NVE-RP