Views
No views yet
1# models:
2# - model: anthracite-org/magnum-v2-12b # instruct model
3# parameters:
4# density: 0.6
5# weight: 0.5
6# # - model: /home/paperspace/projects/project/finetunellm/outputs/nemo-12b-creative/merged # creative writing model
7# # parameters:
8# # density: 0.3
9# # weight: 0.3
10# - model: alexxi19/ft-nemo-base-lora # sft model
11# parameters:
12# density: 0.7
13# weight: 0.5
14# merge_method: dare_ties
15# base_model: alexxi19/ft-nemo-base-lora
16# parameters:
17# int8_mask: true
18# rescale: true
19# # normalize: true
20# dtype: bfloat16
21# chat_template: chatml
22
23
24# models:
25# - model: anthracite-org/magnum-v2-12b # instruct model
26# parameters:
27# density: 0.3
28# weight: 0.5
29# # - model: /home/paperspace/projects/project/finetunellm/outputs/nemo-12b-creative/merged # creative writing model
30# # parameters:
31# # density: 0.3
32# # weight: 0.3
33# - model: alexxi19/ft-nemo-base-lora # sft model
34# parameters:
35# density: 0.7
36# weight: 0.5
37# base_model: alexxi19/ft-nemo-base-lora
38# dtype: bfloat16
39# chat_template: chatml
40# merge_method: della_linear
41# parameters:
42# epsilon: 0.05
43# int8_mask: true
44# rescale: true
45# lambda: 1.0
46
47
48models:
49 - model: Nitral-AI/Captain_BMO-12B # another sft model
50 - model: alexxi19/ft-v1-nemo-base-lora # sft model
51merge_method: slerp
52base_model: alexxi19/ft-v1-nemo-base-lora
53parameters:
54 t:
55 - filter: self_attn
56 value: [0, 0.5, 0.3, 0.7, 1]
57 - filter: mlp
58 value: [1, 0.5, 0.7, 0.3, 0]
59 - value: 0.3 # fallback for rest of tensors
60dtype: float16
61chat_template: chatml
62