Views
No views yet
1dtype: bfloat16
2merge_method: passthrough
3slices:
4- sources:
5 - layer_range: [0, 8]
6 model: nvidia/Mistral-NeMo-Minitron-8B-Base
7- sources:
8 - layer_range: [8, 16]
9 model: nvidia/Mistral-NeMo-Minitron-8B-Base
10 parameters:
11 scale:
12 - filter: o_proj
13 value: 0.5
14 - filter: down_proj
15 value: 0.5
16 - filter: q_proj
17 value: 0.85355339059
18 - filter: k_proj
19 value: 0.85355339059
20 - value: 1.0
21- sources:
22 - layer_range: [8, 16]
23 model: nvidia/Mistral-NeMo-Minitron-8B-Base
24 parameters:
25 scale:
26 - filter: q_proj
27 value: 0.85355339059
28 - filter: k_proj
29 value: 0.85355339059
30 - value: 1.0
31- sources:
32 - layer_range: [16, 17]
33 model: nvidia/Mistral-NeMo-Minitron-8B-Base
34- sources:
35 - layer_range: [17, 24]
36 model: nvidia/Mistral-NeMo-Minitron-8B-Base
37 parameters:
38 scale:
39 - filter: o_proj
40 value: 0.5
41 - filter: down_proj
42 value: 0.5
43 - filter: q_proj
44 value: 0.85355339059
45 - filter: k_proj
46 value: 0.85355339059
47 - value: 1.0
48- sources:
49 - layer_range: [17, 24]
50 model: nvidia/Mistral-NeMo-Minitron-8B-Base
51 parameters:
52 scale:
53 - filter: q_proj
54 value: 0.85355339059
55 - filter: k_proj
56 value: 0.85355339059
57 - value: 1.0
58- sources:
59 - layer_range: [24, 25]
60 model: nvidia/Mistral-NeMo-Minitron-8B-Base
61- sources:
62 - layer_range: [25, 32]
63 model: nvidia/Mistral-NeMo-Minitron-8B-Base
64 parameters:
65 scale:
66 - filter: o_proj
67 value: 0.5
68 - filter: down_proj
69 value: 0.5
70 - filter: q_proj
71 value: 0.85355339059
72 - filter: k_proj
73 value: 0.85355339059
74 - value: 1.0
75- sources:
76 - layer_range: [25, 32]
77 model: nvidia/Mistral-NeMo-Minitron-8B-Base
78 parameters:
79 scale:
80 - filter: q_proj
81 value: 0.85355339059
82 - filter: k_proj
83 value: 0.85355339059
84 - value: 1.0
85- sources:
86 - layer_range: [32, 40]
87 model: nvidia/Mistral-NeMo-Minitron-8B-Base