This is a merge of pre-trained language models created using
mergekit.
This model was merged using the
DARE TIES merge method using
Locutusque/llama-3-neural-chat-v1-8b as a base.
1base_model: Locutusque/llama-3-neural-chat-v1-8b
2dtype: bfloat16
3merge_method: dare_ties
4parameters:
5 int8_mask: 1.0
6 normalize: 0.0
7slices:
8- sources:
9 - layer_range: [0, 4]
10 model: cognitivecomputations/dolphin-2.9-llama3-8b
11 parameters:
12 density: 1.0
13 weight: 0.6
14 - layer_range: [0, 4]
15 model: Weyaxi/Einstein-v6.1-Llama3-8B
16 parameters:
17 density: 0.6
18 weight: 0.5
19 - layer_range: [0, 4]
20 model: Locutusque/llama-3-neural-chat-v1-8b
21 parameters:
22 density: 1.0
23 weight: 0.5
24- sources:
25 - layer_range: [4, 8]
26 model: cognitivecomputations/dolphin-2.9-llama3-8b
27 parameters:
28 density: 0.8
29 weight: 0.1
30 - layer_range: [4, 8]
31 model: Weyaxi/Einstein-v6.1-Llama3-8B
32 parameters:
33 density: 1.0
34 weight: 0.2
35 - layer_range: [4, 8]
36 model: Locutusque/llama-3-neural-chat-v1-8b
37 parameters:
38 density: 1.0
39 weight: 0.7
40- sources:
41 - layer_range: [8, 12]
42 model: cognitivecomputations/dolphin-2.9-llama3-8b
43 parameters:
44 density: 0.7
45 weight: 0.1
46 - layer_range: [8, 12]
47 model: Weyaxi/Einstein-v6.1-Llama3-8B
48 parameters:
49 density: 0.7
50 weight: 0.2
51 - layer_range: [8, 12]
52 model: Locutusque/llama-3-neural-chat-v1-8b
53 parameters:
54 density: 0.7
55 weight: 0.6
56- sources:
57 - layer_range: [12, 16]
58 model: cognitivecomputations/dolphin-2.9-llama3-8b
59 parameters:
60 density: 0.9
61 weight: 0.2
62 - layer_range: [12, 16]
63 model: Weyaxi/Einstein-v6.1-Llama3-8B
64 parameters:
65 density: 0.6
66 weight: 0.6
67 - layer_range: [12, 16]
68 model: Locutusque/llama-3-neural-chat-v1-8b
69 parameters:
70 density: 0.7
71 weight: 0.3
72- sources:
73 - layer_range: [16, 20]
74 model: cognitivecomputations/dolphin-2.9-llama3-8b
75 parameters:
76 density: 1.0
77 weight: 0.2
78 - layer_range: [16, 20]
79 model: Weyaxi/Einstein-v6.1-Llama3-8B
80 parameters:
81 density: 1.0
82 weight: 0.2
83 - layer_range: [16, 20]
84 model: Locutusque/llama-3-neural-chat-v1-8b
85 parameters:
86 density: 0.9
87 weight: 0.4
88- sources:
89 - layer_range: [20, 24]
90 model: cognitivecomputations/dolphin-2.9-llama3-8b
91 parameters:
92 density: 0.7
93 weight: 0.2
94 - layer_range: [20, 24]
95 model: Weyaxi/Einstein-v6.1-Llama3-8B
96 parameters:
97 density: 0.9
98 weight: 0.3
99 - layer_range: [20, 24]
100 model: Locutusque/llama-3-neural-chat-v1-8b
101 parameters:
102 density: 1.0
103 weight: 0.4
104- sources:
105 - layer_range: [24, 28]
106 model: cognitivecomputations/dolphin-2.9-llama3-8b
107 parameters:
108 density: 1.0
109 weight: 0.4
110 - layer_range: [24, 28]
111 model: Weyaxi/Einstein-v6.1-Llama3-8B
112 parameters:
113 density: 0.8
114 weight: 0.2
115 - layer_range: [24, 28]
116 model: Locutusque/llama-3-neural-chat-v1-8b
117 parameters:
118 density: 0.9
119 weight: 0.4
120- sources:
121 - layer_range: [28, 32]
122 model: cognitivecomputations/dolphin-2.9-llama3-8b
123 parameters:
124 density: 1.0
125 weight: 0.3
126 - layer_range: [28, 32]
127 model: Weyaxi/Einstein-v6.1-Llama3-8B
128 parameters:
129 density: 0.9
130 weight: 0.2
131 - layer_range: [28, 32]
132 model: Locutusque/llama-3-neural-chat-v1-8b
133 parameters:
134 density: 1.0
135 weight: 0.3
136
Detailed results can be found
here
