Delete mergekit_config.yml
Browse files- mergekit_config.yml +0 -172
mergekit_config.yml
DELETED
|
@@ -1,172 +0,0 @@
|
|
| 1 |
-
# 32 layers -> VS 40
|
| 2 |
-
#models:
|
| 3 |
-
# - model: G:/7B/L3-8B-Stheno-v3.2
|
| 4 |
-
# - model: G:/7B/Llama-3-Lumimaid-8B-v0.1-OAS
|
| 5 |
-
# - model: G:/7B/L3-Jamet-8B-MK.V-Blackroot
|
| 6 |
-
#merge_method: model_stock
|
| 7 |
-
#base_model: G:/7B/L3-8B-Stheno-v3.2
|
| 8 |
-
#dtype: float32
|
| 9 |
-
|
| 10 |
-
slices:
|
| 11 |
-
- sources:
|
| 12 |
-
- model: G:/7B/L3-8B-Stheno-v3.2
|
| 13 |
-
layer_range: [0, 14]
|
| 14 |
-
parameters:
|
| 15 |
-
scale:
|
| 16 |
-
- filter: o_proj
|
| 17 |
-
value: 1
|
| 18 |
-
- filter: down_proj
|
| 19 |
-
value: 1
|
| 20 |
-
- value: 1
|
| 21 |
-
- sources:
|
| 22 |
-
- model: G:/7B/L3-Lumimaid-8B-v0.1-OAS
|
| 23 |
-
layer_range: [8, 20]
|
| 24 |
-
parameters:
|
| 25 |
-
scale:
|
| 26 |
-
- filter: o_proj
|
| 27 |
-
value: 1
|
| 28 |
-
- filter: down_proj
|
| 29 |
-
value: 1
|
| 30 |
-
- value: 1
|
| 31 |
-
- sources:
|
| 32 |
-
- model: G:/7B/L3-Jamet-8B-MK.V-Blackroot
|
| 33 |
-
layer_range: [12, 24]
|
| 34 |
-
parameters:
|
| 35 |
-
scale:
|
| 36 |
-
- filter: o_proj
|
| 37 |
-
value: 1
|
| 38 |
-
- filter: down_proj
|
| 39 |
-
value: 1
|
| 40 |
-
- value: 1
|
| 41 |
-
- sources:
|
| 42 |
-
- model: G:/7B/L3-8B-Stheno-v3.2
|
| 43 |
-
layer_range: [14, 20]
|
| 44 |
-
parameters:
|
| 45 |
-
scale:
|
| 46 |
-
- filter: o_proj
|
| 47 |
-
value: .8
|
| 48 |
-
- filter: down_proj
|
| 49 |
-
value: .8
|
| 50 |
-
- value: .8
|
| 51 |
-
- sources:
|
| 52 |
-
- model: G:/7B/L3-8B-Stheno-v3.2
|
| 53 |
-
layer_range: [20, 25]
|
| 54 |
-
parameters:
|
| 55 |
-
scale:
|
| 56 |
-
- filter: o_proj
|
| 57 |
-
value: .8
|
| 58 |
-
- filter: down_proj
|
| 59 |
-
value: .8
|
| 60 |
-
- value: .8
|
| 61 |
-
- sources:
|
| 62 |
-
- model: G:/7B/L3-8B-Stheno-v3.2
|
| 63 |
-
layer_range: [25, 27]
|
| 64 |
-
parameters:
|
| 65 |
-
scale:
|
| 66 |
-
- filter: o_proj
|
| 67 |
-
value: .6
|
| 68 |
-
- filter: down_proj
|
| 69 |
-
value: .6
|
| 70 |
-
- value: 1
|
| 71 |
-
- sources:
|
| 72 |
-
- model: G:/7B/L3-8B-Stheno-v3.2
|
| 73 |
-
layer_range: [27, 28]
|
| 74 |
-
parameters:
|
| 75 |
-
scale:
|
| 76 |
-
- filter: o_proj
|
| 77 |
-
value: .9
|
| 78 |
-
- filter: down_proj
|
| 79 |
-
value: .9
|
| 80 |
-
- value: 1
|
| 81 |
-
- sources:
|
| 82 |
-
- model: G:/7B/L3-Lumimaid-8B-v0.1-OAS
|
| 83 |
-
layer_range: [20, 25]
|
| 84 |
-
parameters:
|
| 85 |
-
scale:
|
| 86 |
-
- filter: o_proj
|
| 87 |
-
value: 1
|
| 88 |
-
- filter: down_proj
|
| 89 |
-
value: 1
|
| 90 |
-
- value: 1
|
| 91 |
-
- sources:
|
| 92 |
-
- model: G:/7B/L3-Lumimaid-8B-v0.1-OAS
|
| 93 |
-
layer_range: [25, 27]
|
| 94 |
-
parameters:
|
| 95 |
-
scale:
|
| 96 |
-
- filter: o_proj
|
| 97 |
-
value: .6
|
| 98 |
-
- filter: down_proj
|
| 99 |
-
value: .6
|
| 100 |
-
- value: 1
|
| 101 |
-
- sources:
|
| 102 |
-
- model: G:/7B/L3-Lumimaid-8B-v0.1-OAS
|
| 103 |
-
layer_range: [27, 31]
|
| 104 |
-
parameters:
|
| 105 |
-
scale:
|
| 106 |
-
- filter: o_proj
|
| 107 |
-
value: 1
|
| 108 |
-
- filter: down_proj
|
| 109 |
-
value: 1
|
| 110 |
-
- value: 1
|
| 111 |
-
- sources:
|
| 112 |
-
- model: G:/7B/L3-Jamet-8B-MK.V-Blackroot
|
| 113 |
-
layer_range: [24, 31]
|
| 114 |
-
parameters:
|
| 115 |
-
scale:
|
| 116 |
-
- filter: o_proj
|
| 117 |
-
value: 1
|
| 118 |
-
- filter: down_proj
|
| 119 |
-
value: 1
|
| 120 |
-
- value: 1
|
| 121 |
-
- sources:
|
| 122 |
-
- model: G:/7B/L3-Jamet-8B-MK.V-Blackroot
|
| 123 |
-
layer_range: [31, 32]
|
| 124 |
-
parameters:
|
| 125 |
-
scale:
|
| 126 |
-
- filter: o_proj
|
| 127 |
-
value: 0.3333333333333
|
| 128 |
-
- filter: down_proj
|
| 129 |
-
value: 0.3333333333333
|
| 130 |
-
- value: 0.3333333333333
|
| 131 |
-
- sources:
|
| 132 |
-
- model: G:/7B/L3-Jamet-8B-MK.V-Blackroot
|
| 133 |
-
layer_range: [31, 32]
|
| 134 |
-
parameters:
|
| 135 |
-
scale:
|
| 136 |
-
- filter: o_proj
|
| 137 |
-
value: 0.4444444444444
|
| 138 |
-
- filter: down_proj
|
| 139 |
-
value: 0.4444444444444
|
| 140 |
-
- value: 0.4444444444444
|
| 141 |
-
- sources:
|
| 142 |
-
- model: G:/7B/L3-Jamet-8B-MK.V-Blackroot
|
| 143 |
-
layer_range: [31, 32]
|
| 144 |
-
parameters:
|
| 145 |
-
scale:
|
| 146 |
-
- filter: o_proj
|
| 147 |
-
value: 0.5555555555555
|
| 148 |
-
- filter: down_proj
|
| 149 |
-
value: 0.5555555555555
|
| 150 |
-
- value: 0.5555555555555
|
| 151 |
-
- sources:
|
| 152 |
-
- model: G:/7B/L3-Jamet-8B-MK.V-Blackroot
|
| 153 |
-
layer_range: [31, 32]
|
| 154 |
-
parameters:
|
| 155 |
-
scale:
|
| 156 |
-
- filter: o_proj
|
| 157 |
-
value: 0.6666666666666
|
| 158 |
-
- filter: down_proj
|
| 159 |
-
value: 0.6666666666666
|
| 160 |
-
- value: 0.6666666666666
|
| 161 |
-
- sources:
|
| 162 |
-
- model: G:/7B/L3-Jamet-8B-MK.V-Blackroot
|
| 163 |
-
layer_range: [31, 32]
|
| 164 |
-
parameters:
|
| 165 |
-
scale:
|
| 166 |
-
- filter: o_proj
|
| 167 |
-
value: 0.7777777777777
|
| 168 |
-
- filter: down_proj
|
| 169 |
-
value: 0.7777777777777
|
| 170 |
-
- value: 0.8888888888888
|
| 171 |
-
merge_method: passthrough
|
| 172 |
-
dtype: float16
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|