Views
No views yet
1models:
2 - model: grimjim/mistralai-Mistral-Nemo-Instruct-2407
3 parameters:
4 density: 0.2
5 weight: 0.23
6 - model: nbeerbower/mistral-nemo-bophades-12B
7 parameters:
8 density: 0.2
9 weight: 0.43
10 - model: nbeerbower/mistral-nemo-gutenberg-12B-v4
11 parameters:
12 density: 0.2
13 weight: 0.43
14 - model: TheDrummer/UnslopNemo-12B-v4.1
15 parameters:
16 density: 0.5
17 weight: 0.63
18
19merge_method: task_arithmetic
20base_model: NewEden/nemo-erebus
21parameters:
22 normalize: false
23 int8_mask: true
24dtype: bfloat16
25
26layer_parameters:
27 - range: [0, 10]
28 density_multiplier: 1.2
29 - range: [10, 20]
30 density_multiplier: 1.0
31 - range: [20, 30]
32 density_multiplier: 0.8
33 - range: [30, 40]
34 density_multiplier: 0.6
35
36regularization:
37 - method: gradient_penalty
38 scale: 0.05
39 - method: weight_clipping
40 clip_range: [-0.2, 0.2]
41 - method: random_noise
42 scale: 0.01
43 - method: attention_dropout
44 scale: 0.1
45
46postprocessing:
47 - operation: entropy_regularization
48 scale: 0.05
49 - operation: non_linear_scaling
50 parameters:
51 function: tanh
52 - operation: sharpening
53 intensity: 0.5
54 - operation: gaussian_smoothing
55 sigma: 1.5
56 - operation: normalize
57 - operation: dynamic_scaling
58 scale_range: [0.8, 1.2]
59 - operation: smoothing
60 parameters:
61 adaptive: true
62 range: [0.85, 1.15]
63 kernel_size: 5