Views
No views yet
1slices:
2 - sources:
3 - model: Undi95/PsyMedRP-v1-20B
4 layer_range: [0, 62] # PsyMedRP has 62 layers
5 - model: NeverSleep/Nethena-20B
6 layer_range: [0, 62] # Nethena-20B has 62 layers
7merge_method: slerp # Changing to SLERP method
8base_model: Undi95/PsyMedRP-v1-20B # Focus on reasoning from PsyMedRP
9parameters:
10 t:
11 - filter: self_attn
12 value: [.3, .6, .9, .6, .3] # smooth gradient of focus
13 value: [.3, .6, .9, .6, .3] # consistent level of creativity and abstract reasoning
14 - value: 0.639
15dtype: bfloat16 # Use preferred dtype