metadata
base_model:
- Sao10K/Fimbulvetr-11B-v2
library_name: transformers
tags:
- mergekit
- merge
merge
This is a merge of pre-trained language models created using mergekit.
Merge Details
Merge Method
This model was merged using the passthrough merge method.
Models Merged
The following models were included in the merge:
Configuration
The following YAML configuration was used to produce this model:
slices:
- sources:
- model: Sao10K/Fimbulvetr-11B-v2
layer_range: [0, 16]
parameters:
attention:
- filter: grouped_qk_proj
clusters: 8
value: 1.2
attention_type: hierarchical
local_attention: 0.5
global_attention: 1.5
dynamic_weighting: true
rope_scaling: 10000
significance: 0.85
mlp:
filter: intermediate_proj
value: 14336
compression: true
dynamic_capacity: true
regularization:
filter: attention_dropout
value: 0.0
- sources:
- model: Sao10K/Fimbulvetr-11B-v2
layer_range: [16, 32]
parameters:
attention:
- filter: grouped_qk_proj
clusters: 8
value: 1.3
attention_type: hierarchical
local_attention: 0.6
global_attention: 1.4
dynamic_weighting: true
rope_scaling: 10000
significance: 0.80
mlp:
filter: intermediate_proj
value: 14336
compression: true
dynamic_capacity: true
regularization:
filter: attention_dropout
value: 0.0
- sources:
- model: Sao10K/Fimbulvetr-11B-v2
layer_range: [32, 48]
parameters:
attention:
- filter: grouped_qk_proj
clusters: 8
value: 1.5
attention_type: hierarchical
local_attention: 0.7
global_attention: 1.6
dynamic_weighting: true
rope_scaling: 10000
significance: 0.9
mlp:
filter: intermediate_proj
value: 14336
compression: true
dynamic_capacity: true
regularization:
filter: attention_dropout
value: 0.0
merge_method: passthrough
dtype: bfloat16