|
--- |
|
base_model: |
|
- meta-llama/Meta-Llama-3-70B-Instruct |
|
license: llama3 |
|
language: |
|
- en |
|
pipeline_tag: text-generation |
|
tags: |
|
- merge |
|
- frankenmerge |
|
- 96b |
|
--- |
|
# BigWeave v33 105b |
|
|
|
<img src="https://cdn-uploads.huggingface.co/production/uploads/65a6db055c58475cf9e6def1/4CbbAN-X7ZWj702JrcCGH.png" width=600> |
|
|
|
The BigWeave models aim to experimentally identify merge settings for increasing model performance. The version number merely tracks various attempts and is not a quality indicator. Only results demonstrating good performance are retained and shared. |
|
|
|
# Prompting Format |
|
llamav3 |
|
|
|
# Merge process |
|
This is a self-merge of meta-llama/Meta-Llama-3-70B-Instruct. Middle layers are duplicated and various matrices are scaled according to the template by jukofyork as shown here: https://github.com/arcee-ai/mergekit/issues/198#issuecomment-2079950009 |
|
|
|
Merge configuration: |
|
``` |
|
const_tag: &MODEL meta-llama/Meta-Llama-3-70B-Instruct |
|
|
|
const_tag: &RESIDUAL_SCALE_FACTOR 0.5 |
|
const_tag: &QK_ATTENUATION_FACTOR 0.7071067812 |
|
const_tag: &OUT_FACTOR 0.9 |
|
|
|
scale-filter-env: &scale_filter_env |
|
parameters: |
|
scale: |
|
- filter: o_proj |
|
value: *RESIDUAL_SCALE_FACTOR |
|
- filter: down_proj |
|
value: *RESIDUAL_SCALE_FACTOR |
|
- filter: q_proj |
|
value: *QK_ATTENUATION_FACTOR |
|
- filter: k_proj |
|
value: *QK_ATTENUATION_FACTOR |
|
- filter: v_proj |
|
value: *OUT_FACTOR |
|
- filter: up_proj |
|
value: *OUT_FACTOR |
|
- value: 1.0 |
|
|
|
slices: |
|
- sources: |
|
- model: *MODEL |
|
layer_range: [0, 19] |
|
- sources: |
|
- model: *MODEL |
|
layer_range: [19, 20] |
|
<<: *scale_filter_env |
|
|
|
- sources: |
|
- model: *MODEL |
|
layer_range: [10, 29] |
|
- sources: |
|
- model: *MODEL |
|
layer_range: [29, 30] |
|
<<: *scale_filter_env |
|
|
|
- sources: |
|
- model: *MODEL |
|
layer_range: [20, 39] |
|
- sources: |
|
- model: *MODEL |
|
layer_range: [39, 40] |
|
<<: *scale_filter_env |
|
|
|
- sources: |
|
- model: *MODEL |
|
layer_range: [30, 49] |
|
- sources: |
|
- model: *MODEL |
|
layer_range: [49, 50] |
|
<<: *scale_filter_env |
|
|
|
- sources: |
|
- model: *MODEL |
|
layer_range: [40, 80] |
|
|
|
merge_method: passthrough |
|
dtype: float16 |
|
``` |
