| | --- |
| | base_model: |
| | - brianmatzelle/llama3.1-8b-instruct-political-subreddits |
| | - EryriLabs/DeepSeek-R1-Distill-Llama-UK-Legislation-8B |
| | - charlieduzstuf/SAXON-0 |
| | library_name: transformers |
| | tags: |
| | - mergekit |
| | - merge |
| |
|
| | --- |
| | |
| |  |
| |
|
| | # SAXON 0 I |
| |
|
| | This is an iteration of the SAXON 0 model, I am attempting to minimise the amount of finetuning I need to do by using [mergekit](https://github.com/cg123/mergekit) to let SAXON 0 absorb models trained on Politics and specifically UK Laws etc, since I want the model to be good a those. |
| |
|
| | ## Merge Details |
| | ### Merge Method |
| |
|
| | This model was merged using the [TIES](https://arxiv.org/abs/2306.01708) merge method using [charlieduzstuf/SAXON-0](https://huggingface.co/charlieduzstuf/SAXON-0) as a base. |
| |
|
| | ### Models Merged |
| |
|
| | The following models were included in the merge: |
| | * [Llama 3.1 Politics (Dataset scraped from Reddit)](https://huggingface.co/brianmatzelle/llama3.1-8b-instruct-political-subreddits) |
| | * [Deepkseek R1 Llama Distill UK Legislation Expert](https://huggingface.co/EryriLabs/DeepSeek-R1-Distill-Llama-UK-Legislation-8B) |
| |
|
| | ### Configuration |
| |
|
| | The following YAML configuration was used to produce this model: |
| |
|
| | ```yaml |
| | models: |
| | - model: charlieduzstuf/SAXON-0 |
| | #no parameters necessary for base model |
| | - model: EryriLabs/DeepSeek-R1-Distill-Llama-UK-Legislation-8B |
| | parameters: |
| | density: 0.5 |
| | weight: 0.5 |
| | - model: brianmatzelle/llama3.1-8b-instruct-political-subreddits |
| | parameters: |
| | density: 0.5 |
| | weight: 0.5 |
| | |
| | merge_method: ties |
| | base_model: charlieduzstuf/SAXON-0 |
| | parameters: |
| | normalize: false |
| | int8_mask: true |
| | dtype: float16 |
| | ``` |
| |
|
