sunfall-nemomix-unleashed-12b-v0.6.1-8bit

Sunfall (2024-07-31) v0.6.1 on top of https://huggingface.co/mistralai/Mistral-Nemo-Instruct-2407

Experimental. Please give feedback. Begone if you demand perfection.

This experiment is showing potential but it is still early to tell if it is a viable option. Invest your time accordingly.

Mergers/fine-tuners: there is a LoRA of this model. Consider merging that instead of merging this model.

To use lore book tags (example), make sure you use Status: Blue (constant) and write e.g.

Follow the Diamond Law at all costs.

Tags: humor, dark, complex storytelling, intricate characters, immersive.

This model has been trained on context that mimics that of Silly Tavern's Mistral instruct preset with the Always add character's name to prompt turned off.

Recommended stop strings: ["\nYOURCHARNAME", "\n YOURCHARNAME"]

The card has also been trained on content which includes a narrator card, which was used when the content did not mainly revolve around two characters. Future versions will expand on this idea, so forgive the vagueness at this time.

(The Diamond Law is this, although new rules were added: https://files.catbox.moe/d15m3g.txt -- So far results are unclear, but the training was done with this phrase included, and the training data adheres to the law.)

The model has also been trained to do storywriting. The system message ends up looking something like this:

You are an expert storyteller, who can roleplay or write compelling stories. Follow the Diamond Law at all costs. Below is a scenario with character descriptions and content tags. Write a story based on this scenario.

Scenario: The story is about James, blabla.

James is an overweight 63 year old blabla.

Lucy: James's 62 year old wife.

Tags: tag1, tag2, tag3, ...

MMLU-Pro Benchmark: the model performs slightly better than the base instruct model, except for engineering, law, philosophy, psychology. As the primary purpose is roleplay, this benchmark is in place to demonstrate that the model did not become retarded in the training process. Do not expect miracles in other areas.

Mistral Nemo Instruct:

| overall | biology | business | chemistry | computer science | economics | engineering | health | history |  law  | math  | philosophy | physics | psychology | other |
| ------- | ------- | -------- | --------- | ---------------- | --------- | ----------- | ------ | ------- | ----- | ----- | ---------- | ------- | ---------- | ----- |
|   42.15 |   56.52 |    44.00 |     38.89 |            46.15 |     44.44 |       32.26 |  42.31 |   75.00 | 37.14 | 37.21 |      43.75 |   26.83 |      44.00 | 58.62 |
|     161 |      13 |       11 |        14 |                6 |        12 |          10 |     11 |       9 |    13 |    16 |          7 |      11 |         11 |    17 |
|     382 |      23 |       25 |        36 |               13 |        27 |          31 |     26 |      12 |    35 |    43 |         16 |      41 |         25 |    29 |

Nemo Sunfall v0.6.1:

# | overall | biology | business | chemistry | computer science | economics | engineering | health | history |  law  | math  | philosophy | physics | psychology | other |
# | ------- | ------- | -------- | --------- | ---------------- | --------- | ----------- | ------ | ------- | ----- | ----- | ---------- | ------- | ---------- | ----- |
# |   45.29 |   65.22 |    48.00 |     44.44 |            53.85 |     55.56 |       29.03 |  46.15 |   75.00 | 34.29 | 44.19 |      37.50 |   34.15 |      40.00 | 58.62 |
# |     173 |      15 |       12 |        16 |                7 |        15 |           9 |     12 |       9 |    12 |    19 |          6 |      14 |         10 |    17 |
# |     382 |      23 |       25 |        36 |               13 |        27 |          31 |     26 |      12 |    35 |    43 |         16 |      41 |         25 |    29 |