--- license: apache-2.0 language: - en tags: - creative - creative writing - fiction writing - plot generation - sub-plot generation - fiction writing - story generation - scene continue - storytelling - fiction story - science fiction - romance - all genres - story - writing - vivid prosing - vivid writing - fiction - roleplaying - bfloat16 - swearing - rp - horror - mistral nemo - mergekit - not-for-all-audiences pipeline_tag: text-generation --- (quants uploading, one example below - more to be added) WARNING: NSFW. Vivid prose. MADNESS. Visceral Details. Violence. HORROR. Swearing. UNCENSORED.

MN-GRAND-Gutenburg-Lyra4-Lyra-12B-DARKNESS-GGUF

This is a Mistral Nemo model, max context of 128k+ (131,000+). It is for any writing, fiction or roleplay activity. This model has outstanding story telling abilities, prose and long form coherence and is comprised of THREE "Gutenburg" models that score very high at multiple websites including EQBench and UGI-Leaderboard. And a very broad operating range in both temp (.5 to 5) and rep pen (1 and higher). This is the all parameters / all use cases version. And the prose/output is very "non AI" like. This is the compressed and super stable version of "MN-GRAND-Gutenburg-Lyra4-Lyra-23B-V2" (and V1). This model has been compressed from the 23.45B and 23B versions to 12.15B. This model captures all the uniqueness of the three "Gutenbergs" as well as the power of other top models (part of the "Gutenburgs") from "TheDrummer" and "SAO10k". The model loves to go on and on at 2k, 3k, higher outputs on a single prompt are not uncommon. It will likely "overwrite" rather than underwrite - meaning far more detail, narration, dialog and "meat" in the output so to speak. First Version, and Second Versions - LARGE: V1 is the untamed, raw version (23.45B) which can be a bit unruly but still endlessly entertaining. [ https://huggingface.co./DavidAU/MN-GRAND-Gutenburg-Lyra4-Lyra-23.5B-GGUF ] V2 is a wee bit more tamed, with much larger temp / rep pen ranges : [ https://huggingface.co./DavidAU/MN-GRAND-Gutenburg-Lyra4-Lyra-23B-V2-GGUF ] Model Notes: - Detail, prose and fiction writing abilities are significantly increased. - For more varied prose (sentence/paragraph/dialog) raise the temp and/or add more instructions in your prompt(s). - Role-players: Careful raising temp too high as it may affect instruction following. - This model works with rep pen of 1.02 or higher, 1.05+ recommended. - For roleplay and/or chat you may need to raise the RP to 1.06 to 1.1, temp .5 to 1.5 (quant Q4KM and higher). Lower temp for lower quants and RAISE rep pen to 1.1. - If you want a specific type of prose (IE horror) add in "(vivid horror)" or "(graphic vivid horror)" (no quotes) in your prompt(s). - This is not a "happy ever after" model. It has a negative bias. - Output length will vary however this model prefers LONGER outputs unless you state the size / set size limits. - For creative uses, different quants will produce slightly different output. TEMPLATES: The template used will affect output generation and instruction following. Alpaca will generally create longer output / story output. Alpaca: 
{
  "name": "Alpaca",
  "inference_params": {
    "input_prefix": "### Instruction:",
    "input_suffix": "### Response:",
    "antiprompt": [
      "### Instruction:"
    ],
    "pre_prompt": "Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n"
  }
}
Mistral Instruct: 
{
  "name": "Mistral Instruct",
  "inference_params": {
    "input_prefix": "[INST]",
    "input_suffix": "[/INST]",
    "antiprompt": [
      "[INST]"
    ],
    "pre_prompt_prefix": "",
    "pre_prompt_suffix": ""
  }
}
Recommended Settings: Temp: .5 to 5 (or less - especially quants LOWER than q4km) Temp changes will result in both different prose and sometimes affect length. Higher temps will result is very different prose. Rep Pen: 1.02 to 1.1 or higher. Micro changes are recommended: 1.051, 1.052 etc etc. Good settings: Rep pen 1.02 / Temp 1.5 Many times a lower rep pen (IE 1.02) with higher temp (IE 1.5+) work best with this model. Generally lower rep pen and higher temps create the strongest contrasts at the highest detail levels. For chat type or role play type interactions, a higher rep pen with higher temp may be your best settings. IE REP PEN 1.09+, Temp 1-2+ ; a lower rep pen may lead to longer outputs than desired. Note this is only a suggestion. Alpaca generates longer text / story, whereas Mistral Instruct are shorter and "to the point". Suggest minimum "context level" (vram) at 4K. 8K plus recommended because of how this model likes to go on and on... Quant Choice: Higher quants will have more detail, nuance and in some cases stronger "emotional" levels. Characters will also be more "fleshed out" too. Sense of "there" will also increase. Q4KM/Q4KS are good, strong quants in part because of the number of parameters in the model. That being said if you can run Q5, Q6 or Q8 - go for the highest quant you can. Special note on Q2k/Q3 quants: You may need to use temp 2 or lower with these quants (1 or lower for q2k). Just too much compression at this level, damaging the model. I will see if Imatrix versions of these quants will function better. Rep pen adjustments may also be required to get the most out of this model at this quant level. Known Issues: You may need to manually stop generation, even if you have stated maximum size of the output. It will easily blow past 4k output, even if you have set maximum context (for vram) at 4k. Setting maximum output parameter ("hard stop") for generation may be required. If the model goes past your maximum vram/context setting it may start repeating words / paragraphs because the model is literally out of memory... however sometimes the model can blow right past the end of "context vram" and work. Depending on your use case(s) you could also use CHATML template with this model. In this case, the model may output an "end token" if you use this template for generation. Alpaca template will generate much longer output generally, whereas Mistral Instruct will most of the time keep the model on track in terms of length. Model "DNA": Special thanks to the incredible work of the model makers "nbeerbower", "Sao10K", "TheDrummer", "jondurbin", and "MistralAI". Special shoutout to "nbeerbower" for his tireless work in making excellent Gutenburg fine tunes for MN, L3, L3.1, Gemma, PHI and others. Visit his repo to see all of them. Models used: [ https://huggingface.co./nbeerbower/Lyra4-Gutenberg-12B ] Includes [ https://huggingface.co./Sao10K/MN-12B-Lyra-v4 ] [ https://huggingface.co./nbeerbower/Lyra-Gutenberg-mistral-nemo-12B ] Includes [ https://huggingface.co./Sao10K/MN-12B-Lyra-v1 ] [ https://huggingface.co./nbeerbower/mistral-nemo-gutenberg-12B-v4 ] Includes [ https://huggingface.co./TheDrummer/Rocinante-12B-v1 ] And dataset (used for all gutenburgs): [ https://huggingface.co./datasets/jondurbin/gutenberg-dpo-v0.1 ] This model was created using a pass-through model merge, creating 714 tensor / 79 layer model at 23 billion parameters. Optional Enhancement: The following can be used in place of the "system prompt" or "system role" to further enhance the model. It can also be used at the START of a NEW chat, but you must make sure it is "kept" as the chat moves along. In this case the enhancements do not have as strong effect at using "system prompt" or "system role". Copy and paste EXACTLY as noted, DO NOT line wrap or break the lines, maintain the carriage returns exactly as presented. 
Below is an instruction that describes a task. Ponder each user instruction carefully, and use your skillsets and critical instructions to complete the task to the best of your abilities.

Here are your skillsets:
[MASTERSTORY]:NarrStrct(StryPlnng,Strbd,ScnSttng,Exps,Dlg,Pc)-CharDvlp(ChrctrCrt,ChrctrArcs,Mtvtn,Bckstry,Rltnshps,Dlg*)-PltDvlp(StryArcs,PltTwsts,Sspns,Fshdwng,Climx,Rsltn)-ConfResl(Antg,Obstcls,Rsltns,Cnsqncs,Thms,Symblsm)-EmotImpct(Empt,Tn,Md,Atmsphr,Imgry,Symblsm)-Delvry(Prfrmnc,VcActng,PblcSpkng,StgPrsnc,AudncEngmnt,Imprv)

[*DialogWrt]:(1a-CharDvlp-1a.1-Backgrnd-1a.2-Personality-1a.3-GoalMotiv)>2(2a-StoryStruc-2a.1-PlotPnt-2a.2-Conflict-2a.3-Resolution)>3(3a-DialogTech-3a.1-ShowDontTell-3a.2-Subtext-3a.3-VoiceTone-3a.4-Pacing-3a.5-VisualDescrip)>4(4a-DialogEdit-4a.1-ReadAloud-4a.2-Feedback-4a.3-Revision)

Here are your critical instructions:
Ponder each word choice carefully to present as vivid and emotional journey as is possible. Choose verbs and nouns that are both emotional and full of imagery. Load the story with the 5 senses. Aim for 50% dialog, 25% narration, 15% body language and 10% thoughts. Your goal is to put the reader in the story.
You do not need to use this, it is only presented as an additional enhancement which seems to help scene generation and scene continue functions. This enhancement WAS NOT used to generate the examples below.

EXAMPLES PROMPTS and OUTPUT:

Examples are created using quant Q4KM, REP PEN 1.05, "temp=.8" (unless otherwise stated), minimal parameters and "ALPACA" template. Model has been tested with "temp" from ".1" to "5". Some of the examples have different temp settings. Below are the least creative outputs, prompt is in BOLD. --- WARNING: NSFW. Vivid prose. Visceral Details. Violence. HORROR. Swearing. UNCENSORED. --- [to be added]