Create README.md

README.md

@@ -0,0 +1,228 @@
+---
+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
+- story
+- writing
+- fiction
+- roleplaying
+- swearing
+- rp
+- horror
+- llama3
+- mergekit
+pipeline_tag: text-generation
+---
+
+<h3>L3-Jamet-8B-MK.V-Blackroot-12.2B-V1-INSTRUCT-ULTRA-F32</h3>
+
+Merge of L3-Jamet-8B-MK.V-Blackroot (8b) with Llama3 Instruct (8b) creating a model at 12.1B to improve instruction following and output.
+
+Story / Scene / Fiction:
+
+Unique "pre-amble" / "foreshadowing" of events before they happen instead of "immediate and into the fire" type of prose.
+
+Some improvement in logic/problem solving relative to L3-Jamet-8B-MK.V-Blackroot 8B.
+
+The F32 version exhibits even stronger creativity (detail, place, "there") vs F16 version (not released) 
+
+L3-Jamet-8B-MK.V-Blackroot is a fine tune.
+
+One of the goals of this project was to see if it could be merged with Llama3 Instruct, yet maintain it's unique character YET
+also gain some "brainpower" as well.
+
+The biggest change was removal of most "tells" ( IE: "he stood frozen in horror"). 
+
+In most cases the model will describe the emotion(s) / what is happening in more detail.
+
+Other changes include prose, sentence, and paragraph structure as well as variety.
+
+A simple pass-through merge was used.
+
+See the examples below.
+
+<B>Details:</b>
+
+- Requires Llama 3 Template and/or Command-R Template
+- Context 8192, with rope 32K or higher.
+- No special settings.
+
+Please report any issue(s) and/or feedback via the "Community tab".
+
+This is a LLAMA3 model, and requires Llama3 template, but may work with other template(s) and has maximum context of 8k / 8192. 
+However this can be extended using "rope" settings up to 32k.
+
+For details on "rope" and how to set, see the BOTTOM of this page:
+
+[ https://huggingface.co/DavidAU/TieFighter-Holodeck-Holomax-Mythomax-F1-V1-COMPOS-20B-gguf ]
+
+Here is the standard LLAMA3 template:
+
+<PRE>
+{
+  "name": "Llama 3",
+  "inference_params": {
+    "input_prefix": "<|start_header_id|>user<|end_header_id|>\n\n",
+    "input_suffix": "<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
+    "pre_prompt": "You are a helpful, smart, kind, and efficient AI assistant. You always fulfill the user's requests to the best of your ability.",
+    "pre_prompt_prefix": "<|start_header_id|>system<|end_header_id|>\n\n",
+    "pre_prompt_suffix": "<|eot_id|>",
+    "antiprompt": [
+      "<|start_header_id|>",
+      "<|eot_id|>"
+    ]
+  }
+}
+</PRE>
+
+It is also known, that the "Command-R" template will work too, and will result in radically different prose/output.
+
+<B>Settings / Known Issue(s) and Fix(es):</b>
+
+The default "repetition penalty" (from LMStudio) of 1.1 is recommended. (this was used for examples generations below.)
+
+Use the smallest amount of change possible, as "rep pen" impacts creativity.
+
+Model has been tested with "temp" range of 0 to .8
+
+<b>Optional Enhancement:</B>
+
+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.
+
+<PRE>
+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.
+</PRE>
+
+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.
+
+<h3>MERGE FORMULA: (using MergeKit) </h3>
+
+Special thanks to the incredible work of the model makers "meta-llama", and "bluuwhale".
+
+Models used:
+
+[ https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct ] 
+
+[ https://huggingface.co/Hastagaras/Jamet-8B-L3-MK.V-Blackroot ]
+
+FORMULA:
+
+<PRE>
+slices:
+ - sources:
+   - model: G:/7B/Meta-Llama-3-8B-Instruct
+     layer_range: [0, 12]
+ - sources:
+   - model: G:/7B/Jamet-8B-L3-MK.V-Blackroot
+     layer_range: [6, 19]
+     parameters:
+       scale:
+         - filter: o_proj
+           value: 1
+         - filter: down_proj
+           value: 1
+         - value: 1
+ - sources:
+   - model: G:/7B/Meta-Llama-3-8B-Instruct
+     layer_range: [12, 18]
+     parameters:
+       scale:
+         - filter: o_proj
+           value: .5
+         - filter: down_proj
+           value: .5
+         - value: 1
+ - sources:
+   - model: G:/7B/Meta-Llama-3-8B-Instruct
+     layer_range: [18, 25]
+     parameters:
+       scale:
+         - filter: o_proj
+           value: .75
+         - filter: down_proj
+           value: .75
+         - value: 1
+ - sources:
+   - model: G:/7B/Jamet-8B-L3-MK.V-Blackroot
+     layer_range: [19, 32]
+     parameters:
+       scale:
+         - filter: o_proj
+           value: 1
+         - filter: down_proj
+           value: 1
+         - value: 1
+merge_method: passthrough
+dtype: float32
+</PRE>
+
+MERGEKIT NOTE:
+
+Sub in the "name" of the "creator" (of the model) in place of "G:/7B" to create a mergekit file than can be used in Mergekit Google Colab.
+
+IE: G:/7B/Jamet-8B-L3-MK.V-Blackroot -> Hastagaras/Jamet-8B-L3-MK.V-Blackroot
+
+<h3>EXAMPLES:</h3>
+
+Examples are created using quant Q4_K_M, "temp=0", minimal parameters and "LLAMA3" template. 
+
+Temp=0 was used to assess CORE changes between original SMB and the merge between it and Llama3 Instruct.
+
+Below are the least creative outputs, prompt is in <B>BOLD</B>.
+
+Higher quants will result in better quality.
+
+There will also be some variance between "close" quants like Q4_K_M/Q4_K_S and Q5_K_M/Q5_K_S, so I suggest
+if you are going to use Q4_K_M, you also try Q4_K_S too.
+
+Also, slightly longer / detailed prompts will result in greater creativity (as well as different prose - 
+ie dialog, thoughts, paragraph size differences and so on).
+
+---
+
+<B>
+Start a 1000 word scene (vivid horror, 1st person, include thoughts) with: The sky scraper swayed, as she watched the window in front of her on the 21 floor explode...
+</B>
+
+---
+
+GENERATION from "Jamet-8B-L3-MK.V-Blackroot"
+
+---
+
+
+
+---
+
+GENERATION from "L3-Jamet-8B-MK.V-Blackroot-12.2B-V1-INSTRUCT-ULTRA-F32"
+
+---
+
+