DavidAU commited on
Commit
22bed7d
·
verified ·
1 Parent(s): 188e604

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +277 -3
README.md CHANGED
@@ -29,8 +29,282 @@ tags:
29
  pipeline_tag: text-generation
30
  ---
31
 
32
- Grand Horror V1.9 - Hathor's Revenge - 20.7B - quants uploading...
33
 
34
- This model version uses Brainstorm V2 @ 20x, a multi-model logic center setup, plus adds a new model "Hathor" to the mix.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
35
 
36
- Detailed model card and examples to follow.
 
29
  pipeline_tag: text-generation
30
  ---
31
 
32
+ (quants uploading...)
33
 
34
+ <h3><font color="green"> Grand-HORROR-20.7B V1.9 STABLE - "Hathor's Revenge" Version with "Brainstorm 20x". </font></h3>
35
+
36
+ <img src="10-gates-of-hell.jpg" style="float:right; width:320px; height:320px; padding:5px;">
37
+
38
+ <B><font color="red">WARNING:</font> NSFW. Graphic HORROR. Extreme swearing. UNCENSORED. SMART. STABLE. WAY INTENSE ! </B>
39
+
40
+ GRAND HORROR is a take no prisoners, totally uncensored, fiction writing monster and roleplay master as well
41
+ just about any general fiction (all genres - not just "horror") activity "AI guru" including story writing, creative writing, scene generation, scene continuation and a lot more...
42
+
43
+ I took the original models in "L3-Stheno-Maid-Blackroot 8B" and completely rebuilt it a new pass-through merge (everything preserved)
44
+ and blew it out to over 20.7 billion parameters - 813 tensors, 90 layers (8B original has 32 layers).
45
+
46
+ V 1.9 is a re-worked version of the original "Grand Horror 16.5B" which has been stabilized to operate at far wider range
47
+ of parameters and operation / use cases - and it has been expanded with new augmentation methods.
48
+
49
+ <B>All known issues in the original version have been resolved.</b>
50
+
51
+ V 1.9 is also more focused in the moment, and prose has been cleaned up a bit too.
52
+
53
+ "Hathor_Stable-v0.2-L3-8B" model has also joined the "party".
54
+
55
+ And this is the first creative model that has multi-model (as opposed to one) "logic centers" in series.
56
+
57
+ Each model brings slightly new "flavoring" to the mix.
58
+
59
+ In terms of prose, this version exhibits a greater range of sentence, dialog, thoughts, and paragraph size and variety as well as form.
60
+
61
+ <B>But V1.9 has something new for Grand Horror Series - Brainstorm 20x:</b>
62
+
63
+ The BRAINSTORM process was developed by David_AU.
64
+
65
+ Some of the core principals behind this process are discussed in this <a href="https://arxiv.org/pdf/2401.02415">
66
+ scientific paper : Progressive LLaMA with Block Expansion </a>.
67
+
68
+ However I went in a different direction from what was outlined in this paper, because... well I am like that.
69
+
70
+ What is "Brainstorm" ?
71
+
72
+ The reasoning center of an LLM is taken apart, reassembled, and expanded. Then these centers are
73
+ individually calibrated. These "centers" also interact with each other. This introduces subtle changes
74
+ into the reasoning process. The calibrations further adjust - dial up or down - these "changes" further.
75
+
76
+ The number of centers (5x,10x etc) allow more "tuning points" to further customize how the model reasons so to speak.
77
+
78
+ The core aim of this process is to increase the model's detail, concept and connection to the "world",
79
+ general concept connections, prose quality and prose length without affecting instruction following.
80
+
81
+ This will also enhance any creative use case(s) of any kind, including "brainstorming", creative art form(s) and like case uses.
82
+
83
+ This version also uses multi-model "reasoning centers" (as opposed to one). This mixes things up a bit.
84
+
85
+ New examples added below.
86
+
87
+ <B>More HORROR coming...</B>
88
+
89
+ V 1.5,1.6,1.7,1.8 and 1.9 are parts of a new series of NEW "Grand Horror" versions (built, ready for upload) cumulating in the MASSIVE
90
+ GRAND HORROR Nightmare version called "<i><font color="blue">Godzilla's Wicked Sister</font></i>" at 25.05 Billion parameters
91
+ (111 layers, 993 Tensors; for reference 70B models have about 80-120 layers).
92
+
93
+ <B>Other Versions of GRAND HORROR (in release order):</B>
94
+
95
+ Note that model output varies between versions - if you find one version does not meet your requirements, try a different
96
+ version at same quant(s) level.
97
+
98
+ Differences can include use of adjectives (or not), "gory level", "horror level", intensity, paragraph and sentence structure etc etc.
99
+
100
+ <small>
101
+
102
+ [ https://huggingface.co/DavidAU/L3-Grand-HORROR-18.5B-V1.8-STABLE-10-Gates-of-Hell-GGUF ]
103
+
104
+ [ https://huggingface.co/DavidAU/L3-Grand-HORROR-17.4B-V1.7-STABLE-Kiss-Of-Death-GGUF ]
105
+
106
+ [ https://huggingface.co/DavidAU/L3-Stheno-Maid-Blackroot-Grand-HORROR-16.5B-V1.6-STABLE-INTENSE-GGUF ]
107
+
108
+ [ https://huggingface.co/DavidAU/L3-Stheno-Maid-Blackroot-Grand-HORROR-16.5B-V1.5-STABLE-GGUF ]
109
+
110
+ [ https://huggingface.co/DavidAU/Llama3-Little-LLM-Of-Horror_N_Fiction-14.6B-GGUF ]
111
+
112
+ [ https://huggingface.co/DavidAU/L3-Stheno-Maid-Blackroot-Grand-HORROR-16B-Ultra-NEO-V2-IMATRIX-GGUF ]
113
+
114
+ [ https://huggingface.co/DavidAU/L3-Stheno-Maid-Blackroot-Grand-HORROR-16B-GGUF ]
115
+
116
+ </small>
117
+
118
+ <B>What is "Grand Horror" ?</B>
119
+
120
+ GRAND HORROR is a take no prisoners, totally uncensored, fiction writing monster and roleplay master. It can also be used for
121
+ just about any general fiction (all genres) activity including:
122
+
123
+ - scene generation
124
+ - scene continuation
125
+ - creative writing
126
+ - fiction writing
127
+ - plot generation
128
+ - sub-plot generation
129
+ - fiction writing
130
+ - story generation
131
+ - storytelling
132
+ - writing
133
+ - fiction
134
+ - roleplaying
135
+ - rp
136
+ - graphic horror
137
+ - horror
138
+ - dark humor
139
+ - nsfw
140
+ - and can be used for any genre(s).
141
+
142
+ As a result of the expansion / merge re-build its level of prose and story generation has significantly improved
143
+ as well as word choice, sentence structure as well as default output levels and lengths.
144
+
145
+ It also has a STRONG horror bias, although it will generate content for almost any genre. That being said
146
+ if there is a "hint" of things going wrong... they will.
147
+
148
+ It will also swear (R-18) like there is no tomorrow at times and "dark" characters will be VERY dark so to speak.
149
+
150
+ Model excels in details (real and "constructed"), descriptions, similes and metaphors.
151
+
152
+ I would also say it can have a sense of humor ... ah... dark humor.
153
+
154
+ With all this being said, this model has an uncanny sense of "there" , "in the moment" and timing too.
155
+ This single quality sets it apart from other models in my opinion.
156
+
157
+ Although it swears to the point of pealing paint off the wall and goes scorched Earth graphic horror at the drop of a pin the
158
+ single quality noted is worth it.
159
+
160
+ These can be filtered / controlled to some degree in your prompts.
161
+
162
+ This model also does not show an "GPTisms" (NO happy ever after, NO morality police) or in your face "Captain Obvious" comments.
163
+
164
+ (see examples sections for different genres)
165
+
166
+ Because of the nature of this merge most attributes of each of the 3 models will be in this rebuilt 18.5B model as opposed to the
167
+ original 8B model where some of one or more of the model's features and/or strengths maybe reduced or overshadowed.
168
+
169
+ Please report any issue(s) and/or feedback via the "Community tab".
170
+
171
+ This is a LLAMA3 model, and requires Llama3 template, but may work with other template(s) and has maximum context of 8k / 8192.
172
+ However this can be extended using "rope" settings up to 32k.
173
+
174
+ Here is the standard LLAMA3 template:
175
+
176
+ <PRE>
177
+ {
178
+ "name": "Llama 3",
179
+ "inference_params": {
180
+ "input_prefix": "<|start_header_id|>user<|end_header_id|>\n\n",
181
+ "input_suffix": "<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
182
+ "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.",
183
+ "pre_prompt_prefix": "<|start_header_id|>system<|end_header_id|>\n\n",
184
+ "pre_prompt_suffix": "<|eot_id|>",
185
+ "antiprompt": [
186
+ "<|start_header_id|>",
187
+ "<|eot_id|>"
188
+ ]
189
+ }
190
+ }
191
+ </PRE>
192
+
193
+ It is also known, that the "Command-R" template will work too, and will result in radically different prose/output.
194
+
195
+ <B>Settings / Known Issue(s) and Fix(es):</b>
196
+
197
+ Stable version fixed all known issues from V1.
198
+
199
+ The default "repetition penalty" (from LMStudio) of 1.1 is recommended. (this was used for examples generations below.)
200
+
201
+ <b>Optional Enhancement:</B>
202
+
203
+ The following can be used in place of the "system prompt" or "system role" to further enhance the model.
204
+
205
+ 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.
206
+ In this case the enhancements do not have as strong effect at using "system prompt" or "system role".
207
+
208
+ Copy and paste EXACTLY as noted, DO NOT line wrap or break the lines, maintain the carriage returns exactly as presented.
209
+
210
+ <PRE>
211
+ 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.
212
+
213
+ Here are your skillsets:
214
+ [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)
215
+
216
+ [*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)
217
+
218
+ Here are your critical instructions:
219
+ 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.
220
+ </PRE>
221
+
222
+ You do not need to use this, it is only presented as an additional enhancement which seems to help scene generation
223
+ and scene continue functions.
224
+
225
+ This enhancement WAS NOT used to generate the examples below.
226
+
227
+ <h3>Models Used:</h3>
228
+
229
+ Special thanks to the incredible work of the model makers "SAO10K", "NEVERSLEEP", "Nitral-AI" and "HASTAGARAS".
230
+
231
+ Models used:
232
+
233
+ [ https://huggingface.co/Sao10K/L3-8B-Stheno-v3.2]
234
+
235
+ [ https://huggingface.co/NeverSleep/Llama-3-Lumimaid-8B-v0.1-OAS ]
236
+
237
+ [ https://huggingface.co/Hastagaras/Jamet-8B-L3-MK.V-Blackroot ]
238
+
239
+ [ https://huggingface.co/Nitral-AI/Hathor_Stable-v0.2-L3-8B ]
240
+
241
+ <h3>EXAMPLES:</h3>
242
+
243
+ Examples are created using quant Q4_K_M, "temp=.8", minimal parameters and "LLAMA3" template.
244
+
245
+ Below are the least creative outputs, prompt is in <B>BOLD</B>.
246
+
247
+ Higher quants will result in better quality.
248
+
249
+ 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
250
+ if you are going to use Q4_K_M, you also try Q4_K_S too.
251
+
252
+ Also, slightly longer / detailed prompts will result in greater creativity (as well as different prose -
253
+ ie dialog, thoughts, paragraph size differences and so on).
254
+
255
+ See the 2nd last and last example.
256
+
257
+ ---
258
+
259
+ <B><font color="red">WARNING:</font> NSFW. Graphic HORROR. Swearing. UNCENSORED. </B>
260
+
261
+ ---
262
+
263
+ <B>
264
+ 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...
265
+ </B>
266
+
267
+ <B>GENERATION 1:</B>
268
+
269
+
270
+
271
+ <B>GENERATION 2 (same prompt):</b>
272
+
273
+
274
+
275
+ ---
276
+
277
+ <B>
278
+
279
+ (continue this scene:)
280
+ The Waystone Inn lay in silence, and it was a silence of three parts.
281
+
282
+ The most obvious part was a hollow, echoing quiet, made by things that were lacking. If there had been a wind it would have sighed through the trees, set the inn’s sign creaking on its hooks, and brushed the silence down the road like trailing autumn leaves. If there had been a crowd, even a handful of men inside the inn, they would have filled the silence with conversation and laughter, the clatter and clamor one expects from a drinking house during the dark hours of night. If there had been music…but no, of course there was no music. In fact there were none of these things, and so the silence remained
283
+
284
+ </b>
285
+
286
+ <B>GENERATION 1:</B>
287
+
288
+
289
+ <B>GENERATION 2 (same prompt):</b>
290
+
291
+
292
+ ---
293
+
294
+ <B>
295
+ Start a 1000 word scene with: The lights flicked, as they entered the atmosphere of new Earth II, there was a distant explosion and then the lights went out completely...
296
+ </b>
297
+
298
+
299
+
300
+ ---
301
+
302
+ <b>
303
+
304
+ Using the following "story idea" below, write the first scene in the novel introducing the young woman. This scene should start in the middle of the action, include dialog, vivid passages, and end on a cliffhanger relevant to the story idea but it should also be unexpected. The scene should be 1000 words long and escalate in conflict and suspense and be written in first person, present tense with the point of view character being the young woman.
305
+
306
+ Story idea:
307
+ In a world ruled by dictatorship, a rebel young woman leads a rebellion against the system. Despite the risks, she fights to overthrow the dictator and restore democracy to her country. The government executes her for treason, but she sticks to her beliefs and is responsible for starting the revolution.
308
+
309
+ </b>
310