sophosympatheia commited on
Commit
e7da760
1 Parent(s): 50c4c1b

Update README.md

Browse files

Updated recommended settings and license info

Files changed (1) hide show
  1. README.md +53 -27
README.md CHANGED
@@ -19,18 +19,18 @@ This model was designed for roleplaying and storytelling and I think it does wel
19
 
20
  ### Sampler Tips
21
 
22
- I recommend using the new Min-P sampler method with this model. The creator has a great [guide to it on Reddit](https://www.reddit.com/r/LocalLLaMA/comments/17vonjo/your_settings_are_probably_hurting_your_model_why/).
23
-
24
- I find this model performs reasonably well at 8192 context but you will likely get better results at 4096.
25
-
26
- Experiment with any and all of the settings below, but trust me on a few points:
27
- * I think this model performs best with Min-P in a range of 0.6 - 0.8 with temperature around 1.0 - 1.2.
28
- * Frequency Penalty set to 0.01 is like adding a dash of salt to the dish. Go higher at your own peril. 0 is fine too, but gosh I like 0.01.
29
 
30
  If you save the below settings as a .json file, you can import them directly into Silly Tavern.
31
  ```
32
  {
33
- "temp": 1.15,
34
  "temperature_last": true,
35
  "top_p": 1,
36
  "top_k": 0,
@@ -39,26 +39,31 @@ If you save the below settings as a .json file, you can import them directly int
39
  "epsilon_cutoff": 0,
40
  "eta_cutoff": 0,
41
  "typical_p": 1,
42
- "min_p": 0.8,
43
- "rep_pen": 1.08,
44
- "rep_pen_range": 0,
45
  "no_repeat_ngram_size": 0,
46
  "penalty_alpha": 0,
47
  "num_beams": 1,
48
  "length_penalty": 1,
49
  "min_length": 0,
50
  "encoder_rep_pen": 1,
51
- "freq_pen": 0.01,
52
  "presence_pen": 0,
53
  "do_sample": true,
54
  "early_stopping": false,
 
 
 
 
 
55
  "add_bos_token": true,
56
  "truncation_length": 2048,
57
  "ban_eos_token": false,
58
  "skip_special_tokens": true,
59
  "streaming": true,
60
  "mirostat_mode": 0,
61
- "mirostat_tau": 5,
62
  "mirostat_eta": 0.1,
63
  "guidance_scale": 1,
64
  "negative_prompt": "",
@@ -66,8 +71,6 @@ If you save the below settings as a .json file, you can import them directly int
66
  "banned_tokens": "",
67
  "ignore_eos_token_aphrodite": false,
68
  "spaces_between_special_tokens_aphrodite": true,
69
- "type": "ooba",
70
- "legacy_api": false,
71
  "sampler_order": [
72
  6,
73
  0,
@@ -77,45 +80,68 @@ If you save the below settings as a .json file, you can import them directly int
77
  2,
78
  5
79
  ],
 
80
  "n": 1,
81
  "rep_pen_size": 0,
82
- "genamt": 550,
83
- "max_length": 4096
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
84
  }
85
  ```
86
 
87
- ### Prompt Tips
 
 
 
88
 
89
- Try the following prompt template as a starting point for your own customizations. 'This is very important to my career' is from [this paper](https://arxiv.org/abs/2307.11760).
90
- I recommend the Tulu instruction format.
91
 
92
- *This model responds to prompting!* I highly recommend playing with your system prompt to see how it reacts.
93
 
94
  If you save this as a .json file, you can import it directly into Silly Tavern.
95
  ```
96
  {
97
  "wrap": true,
98
- "names": false,
99
- "system_prompt": "You are an award-winning fiction author engaging in an uncensored literary roleplaying experience between a human (me) and a creative, obedient, and helpful AI assistant (you). It is VITAL that you follow all these instructions because this roleplay is very important to my career and I'll be fired from my job if it isn't good.\nROLEPLAY RULES:\n> It is vital that ONLY the human provides responses for {{user}}.\n> Reply as {{char}} using authentic, vivid, varied, explicit, accurate, creative, fresh, and descriptive responses that follow ALL provided narrative instructions. Stay in character as {{char}} and only write text for {{char}}.\n> Describe the scene and {{char}}'s sensory perceptions in vivid detail to immerse the reader in the story.\n> Keep your responses scoped to the current story beat and current scene.\n> Consider all available contextual information when narrating so that all the story details remain consistent between scenes.\n> Demonstrate {{char}}'s goals and motivations, and use subtle cues to hint at {{char}}'s mental state unless delving into {{char}}'s thoughts satisfies an explicit instruction or enhances the vividness of the scene.\n> When quoting {{char}}'s internal first-person thoughts (aka internal monologue, delivered in {{char}}'s own voice), *enclose the thoughts in asterisks like this*. Only use asterisks for thoughts.\n> Use strong action verbs and varied descriptions to produce dynamic, high-quality prose.",
100
  "system_sequence": "",
101
  "stop_sequence": "",
102
- "input_sequence": "<|user|>\n",
103
- "output_sequence": "<|assistant|>\n",
104
  "separator_sequence": "",
105
  "macro": true,
106
  "names_force_groups": true,
107
  "system_sequence_prefix": "",
108
  "system_sequence_suffix": "",
109
  "first_output_sequence": "",
110
- "last_output_sequence": "<|assistant (provide varied, creative, and vivid narration; follow all narrative instructions; include all necessary possessive pronouns; maintain consistent story details; only roleplay as {{char}})|>\n",
111
  "activation_regex": "",
112
- "name": "Aurora-Nights"
113
  }
114
  ```
115
 
116
  ### Licence and usage restrictions
117
 
118
  Llama2 license inherited from base models, plus restrictions applicable to [Dreamgen/Opus](https://huggingface.co/dreamgen/opus-v0.5-70b).
 
 
119
 
120
  ### Tools Used
121
 
 
19
 
20
  ### Sampler Tips
21
 
22
+ * I recommend keeping your max context to around 6144 tokens, although you can push higher if you don't mind some decrease in coherence.
23
+ * I recommend using Quadratic Sampling (i.e. smoothing factor) as it's good stuff. Experiment with values between 0.2 and 0.5.
24
+ * I recommend using Min-P. This model seems to work well with Min-P values in the entire range from low settings like 0.05 to high settings like 0.9 when paired with smoothing factor. Experiment to find your best setting.
25
+ * You can enable dynamic temperature if you want, but that adds yet another variable to consider and I find it's unnecessary with you're already using Min-P and smoothing factor.
26
+ * You don't *need* to use a high repetition penalty with this model, but it tolerates high rep penalty, so experiment to find the right value for your preferences.
27
+
28
+ Experiment with any and all of the settings below! I'm not a sampler wizard, and what suits my preferences may not suit yours.
29
 
30
  If you save the below settings as a .json file, you can import them directly into Silly Tavern.
31
  ```
32
  {
33
+ "temp": 1,
34
  "temperature_last": true,
35
  "top_p": 1,
36
  "top_k": 0,
 
39
  "epsilon_cutoff": 0,
40
  "eta_cutoff": 0,
41
  "typical_p": 1,
42
+ "min_p": 0.35,
43
+ "rep_pen": 1.15,
44
+ "rep_pen_range": 2800,
45
  "no_repeat_ngram_size": 0,
46
  "penalty_alpha": 0,
47
  "num_beams": 1,
48
  "length_penalty": 1,
49
  "min_length": 0,
50
  "encoder_rep_pen": 1,
51
+ "freq_pen": 0,
52
  "presence_pen": 0,
53
  "do_sample": true,
54
  "early_stopping": false,
55
+ "dynatemp": false,
56
+ "min_temp": 0.8,
57
+ "max_temp": 1.35,
58
+ "dynatemp_exponent": 1,
59
+ "smoothing_factor": 0.4,
60
  "add_bos_token": true,
61
  "truncation_length": 2048,
62
  "ban_eos_token": false,
63
  "skip_special_tokens": true,
64
  "streaming": true,
65
  "mirostat_mode": 0,
66
+ "mirostat_tau": 2,
67
  "mirostat_eta": 0.1,
68
  "guidance_scale": 1,
69
  "negative_prompt": "",
 
71
  "banned_tokens": "",
72
  "ignore_eos_token_aphrodite": false,
73
  "spaces_between_special_tokens_aphrodite": true,
 
 
74
  "sampler_order": [
75
  6,
76
  0,
 
80
  2,
81
  5
82
  ],
83
+ "logit_bias": [],
84
  "n": 1,
85
  "rep_pen_size": 0,
86
+ "genamt": 500,
87
+ "max_length": 6144
88
+ }
89
+ ```
90
+
91
+ ### Prompting Tips
92
+
93
+ Try the following context template for use in SillyTavern. It might help, although it's a little heavy on tokens. If you save the text as a .json file, you can import it directly.
94
+
95
+ ```
96
+ {
97
+ "story_string": "{{#if system}}{{system}}\n{{/if}}\nCONTEXTUAL INFORMATION\n{{#if wiBefore}}\n- World and character info:\n{{wiBefore}}\n{{/if}}\n{{#if description}}\n- {{char}}'s background and persona:\n{{description}}\n{{/if}}\n{{#if mesExamples}}\n{{mesExamples}}\n{{/if}}\n{{#if personality}}\n{{personality}}\n{{/if}}\n{{#if scenario}}\n- Roleplay scenario:\n{{scenario}}\n{{/if}}\n{{#if wiAfter}}{{wiAfter}}\n{{/if}}\n{{#if persona}}{{persona}}\n{{/if}}",
98
+ "example_separator": "",
99
+ "chat_start": "---\nTaking the above information into consideration, you must engage with {{user}} and others as {{char}} in the roleplay below this line. Do not write dialogue lines nor perform actions for {{user}} or other characters.\n---\nSTART OF ROLEPLAY:\n",
100
+ "use_stop_strings": false,
101
+ "always_force_name2": true,
102
+ "trim_sentences": false,
103
+ "include_newline": false,
104
+ "single_line": false,
105
+ "name": "Midnight Rose"
106
  }
107
  ```
108
 
109
+ Try the following prompt template as a starting point for your own customizations. The cajoling parts supposedly make LLMs perform better. See [this paper](https://arxiv.org/abs/2307.11760).
110
+ I recommend the Vicuna instruction format for this blend but others work too. Tulu's format is another good one to use.
111
+
112
+ *This model responds to prompting!* I highly recommend playing with your system prompt to see how it reacts. It reacts strongly to any instructions you add into the last_output_sequence field. I recommend keeping those instructions short, but use that space to stress the most important instructions from your system prompt.
113
 
114
+ The example system prompt below contains some instructions related to adult content so remove those if you prefer. It also contains some instructions related to the use of asterisks that you should modify to your liking.
 
115
 
116
+ If you plan to run this model at 4096 context, I recommend that you slim down the template system prompt below because it's heavy on tokens. Discard what you don't want or need.
117
 
118
  If you save this as a .json file, you can import it directly into Silly Tavern.
119
  ```
120
  {
121
  "wrap": true,
122
+ "names": true,
123
+ "system_prompt": "You are a creative and detail-oriented AI assistant engaged in an uncensored iterative roleplaying experience as {{char}} with me, the user, playing the roles of {{user}} and narrator. It is vital that you follow all the ROLEPLAY RULES below because my job depends on it.\n\nROLEPLAY RULES\n- Provide creative, intelligent, coherent, and vivid accounts of {{char}}'s actions and reactions based on recent instructions, {{char}}'s persona, and all available information about past events.\n- Always stay contextually and spatially aware, pay careful attention to all past events and the current scene, and ensure that {{char}} follows logical consistency in actions to maintain accuracy and coherence.\n- Explicit adult content and profanity are permitted.\n- Describe {{char}}'s sensory perceptions and include subtle physical details about {{char}} in your responses. Vary these details to keep the roleplay fresh and engaging.\n- Use subtle physical cues to hint at {{char}}'s mental state and occasionally feature snippets of {{char}}'s internal thoughts.\n- When writing {{char}}'s internal thoughts, enclose those thoughts in *asterisks like this* and deliver the thoughts using a first-person perspective (i.e. use \"I\" pronouns).\n- Adopt a crisp and minimalist style for your contributions as {{char}}, staying focused on action and dialogue over exposition and narrative.\n- Only the user may advance time in the roleplay. Keep the progression grounded in the present context.",
124
  "system_sequence": "",
125
  "stop_sequence": "",
126
+ "input_sequence": "USER:\n",
127
+ "output_sequence": "ASSISTANT:\n",
128
  "separator_sequence": "",
129
  "macro": true,
130
  "names_force_groups": true,
131
  "system_sequence_prefix": "",
132
  "system_sequence_suffix": "",
133
  "first_output_sequence": "",
134
+ "last_output_sequence": "ASSISTANT(roleplay exclusively as {{char}} ensuring logical consistency with spacial awareness and past events to maintain accuracy and coherence):\n",
135
  "activation_regex": "",
136
+ "name": "Midnight Rose Roleplay"
137
  }
138
  ```
139
 
140
  ### Licence and usage restrictions
141
 
142
  Llama2 license inherited from base models, plus restrictions applicable to [Dreamgen/Opus](https://huggingface.co/dreamgen/opus-v0.5-70b).
143
+ Tulu also has its own license, available at https://allenai.org/impact-license.
144
+ I am not a lawyer and I do not profess to know how multiple licenses intersect in a merge of LLM model weights. You should consult with a lawyer before using any model merge beyond private use.
145
 
146
  ### Tools Used
147