Delta-Vector commited on
Commit
1c93d5d
·
verified ·
1 Parent(s): 1319fb4

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +324 -0
README.md ADDED
@@ -0,0 +1,324 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ tags:
4
+ - fine-tuning
5
+ - prose
6
+ - KTO
7
+ - axolotl
8
+ - finetune
9
+ - roleplaying
10
+ - creative-writing
11
+ ---
12
+ <style>
13
+ @import url('https://fonts.googleapis.com/css2?family=Share+Tech+Mono&display=swap');
14
+
15
+ body {
16
+ font-family: 'Share Tech Mono', monospace;
17
+ background: #121212;
18
+ color: #E0E0E0;
19
+ margin: 0;
20
+ padding: 0;
21
+ font-size: 16px;
22
+ }
23
+ .container {
24
+ margin: 20px auto;
25
+ background-color: #1E1E1E;
26
+ padding: 20px;
27
+ border-radius: 8px;
28
+ box-shadow: 0 0 25px rgba(255, 0, 51, 0.3);
29
+ border: 1px solid #FF0033;
30
+ position: relative;
31
+ max-width: 1000px;
32
+ }
33
+ .header h1 {
34
+ font-size: 32px;
35
+ color: #FF0033;
36
+ margin: 0 0 20px 0;
37
+ text-align: center;
38
+ text-shadow: 0 0 10px rgba(255, 0, 51, 0.7);
39
+ text-transform: uppercase;
40
+ }
41
+ .section {
42
+ margin-top: 30px;
43
+ }
44
+ .section h2 {
45
+ font-size: 26px;
46
+ color: #FF0033;
47
+ text-align: center;
48
+ border-bottom: 1px solid #FF0033;
49
+ padding-bottom: 10px;
50
+ margin-bottom: 20px;
51
+ }
52
+ .info p {
53
+ color: #E0E0E0;
54
+ line-height: 1.7;
55
+ font-size: 16px;
56
+ }
57
+ .info img {
58
+ width: 100%;
59
+ max-width: 600px;
60
+ border-radius: 4px;
61
+ margin: 0 auto 15px;
62
+ display: block;
63
+ box-shadow: 0 0 20px rgba(0, 0, 0, 0.5);
64
+ border: 1px solid #444;
65
+ }
66
+ a {
67
+ color: #FF0033;
68
+ text-decoration: none;
69
+ transition: color 0.2s ease, text-shadow 0.2s ease;
70
+ }
71
+ a:hover {
72
+ color: #FFFFFF;
73
+ text-shadow: 0 0 8px rgba(255, 0, 51, 0.8);
74
+ }
75
+ .button {
76
+ display: inline-block;
77
+ background-color: transparent;
78
+ color: #FF0033;
79
+ padding: 10px 20px;
80
+ border-radius: 4px;
81
+ cursor: pointer;
82
+ text-decoration: none;
83
+ border: 1px solid #FF0033;
84
+ transition: background-color 0.2s ease, box-shadow 0.2s ease, color 0.2s ease;
85
+ }
86
+ .button:hover {
87
+ background-color: #FF0033;
88
+ color: #121212;
89
+ box-shadow: 0 0 15px rgba(255, 0, 51, 0.7);
90
+ }
91
+ pre {
92
+ background-color: #0d0d0d;
93
+ padding: 15px;
94
+ border-radius: 4px;
95
+ overflow-x: auto;
96
+ border: 1px solid #444;
97
+ border-left: 3px solid #FF0033;
98
+ color: #E0E0E0;
99
+ }
100
+ code {
101
+ font-family: 'Share Tech Mono', monospace;
102
+ color: #E0E0E0;
103
+ }
104
+ .info-card {
105
+ background: #2a2a2a;
106
+ border: 1px solid #444;
107
+ border-radius: 8px;
108
+ overflow: hidden;
109
+ }
110
+ .info-header {
111
+ background: #1a1a1a;
112
+ padding: 20px;
113
+ border-bottom: 1px solid #444;
114
+ }
115
+ .info-header h3 {
116
+ color: #FFFFFF;
117
+ margin: 0 0 10px 0;
118
+ font-size: 22px;
119
+ text-align: center;
120
+ }
121
+ .model-tags {
122
+ display: flex;
123
+ justify-content: center;
124
+ gap: 10px;
125
+ flex-wrap: wrap;
126
+ }
127
+ .model-tag {
128
+ background: #121212;
129
+ color: #FF0033;
130
+ padding: 5px 10px;
131
+ border-radius: 4px;
132
+ font-size: 12px;
133
+ border: 1px solid #FF0033;
134
+ text-transform: uppercase;
135
+ }
136
+ .model-composition {
137
+ padding: 20px;
138
+ border-bottom: 1px solid #444;
139
+ }
140
+ .model-composition h4 {
141
+ color: #FFFFFF;
142
+ margin: 0 0 15px 0;
143
+ font-size: 18px;
144
+ text-align: center;
145
+ }
146
+ .composition-list {
147
+ list-style: none;
148
+ padding: 0;
149
+ margin: 0;
150
+ display: grid;
151
+ gap: 10px;
152
+ }
153
+ .composition-list li {
154
+ color: #E0E0E0;
155
+ display: flex;
156
+ align-items: baseline;
157
+ gap: 8px;
158
+ }
159
+ .model-component {
160
+ font-weight: bold;
161
+ min-width: 120px;
162
+ }
163
+ .model-description {
164
+ padding: 20px;
165
+ background: #2a2a2a;
166
+ }
167
+ .metrics-section {
168
+ margin-bottom: 30px;
169
+ }
170
+ .metrics-section details {
171
+ background: #2a2a2a;
172
+ border: 1px solid #444;
173
+ border-radius: 8px;
174
+ padding: 15px;
175
+ margin-bottom: 15px;
176
+ }
177
+ .metrics-section summary {
178
+ color: #FF0033;
179
+ font-size: 18px;
180
+ cursor: pointer;
181
+ outline: none;
182
+ padding: 5px 0;
183
+ text-align: center;
184
+ font-weight: bold;
185
+ }
186
+ .creator-section {
187
+ margin: 20px 0;
188
+ }
189
+ .creator-badge {
190
+ display: inline-flex;
191
+ align-items: center;
192
+ background: #1a1a1a;
193
+ border: 1px solid #444;
194
+ border-radius: 4px;
195
+ padding: 10px 15px;
196
+ }
197
+ .creator-label {
198
+ color: #E0E0E0;
199
+ font-size: 14px;
200
+ margin-right: 8px;
201
+ }
202
+ .creator-link {
203
+ display: flex;
204
+ align-items: center;
205
+ gap: 5px;
206
+ color: #FF0033;
207
+ text-decoration: none;
208
+ transition: all 0.2s ease;
209
+ }
210
+ .creator-name {
211
+ font-weight: 600;
212
+ }
213
+ .creator-arrow {
214
+ font-size: 16px;
215
+ transition: transform 0.2s ease;
216
+ }
217
+ .creator-link:hover .creator-arrow {
218
+ transform: translateX(5px);
219
+ }
220
+ .link-arrow {
221
+ display: inline-block;
222
+ transition: transform 0.2s ease;
223
+ }
224
+ a:hover .link-arrow {
225
+ transform: translateX(5px);
226
+ }
227
+ .axolotl-container {
228
+ text-align: center;
229
+ margin: 30px 0;
230
+ filter: invert(1) hue-rotate(180deg);
231
+ }
232
+ .axolotl-container img {
233
+ max-width: 300px;
234
+ border-radius: 8px;
235
+ box-shadow: 0 2px 10px rgba(0, 0, 0, 0.3);
236
+ border: 1px solid #000000;
237
+ }
238
+ </style>
239
+ <html lang="en">
240
+ <head>
241
+ <meta charset="UTF-8">
242
+ <meta name="viewport" content="width=device-width, initial-scale=1.0">
243
+ <title>Rei-KTO-24B</title>
244
+ </head>
245
+ <body>
246
+ <div class="container">
247
+ <div class="header">
248
+ <h1>Rei-KTO-24B</h1>
249
+ </div>
250
+ <div class="info">
251
+ <img src="https://cdn-uploads.huggingface.co/production/uploads/66c26b6fb01b19d8c3c2467b/qUdM9qgZWeyfxwdds9QRZ.jpeg" alt="Model banner">
252
+ <div style="text-align: center;">
253
+ <div class="creator-section">
254
+ <div class="creator-badge">
255
+ <span class="creator-label">Created by</span>
256
+ <a href="https://huggingface.co/Delta-Vector" target="_blank" class="creator-link">
257
+ <span class="creator-name">Delta-Vector</span>
258
+ <span class="creator-arrow">→</span>
259
+ </a>
260
+ </div>
261
+ </div>
262
+ <div class="model-info">
263
+ <h2>Model Information</h2>
264
+ <div class="info-card">
265
+ <div class="info-header">
266
+ <h>Rei-KTO-24B</h3>
267
+ <div class="model-tags">
268
+ <span class="model-tag">KTO enhanced</span>
269
+ <span class="model-tag">Painted Fantasy Finetune</span>
270
+ <span class="model-tag">Creative Prose</span>
271
+ </div>
272
+ </div>
273
+ <div class="model-description">
274
+ <p>A model meant to replicate the style and prose of the Anthropic Claude models, Opus and Sonnet. This model is meant for Roleplaying/Creative-writing, Has some nice smarts without being too sloppy, etc - It's pretty good. Trained in 2 steps, Firstly SFT trained on Zerofata's PaintedFantasy which i found great at anime-otaku-esque characters, and then KTO'd to improve coherency and Instruct Following </p>
275
+ </div>
276
+ </div>
277
+ </div>
278
+ <div class="section">
279
+ <h2>Quantized Versions</h2>
280
+ <div class="info-card">
281
+ <div class="model-composition">
282
+ <h4>Available Downloads</h4>
283
+ <ul class="composition-list">
284
+ <li><span class="model-component"><a href="" target="_blank">GGUF Format</a></span>For use with LLama.cpp & Forks (Coming Soon!)</li>
285
+ <li><span class="model-component"><a href="" target="_blank">EXL2 Format</a></span>For use with TabbyAPI (Coming Soon!)</li>
286
+ </ul>
287
+ </div>
288
+ </div>
289
+ </div>
290
+ <div class="section">
291
+ <h2>Prompting</h2>
292
+ <p>The model is tuned with V7 Tekken formatting. A typical input would look like this:</p>
293
+ <pre><code><span style="color:#FF0033;"><s>[SYSTEM_PROMPT]system_prompt[/SYSTEM_PROMPT][INST]Hi there![/INST]Nice to meet you!</s>[INST]Can I ask a question?[/INST]
294
+ </code></pre>
295
+ </div>
296
+ <div class="section">
297
+ <h2>Training</h2>
298
+ <p>Training was done in 2 steps, SFT>KTO</p>
299
+ <div class="metrics-section">
300
+ <details>
301
+ <summary>Access Configs</summary>
302
+ <pre><code> SFT: https://wandb.ai/new-eden/Painted-Fantasy-Rei/artifacts/axolotl-config/config-u7to9d5q/v0/files/axolotl_config_f0p7vnaf.yml
303
+ KTO : https://wandb.ai/new-eden/Painted-Rei/artifacts/axolotl-config/config-8n37w77c/v0/files/axolotl_config_hvrd2tzn.yml
304
+ </code></pre>
305
+ </details>
306
+ </div>
307
+ </div>
308
+ <div class="section">
309
+ <h2>Training</h2>
310
+ <p>The training was done for 2 epochs using 8 x A100s for 24 hours/p>
311
+ <p style="text-align: center; margin-top: 20px;">
312
+ <div class="axolotl-container">
313
+ <a href="https://github.com/OpenAccess-AI-Collective/axolotl" target="_blank">
314
+ <img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl">
315
+ </a>
316
+ </div>
317
+ <div class="section">
318
+ <h2>Credits</h2>
319
+ <p>Thank you to <a href="https://huggingface.co/lucyknada">Lucy Knada</a>, <a href="https://huggingface.co/Ateron">Ateron</a>, <a href="https://huggingface.co/AliCat2">Alicat</a>, <a href="https://huggingface.co/intervitens">Intervitens</a>, <a href="https://huggingface.co/cgato">Cgato</a>, <a href="https://huggingface.co/kubernetes-bad">Kubernetes Bad</a> and the rest of <a href="https://huggingface.co/anthracite-org">Anthracite</a>.</p>
320
+ </div>
321
+ </div>
322
+ </div>
323
+ </body>
324
+ </html>