File size: 42,632 Bytes
f8f1add 7808abd 6c0b715 a2073e1 6c0b715 a2073e1 6c0b715 a2073e1 6c0b715 7808abd f8f1add 8e7a56f f8f1add 4aae6d1 f8f1add a2073e1 7808abd 6c0b715 7808abd a2073e1 7808abd a2073e1 6c0b715 a2073e1 7808abd a2073e1 7808abd a2073e1 7808abd 6c0b715 7808abd 6c0b715 7808abd f8f1add |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 |
---
license: other
inference: false
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<img src="https://i.imgur.com/EBdldam.jpg" alt="TheBlokeAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</div>
<div style="display: flex; justify-content: space-between; width: 100%;">
<div style="display: flex; flex-direction: column; align-items: flex-start;">
<p style="margin-top: 0.5em; margin-bottom: 0em;"><a href="https://discord.gg/theblokeai">Chat & support: TheBloke's Discord server</a></p>
</div>
<div style="display: flex; flex-direction: column; align-items: flex-end;">
<p style="margin-top: 0.5em; margin-bottom: 0em;"><a href="https://www.patreon.com/TheBlokeAI">Want to contribute? TheBloke's Patreon page</a></p>
</div>
</div>
<div style="text-align:center; margin-top: 0em; margin-bottom: 0em"><p style="margin-top: 0.25em; margin-bottom: 0em;">TheBloke's LLM work is generously supported by a grant from <a href="https://a16z.com">andreessen horowitz (a16z)</a></p></div>
<hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
<!-- header end -->
# Airoboros 13B GPTQ 4bit
These files are GPTQ 4bit model files for [Jon Durbin's Airoboros 13B](https://huggingface.co/jondurbin/airoboros-13b).
It is the result of quantising to 4bit using [GPTQ-for-LLaMa](https://github.com/qwopqwop200/GPTQ-for-LLaMa).
## Other repositories available
* [4bit GPTQ models for GPU inference](https://huggingface.co/TheBloke/airoboros-13B-GPTQ)
* [Unquantised model in HF fp16 format](https://huggingface.co/TheBloke/airoboros-13B-HF)
* [latimar's GGML models for CPU (+CUDA) inference](https://huggingface.co/latimar/airoboros-13b-ggml)
## How to easily download and use this model in text-generation-webui
Open the text-generation-webui UI as normal.
1. Click the **Model tab**.
2. Under **Download custom model or LoRA**, enter `TheBloke/Airoboros-13B-GPTQ`.
3. Click **Download**.
4. Wait until it says it's finished downloading.
5. Click the **Refresh** icon next to **Model** in the top left.
6. In the **Model drop-down**: choose the model you just downloaded, `Airoboros-13B-GPTQ`.
7. If you see an error in the bottom right, ignore it - it's temporary.
8. Fill out the `GPTQ parameters` on the right: `Bits = 4`, `Groupsize = 128`, `model_type = Llama`
9. Click **Save settings for this model** in the top right.
10. Click **Reload the Model** in the top right.
11. Once it says it's loaded, click the **Text Generation tab** and enter a prompt!
## Provided files
**Compatible file - Airoboros-13B-GPTQ-4bit-128g.no-act-order.safetensors**
In the `main` branch - the default one - you will find `Airoboros-13B-GPTQ-4bit-128g.no-act-order.safetensors`
This will work with all versions of GPTQ-for-LLaMa. It has maximum compatibility.
It was created without the `--act-order` parameter to ensure full compatibility.
* `wizard-vicuna-13B-GPTQ-4bit.compat.no-act-order.safetensors`
* Works with all versions of GPTQ-for-LLaMa code, both Triton and CUDA branches
* Works with AutoGPTQ.
* Works with text-generation-webui one-click-installers
* Parameters: Groupsize = 128. No act-order.
* Command used to create the GPTQ:
```
python llama.py /workspace/models/jondurbin_airoboros-13b wikitext2 --wbits 4 --true-sequential --groupsize 128 --save_safetensors /workspace/jon-13b/gptq/Airoboros-13B-GPTQ-4bit-128g.no-act-order.safetensors
```
<!-- footer start -->
<!-- 200823 -->
## Discord
For further support, and discussions on these models and AI in general, join us at:
[TheBloke AI's Discord server](https://discord.gg/theblokeai)
## Thanks, and how to contribute.
Thanks to the [chirper.ai](https://chirper.ai) team!
I've had a lot of people ask if they can contribute. I enjoy providing models and helping people, and would love to be able to spend even more time doing it, as well as expanding into new projects like fine tuning/training.
If you're able and willing to contribute it will be most gratefully received and will help me to keep providing more models, and to start work on new AI projects.
Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits.
* Patreon: https://patreon.com/TheBlokeAI
* Ko-Fi: https://ko-fi.com/TheBlokeAI
**Special thanks to**: Aemon Algiz.
**Patreon special mentions**: Sam, theTransient, Jonathan Leane, Steven Wood, webtim, Johann-Peter Hartmann, Geoffrey Montalvo, Gabriel Tamborski, Willem Michiel, John Villwock, Derek Yates, Mesiah Bishop, Eugene Pentland, Pieter, Chadd, Stephen Murray, Daniel P. Andersen, terasurfer, Brandon Frisco, Thomas Belote, Sid, Nathan LeClaire, Magnesian, Alps Aficionado, Stanislav Ovsiannikov, Alex, Joseph William Delisle, Nikolai Manek, Michael Davis, Junyu Yang, K, J, Spencer Kim, Stefan Sabev, Olusegun Samson, transmissions 11, Michael Levine, Cory Kujawski, Rainer Wilmers, zynix, Kalila, Luke @flexchar, Ajan Kanaga, Mandus, vamX, Ai Maven, Mano Prime, Matthew Berman, subjectnull, Vitor Caleffi, Clay Pascal, biorpg, alfie_i, 阿明, Jeffrey Morgan, ya boyyy, Raymond Fosdick, knownsqashed, Olakabola, Leonard Tan, ReadyPlayerEmma, Enrico Ros, Dave, Talal Aujan, Illia Dulskyi, Sean Connelly, senxiiz, Artur Olbinski, Elle, Raven Klaugh, Fen Risland, Deep Realms, Imad Khwaja, Fred von Graf, Will Dee, usrbinkat, SuperWojo, Alexandros Triantafyllidis, Swaroop Kallakuri, Dan Guido, John Detwiler, Pedro Madruga, Iucharbius, Viktor Bowallius, Asp the Wyvern, Edmond Seymore, Trenton Dambrowitz, Space Cruiser, Spiking Neurons AB, Pyrater, LangChain4j, Tony Hughes, Kacper Wikieł, Rishabh Srivastava, David Ziegler, Luke Pendergrass, Andrey, Gabriel Puliatti, Lone Striker, Sebastain Graf, Pierre Kircher, Randy H, NimbleBox.ai, Vadim, danny, Deo Leter
Thank you to all my generous patrons and donaters!
And thank you again to a16z for their generous grant.
<!-- footer end -->
# Airoboros-13B original model card
# Overview
This is a fine-tuned 13b parameter LlaMa model, using completely synthetic training data created by https://github.com/jondurbin/airoboros
### Eval (gpt4 judging)
| model | raw score | gpt-3.5 adjusted score |
| --- | --- | --- |
| __airoboros-13b__ | __17947__ | __98.087__ |
| gpt35 | 18297 | 100.0 |
| gpt4-x-alpasta-30b | 15612 | 85.33 |
| manticore-13b | 15856 | 86.66 |
| vicuna-13b-1.1 | 16306 | 89.12 |
| wizard-vicuna-13b-uncensored | 16287 | 89.01 |
<details>
<summary>individual question scores, with shareGPT links (200 prompts generated by gpt-4)</summary>
*wb-13b-u is Wizard-Vicuna-13b-Uncensored*
| airoboros-13b | gpt35 | gpt4-x-alpasta-30b | manticore-13b | vicuna-13b-1.1 | wv-13b-u | link |
|----------------:|--------:|---------------------:|----------------:|-----------------:|-------------------------------:|:---------------------------------------|
| 80 | 95 | 70 | 90 | 85 | 60 | [eval](https://sharegpt.com/c/PIbRQD3) |
| 20 | 95 | 40 | 30 | 90 | 80 | [eval](https://sharegpt.com/c/fSzwzzd) |
| 100 | 100 | 100 | 95 | 95 | 100 | [eval](https://sharegpt.com/c/AXMzZiO) |
| 90 | 100 | 85 | 60 | 95 | 100 | [eval](https://sharegpt.com/c/7obzJm2) |
| 95 | 90 | 80 | 85 | 95 | 75 | [eval](https://sharegpt.com/c/cRpj6M1) |
| 100 | 95 | 90 | 95 | 98 | 92 | [eval](https://sharegpt.com/c/p0by1T7) |
| 50 | 100 | 80 | 95 | 60 | 55 | [eval](https://sharegpt.com/c/rowNlKx) |
| 70 | 90 | 80 | 60 | 85 | 40 | [eval](https://sharegpt.com/c/I4POj4I) |
| 100 | 95 | 50 | 85 | 40 | 60 | [eval](https://sharegpt.com/c/gUAeiRp) |
| 85 | 60 | 55 | 65 | 50 | 70 | [eval](https://sharegpt.com/c/Lgw4QQL) |
| 95 | 100 | 85 | 90 | 60 | 75 | [eval](https://sharegpt.com/c/X9tDYft) |
| 100 | 95 | 70 | 80 | 50 | 85 | [eval](https://sharegpt.com/c/9V2ElkH) |
| 100 | 95 | 80 | 70 | 60 | 90 | [eval](https://sharegpt.com/c/D5xg6qt) |
| 95 | 100 | 70 | 85 | 90 | 90 | [eval](https://sharegpt.com/c/lQnSfDs) |
| 80 | 95 | 90 | 60 | 30 | 85 | [eval](https://sharegpt.com/c/1hpHGNc) |
| 60 | 95 | 0 | 75 | 50 | 40 | [eval](https://sharegpt.com/c/an6TqE4) |
| 100 | 95 | 90 | 98 | 95 | 95 | [eval](https://sharegpt.com/c/7vr6n3F) |
| 60 | 85 | 40 | 50 | 20 | 0 | [eval](https://sharegpt.com/c/TOkMkgE) |
| 100 | 90 | 85 | 95 | 95 | 80 | [eval](https://sharegpt.com/c/Qu7ak0r) |
| 100 | 95 | 100 | 95 | 90 | 95 | [eval](https://sharegpt.com/c/hMD4gPo) |
| 95 | 90 | 96 | 80 | 92 | 88 | [eval](https://sharegpt.com/c/HTlicNh) |
| 95 | 92 | 90 | 93 | 89 | 91 | [eval](https://sharegpt.com/c/MjxHpAf) |
| 95 | 93 | 90 | 94 | 96 | 92 | [eval](https://sharegpt.com/c/4RvxOR9) |
| 95 | 90 | 93 | 88 | 92 | 85 | [eval](https://sharegpt.com/c/PcAIU9r) |
| 95 | 90 | 85 | 96 | 88 | 92 | [eval](https://sharegpt.com/c/MMqul3q) |
| 95 | 95 | 90 | 93 | 92 | 91 | [eval](https://sharegpt.com/c/YQsLyzJ) |
| 95 | 98 | 80 | 97 | 99 | 96 | [eval](https://sharegpt.com/c/UDhSTMq) |
| 95 | 93 | 90 | 87 | 92 | 89 | [eval](https://sharegpt.com/c/4gCfdCV) |
| 90 | 85 | 95 | 80 | 92 | 75 | [eval](https://sharegpt.com/c/bkQs4SP) |
| 90 | 85 | 95 | 93 | 80 | 92 | [eval](https://sharegpt.com/c/LeLCEEt) |
| 95 | 92 | 90 | 91 | 93 | 89 | [eval](https://sharegpt.com/c/DFxNzVu) |
| 100 | 95 | 90 | 85 | 80 | 95 | [eval](https://sharegpt.com/c/gnVzNML) |
| 95 | 97 | 93 | 92 | 96 | 94 | [eval](https://sharegpt.com/c/y7pxMIy) |
| 95 | 93 | 94 | 90 | 88 | 92 | [eval](https://sharegpt.com/c/5UeCvTY) |
| 90 | 95 | 98 | 85 | 96 | 92 | [eval](https://sharegpt.com/c/T4oL9I5) |
| 90 | 88 | 85 | 80 | 82 | 84 | [eval](https://sharegpt.com/c/HnGyTAG) |
| 90 | 95 | 85 | 87 | 92 | 88 | [eval](https://sharegpt.com/c/ZbRMBNj) |
| 95 | 97 | 96 | 90 | 93 | 92 | [eval](https://sharegpt.com/c/iTmFJqd) |
| 95 | 93 | 92 | 90 | 89 | 91 | [eval](https://sharegpt.com/c/VuPifET) |
| 90 | 95 | 93 | 92 | 94 | 91 | [eval](https://sharegpt.com/c/AvFAH1x) |
| 90 | 85 | 95 | 80 | 88 | 75 | [eval](https://sharegpt.com/c/4ealKGN) |
| 85 | 90 | 95 | 88 | 92 | 80 | [eval](https://sharegpt.com/c/bE1b2vX) |
| 90 | 95 | 92 | 85 | 80 | 87 | [eval](https://sharegpt.com/c/I3nMPBC) |
| 85 | 90 | 95 | 80 | 88 | 75 | [eval](https://sharegpt.com/c/as7r3bW) |
| 85 | 80 | 75 | 90 | 70 | 82 | [eval](https://sharegpt.com/c/qYceaUa) |
| 90 | 85 | 95 | 92 | 93 | 80 | [eval](https://sharegpt.com/c/g4FXchU) |
| 90 | 95 | 75 | 85 | 80 | 70 | [eval](https://sharegpt.com/c/6kGLvL5) |
| 85 | 90 | 80 | 88 | 82 | 83 | [eval](https://sharegpt.com/c/SRozqaF) |
| 85 | 90 | 95 | 92 | 88 | 80 | [eval](https://sharegpt.com/c/GoKydf6) |
| 85 | 90 | 80 | 75 | 95 | 88 | [eval](https://sharegpt.com/c/37aXkHQ) |
| 85 | 90 | 80 | 88 | 84 | 92 | [eval](https://sharegpt.com/c/nVuUaTj) |
| 80 | 90 | 75 | 85 | 70 | 95 | [eval](https://sharegpt.com/c/TkAQKLC) |
| 90 | 88 | 85 | 80 | 92 | 83 | [eval](https://sharegpt.com/c/55cO2y0) |
| 85 | 75 | 90 | 80 | 78 | 88 | [eval](https://sharegpt.com/c/tXtq5lT) |
| 85 | 90 | 80 | 82 | 75 | 88 | [eval](https://sharegpt.com/c/TfMjeJQ) |
| 90 | 85 | 40 | 95 | 80 | 88 | [eval](https://sharegpt.com/c/2jQ6K2S) |
| 85 | 95 | 90 | 75 | 88 | 80 | [eval](https://sharegpt.com/c/aQtr2ca) |
| 85 | 95 | 90 | 92 | 89 | 88 | [eval](https://sharegpt.com/c/tbWLyZ7) |
| 80 | 85 | 75 | 60 | 90 | 70 | [eval](https://sharegpt.com/c/moHC7i2) |
| 85 | 90 | 87 | 80 | 88 | 75 | [eval](https://sharegpt.com/c/GK6GShh) |
| 85 | 80 | 75 | 50 | 90 | 80 | [eval](https://sharegpt.com/c/ugcW4qG) |
| 95 | 80 | 90 | 85 | 75 | 82 | [eval](https://sharegpt.com/c/WL8iq6F) |
| 85 | 90 | 80 | 70 | 95 | 88 | [eval](https://sharegpt.com/c/TZJKnvS) |
| 90 | 95 | 70 | 85 | 80 | 75 | [eval](https://sharegpt.com/c/beNOKb5) |
| 90 | 85 | 70 | 75 | 80 | 60 | [eval](https://sharegpt.com/c/o2oRCF5) |
| 95 | 90 | 70 | 50 | 85 | 80 | [eval](https://sharegpt.com/c/TNjbK6D) |
| 80 | 85 | 40 | 60 | 90 | 95 | [eval](https://sharegpt.com/c/rJvszWJ) |
| 75 | 60 | 80 | 55 | 70 | 85 | [eval](https://sharegpt.com/c/HJwRkro) |
| 90 | 85 | 60 | 50 | 80 | 95 | [eval](https://sharegpt.com/c/AeFoSDK) |
| 45 | 85 | 60 | 20 | 65 | 75 | [eval](https://sharegpt.com/c/KA1cgOl) |
| 85 | 90 | 30 | 60 | 80 | 70 | [eval](https://sharegpt.com/c/RTy8n0y) |
| 90 | 95 | 80 | 40 | 85 | 70 | [eval](https://sharegpt.com/c/PJMJoXh) |
| 85 | 90 | 70 | 75 | 80 | 95 | [eval](https://sharegpt.com/c/Ib3jzyC) |
| 90 | 70 | 50 | 20 | 60 | 40 | [eval](https://sharegpt.com/c/oMmqqtX) |
| 90 | 95 | 75 | 60 | 85 | 80 | [eval](https://sharegpt.com/c/qRNhNTw) |
| 85 | 80 | 60 | 70 | 65 | 75 | [eval](https://sharegpt.com/c/3MAHQIy) |
| 90 | 85 | 80 | 75 | 82 | 70 | [eval](https://sharegpt.com/c/0Emc5HS) |
| 90 | 95 | 80 | 70 | 85 | 75 | [eval](https://sharegpt.com/c/UqAxRWF) |
| 85 | 75 | 30 | 80 | 90 | 70 | [eval](https://sharegpt.com/c/eywxGAw) |
| 85 | 90 | 50 | 70 | 80 | 60 | [eval](https://sharegpt.com/c/A2KSEWP) |
| 100 | 95 | 98 | 99 | 97 | 96 | [eval](https://sharegpt.com/c/C8rebQf) |
| 95 | 90 | 92 | 93 | 91 | 89 | [eval](https://sharegpt.com/c/cd9HF4V) |
| 95 | 92 | 90 | 85 | 88 | 91 | [eval](https://sharegpt.com/c/LHkjvQJ) |
| 100 | 95 | 98 | 97 | 96 | 99 | [eval](https://sharegpt.com/c/o5PdoyZ) |
| 100 | 100 | 100 | 90 | 100 | 95 | [eval](https://sharegpt.com/c/rh8pZVg) |
| 100 | 95 | 98 | 97 | 94 | 99 | [eval](https://sharegpt.com/c/T5DYL83) |
| 95 | 90 | 92 | 93 | 94 | 91 | [eval](https://sharegpt.com/c/G5Osg3X) |
| 100 | 95 | 98 | 90 | 96 | 95 | [eval](https://sharegpt.com/c/9ZqI03V) |
| 95 | 96 | 92 | 90 | 89 | 93 | [eval](https://sharegpt.com/c/4tFfwZU) |
| 100 | 95 | 93 | 90 | 92 | 88 | [eval](https://sharegpt.com/c/mG1JqPH) |
| 100 | 100 | 98 | 97 | 99 | 100 | [eval](https://sharegpt.com/c/VDdtgCu) |
| 95 | 90 | 92 | 85 | 93 | 94 | [eval](https://sharegpt.com/c/uKtGkvg) |
| 95 | 93 | 90 | 92 | 96 | 91 | [eval](https://sharegpt.com/c/9B92N6P) |
| 95 | 96 | 92 | 90 | 93 | 91 | [eval](https://sharegpt.com/c/GeIFfOu) |
| 95 | 90 | 92 | 93 | 91 | 89 | [eval](https://sharegpt.com/c/gn3E9nN) |
| 100 | 98 | 95 | 97 | 96 | 99 | [eval](https://sharegpt.com/c/Erxa46H) |
| 90 | 95 | 85 | 88 | 92 | 87 | [eval](https://sharegpt.com/c/oRHVOvK) |
| 95 | 93 | 90 | 92 | 89 | 88 | [eval](https://sharegpt.com/c/ghtKLUX) |
| 100 | 95 | 97 | 90 | 96 | 94 | [eval](https://sharegpt.com/c/ZL4KjqP) |
| 95 | 93 | 90 | 92 | 94 | 91 | [eval](https://sharegpt.com/c/YOnqIQa) |
| 95 | 92 | 90 | 93 | 94 | 88 | [eval](https://sharegpt.com/c/3BKwKho) |
| 95 | 92 | 60 | 97 | 90 | 96 | [eval](https://sharegpt.com/c/U1i31bn) |
| 95 | 90 | 92 | 93 | 91 | 89 | [eval](https://sharegpt.com/c/etfRoAE) |
| 95 | 90 | 97 | 92 | 91 | 93 | [eval](https://sharegpt.com/c/B0OpVxR) |
| 90 | 95 | 93 | 85 | 92 | 91 | [eval](https://sharegpt.com/c/MBgGJ5A) |
| 95 | 90 | 40 | 92 | 93 | 85 | [eval](https://sharegpt.com/c/eQKTYO7) |
| 100 | 100 | 95 | 90 | 95 | 90 | [eval](https://sharegpt.com/c/szKWCBt) |
| 90 | 95 | 96 | 98 | 93 | 92 | [eval](https://sharegpt.com/c/8ZhUcAv) |
| 90 | 95 | 92 | 89 | 93 | 94 | [eval](https://sharegpt.com/c/VQWdy99) |
| 100 | 95 | 100 | 98 | 96 | 99 | [eval](https://sharegpt.com/c/g1DHUSM) |
| 100 | 100 | 95 | 90 | 100 | 90 | [eval](https://sharegpt.com/c/uYgfJC3) |
| 90 | 85 | 88 | 92 | 87 | 91 | [eval](https://sharegpt.com/c/crk8BH3) |
| 95 | 97 | 90 | 92 | 93 | 94 | [eval](https://sharegpt.com/c/95F9afQ) |
| 90 | 95 | 85 | 88 | 92 | 89 | [eval](https://sharegpt.com/c/otioHUo) |
| 95 | 93 | 90 | 92 | 94 | 91 | [eval](https://sharegpt.com/c/KSiL9F6) |
| 90 | 95 | 85 | 80 | 88 | 82 | [eval](https://sharegpt.com/c/GmGq3b3) |
| 95 | 90 | 60 | 85 | 93 | 70 | [eval](https://sharegpt.com/c/VOhklyz) |
| 95 | 92 | 94 | 93 | 96 | 90 | [eval](https://sharegpt.com/c/wqy8m6k) |
| 95 | 90 | 85 | 93 | 87 | 92 | [eval](https://sharegpt.com/c/iWKrIuS) |
| 95 | 96 | 93 | 90 | 97 | 92 | [eval](https://sharegpt.com/c/o1h3w8N) |
| 100 | 0 | 0 | 100 | 0 | 0 | [eval](https://sharegpt.com/c/3UH9eed) |
| 60 | 100 | 0 | 80 | 0 | 0 | [eval](https://sharegpt.com/c/44g0FAh) |
| 0 | 100 | 60 | 0 | 0 | 90 | [eval](https://sharegpt.com/c/PaQlcrU) |
| 100 | 100 | 0 | 100 | 100 | 100 | [eval](https://sharegpt.com/c/51icV4o) |
| 100 | 100 | 100 | 100 | 95 | 100 | [eval](https://sharegpt.com/c/1VnbGAR) |
| 100 | 100 | 100 | 50 | 90 | 100 | [eval](https://sharegpt.com/c/EYGBrgw) |
| 100 | 100 | 100 | 100 | 95 | 90 | [eval](https://sharegpt.com/c/EGRduOt) |
| 100 | 100 | 100 | 95 | 0 | 100 | [eval](https://sharegpt.com/c/O3JJfnK) |
| 50 | 95 | 20 | 10 | 30 | 85 | [eval](https://sharegpt.com/c/2roVtAu) |
| 100 | 100 | 60 | 20 | 30 | 40 | [eval](https://sharegpt.com/c/sphFpfx) |
| 100 | 0 | 0 | 0 | 0 | 100 | [eval](https://sharegpt.com/c/OeWGKBo) |
| 0 | 100 | 60 | 0 | 0 | 80 | [eval](https://sharegpt.com/c/TOUsuFA) |
| 50 | 100 | 20 | 90 | 0 | 10 | [eval](https://sharegpt.com/c/Y3P6DCu) |
| 100 | 100 | 100 | 100 | 100 | 100 | [eval](https://sharegpt.com/c/hkbdeiM) |
| 100 | 100 | 100 | 100 | 100 | 100 | [eval](https://sharegpt.com/c/eubbaVC) |
| 40 | 100 | 95 | 0 | 100 | 40 | [eval](https://sharegpt.com/c/QWiF49v) |
| 100 | 100 | 100 | 100 | 80 | 100 | [eval](https://sharegpt.com/c/dKTapBu) |
| 100 | 100 | 100 | 0 | 90 | 40 | [eval](https://sharegpt.com/c/P8NGwFZ) |
| 0 | 100 | 100 | 50 | 70 | 20 | [eval](https://sharegpt.com/c/v96BtBL) |
| 100 | 100 | 50 | 90 | 0 | 95 | [eval](https://sharegpt.com/c/YRlzj1t) |
| 100 | 95 | 90 | 85 | 98 | 80 | [eval](https://sharegpt.com/c/76VX3eB) |
| 95 | 98 | 90 | 92 | 96 | 89 | [eval](https://sharegpt.com/c/JK1uNef) |
| 90 | 95 | 75 | 85 | 80 | 82 | [eval](https://sharegpt.com/c/ku6CKmx) |
| 95 | 98 | 50 | 92 | 96 | 94 | [eval](https://sharegpt.com/c/0iAFuKW) |
| 95 | 90 | 0 | 93 | 92 | 94 | [eval](https://sharegpt.com/c/6uGnKio) |
| 95 | 90 | 85 | 92 | 80 | 88 | [eval](https://sharegpt.com/c/lfpRBw8) |
| 95 | 93 | 75 | 85 | 90 | 92 | [eval](https://sharegpt.com/c/mKu70jb) |
| 90 | 95 | 88 | 85 | 92 | 89 | [eval](https://sharegpt.com/c/GkYzJHO) |
| 100 | 100 | 100 | 95 | 97 | 98 | [eval](https://sharegpt.com/c/mly2k0z) |
| 85 | 40 | 30 | 95 | 90 | 88 | [eval](https://sharegpt.com/c/5td2ob0) |
| 90 | 95 | 92 | 85 | 88 | 93 | [eval](https://sharegpt.com/c/0ISpWfy) |
| 95 | 96 | 92 | 90 | 89 | 93 | [eval](https://sharegpt.com/c/kdUDUn7) |
| 90 | 95 | 85 | 80 | 92 | 88 | [eval](https://sharegpt.com/c/fjMNYr2) |
| 95 | 98 | 65 | 90 | 85 | 93 | [eval](https://sharegpt.com/c/6xBIf2Q) |
| 95 | 92 | 96 | 97 | 90 | 89 | [eval](https://sharegpt.com/c/B9GY8Ln) |
| 95 | 90 | 92 | 91 | 89 | 93 | [eval](https://sharegpt.com/c/vn1FPU4) |
| 95 | 90 | 80 | 75 | 95 | 90 | [eval](https://sharegpt.com/c/YurEMYg) |
| 92 | 40 | 30 | 95 | 90 | 93 | [eval](https://sharegpt.com/c/D19Qeui) |
| 90 | 92 | 85 | 88 | 89 | 87 | [eval](https://sharegpt.com/c/5QRFfrt) |
| 95 | 80 | 90 | 92 | 91 | 88 | [eval](https://sharegpt.com/c/pYWPRi4) |
| 95 | 93 | 92 | 90 | 91 | 94 | [eval](https://sharegpt.com/c/wPRTntL) |
| 100 | 98 | 95 | 90 | 92 | 96 | [eval](https://sharegpt.com/c/F6PLYKE) |
| 95 | 92 | 80 | 85 | 90 | 93 | [eval](https://sharegpt.com/c/WeJnMGv) |
| 95 | 98 | 90 | 88 | 97 | 96 | [eval](https://sharegpt.com/c/zNKL49e) |
| 90 | 95 | 85 | 88 | 86 | 92 | [eval](https://sharegpt.com/c/kIKmA1b) |
| 100 | 100 | 100 | 100 | 100 | 100 | [eval](https://sharegpt.com/c/1btWd4O) |
| 90 | 95 | 85 | 96 | 92 | 88 | [eval](https://sharegpt.com/c/s9sf1Lp) |
| 100 | 98 | 95 | 99 | 97 | 96 | [eval](https://sharegpt.com/c/RWzv8py) |
| 95 | 92 | 70 | 90 | 93 | 89 | [eval](https://sharegpt.com/c/bYF7FqA) |
| 95 | 90 | 88 | 92 | 94 | 93 | [eval](https://sharegpt.com/c/SuUqjMj) |
| 95 | 90 | 93 | 92 | 85 | 94 | [eval](https://sharegpt.com/c/r0aRdYY) |
| 95 | 93 | 90 | 87 | 92 | 91 | [eval](https://sharegpt.com/c/VuMfkkd) |
| 95 | 93 | 90 | 96 | 92 | 91 | [eval](https://sharegpt.com/c/rhm6fa4) |
| 95 | 97 | 85 | 96 | 98 | 90 | [eval](https://sharegpt.com/c/DwXnyqG) |
| 95 | 92 | 90 | 85 | 93 | 94 | [eval](https://sharegpt.com/c/0ScdkGS) |
| 95 | 96 | 92 | 90 | 97 | 93 | [eval](https://sharegpt.com/c/6yIoCDU) |
| 95 | 93 | 96 | 94 | 90 | 92 | [eval](https://sharegpt.com/c/VubEvp9) |
| 95 | 94 | 93 | 92 | 90 | 89 | [eval](https://sharegpt.com/c/RHzmZWG) |
| 90 | 85 | 95 | 80 | 87 | 75 | [eval](https://sharegpt.com/c/IMiP9Zm) |
| 95 | 94 | 92 | 93 | 90 | 96 | [eval](https://sharegpt.com/c/bft4PIL) |
| 95 | 100 | 90 | 95 | 95 | 95 | [eval](https://sharegpt.com/c/iHXB34b) |
| 100 | 95 | 85 | 100 | 0 | 90 | [eval](https://sharegpt.com/c/vCGn9R7) |
| 100 | 95 | 90 | 95 | 100 | 95 | [eval](https://sharegpt.com/c/be8crZL) |
| 95 | 90 | 60 | 95 | 85 | 80 | [eval](https://sharegpt.com/c/33elmDz) |
| 100 | 95 | 90 | 98 | 97 | 99 | [eval](https://sharegpt.com/c/RWD3Zx7) |
| 95 | 90 | 85 | 95 | 80 | 92 | [eval](https://sharegpt.com/c/GiwBvM7) |
| 100 | 95 | 100 | 98 | 100 | 90 | [eval](https://sharegpt.com/c/hX2pYxk) |
| 100 | 95 | 80 | 85 | 90 | 85 | [eval](https://sharegpt.com/c/MfxdGd7) |
| 100 | 90 | 95 | 85 | 95 | 100 | [eval](https://sharegpt.com/c/28hQjmS) |
| 95 | 90 | 85 | 80 | 88 | 92 | [eval](https://sharegpt.com/c/fzy5EPe) |
| 100 | 100 | 0 | 0 | 100 | 0 | [eval](https://sharegpt.com/c/vwxPjbR) |
| 100 | 100 | 100 | 50 | 100 | 75 | [eval](https://sharegpt.com/c/FAYfFWy) |
| 100 | 100 | 0 | 0 | 100 | 0 | [eval](https://sharegpt.com/c/SoudGsQ) |
| 0 | 100 | 0 | 0 | 0 | 0 | [eval](https://sharegpt.com/c/mkwEgVn) |
| 100 | 100 | 50 | 0 | 0 | 0 | [eval](https://sharegpt.com/c/q8MQEsz) |
| 100 | 100 | 100 | 100 | 100 | 95 | [eval](https://sharegpt.com/c/tzHpsKh) |
| 100 | 100 | 50 | 0 | 0 | 0 | [eval](https://sharegpt.com/c/3ugYBtJ) |
| 100 | 100 | 0 | 0 | 100 | 0 | [eval](https://sharegpt.com/c/I6KfOJT) |
| 90 | 85 | 80 | 95 | 70 | 75 | [eval](https://sharegpt.com/c/enaV1CK) |
| 100 | 100 | 0 | 0 | 0 | 0 | [eval](https://sharegpt.com/c/JBk7oSh) |
</details>
### Training data
I used a jailbreak prompt to generate the synthetic instructions, which resulted in some training data that would likely be censored by other models, such as how-to prompts about synthesizing drugs, making homemade flamethrowers, etc. Mind you, this is all generated by ChatGPT, not me. My goal was to simply test some of the capabilities of ChatGPT when unfiltered (as much as possible), and not to intentionally produce any harmful/dangerous/etc. content.
The jailbreak prompt I used is the default prompt in the python code when using the `--uncensored` flag: https://github.com/jondurbin/airoboros/blob/main/airoboros/self_instruct.py#L39
I also did a few passes of manually cleanup to remove some bad prompts, but mostly I left the data as-is. Initially, the model was fairly bad at math/extrapolation, closed question-answering (heavy hallucination), and coding, so I did one more fine tuning pass with additional synthetic instructions aimed at those types of problems.
Both the initial instructions and final-pass fine-tuning instructions will be published soon.
### Fine-tuning method
I used the excellent [FastChat](https://github.com/lm-sys/FastChat) module, running with:
```
source /workspace/venv/bin/activate
export NCCL_P2P_DISABLE=1
export NCCL_P2P_LEVEL=LOC
torchrun --nproc_per_node=8 --master_port=20001 /workspace/FastChat/fastchat/train/train_mem.py \
--model_name_or_path /workspace/llama-13b \
--data_path /workspace/as_conversations.json \
--bf16 True \
--output_dir /workspace/airoboros-uncensored-13b \
--num_train_epochs 3 \
--per_device_train_batch_size 20 \
--per_device_eval_batch_size 20 \
--gradient_accumulation_steps 2 \
--evaluation_strategy "steps" \
--eval_steps 500 \
--save_strategy "steps" \
--save_steps 500 \
--save_total_limit 10 \
--learning_rate 2e-5 \
--weight_decay 0. \
--warmup_ratio 0.04 \
--lr_scheduler_type "cosine" \
--logging_steps 1 \
--fsdp "full_shard auto_wrap offload" \
--fsdp_transformer_layer_cls_to_wrap 'LlamaDecoderLayer' \
--tf32 True \
--model_max_length 2048 \
--gradient_checkpointing True \
--lazy_preprocess True
```
This ran on 8x nvidia 80gb a100's for about 40 hours.
### Prompt format
The prompt should be 1:1 compatible with the FastChat/vicuna format, e.g.:
With a preamble:
```
A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions.
USER: [prompt]
<\s>
ASSISTANT:
```
Or just:
```
USER: [prompt]
<\s>
ASSISTANT:
```
### License
The model is licensed under the LLaMA model, and the dataset is licensed under the terms of OpenAI because it uses ChatGPT. Everything else is free.
|