TheBloke commited on
Commit
6162a27
1 Parent(s): 6c21ccf

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +463 -0
README.md ADDED
@@ -0,0 +1,463 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: CalderaAI/Naberius-7B
3
+ inference: false
4
+ language:
5
+ - en
6
+ license: apache-2.0
7
+ model_creator: Caldera AI
8
+ model_name: Naberius 7B
9
+ model_type: mistral
10
+ prompt_template: '<|im_start|>system
11
+
12
+ {system_message}<|im_end|>
13
+
14
+ <|im_start|>user
15
+
16
+ {prompt}<|im_end|>
17
+
18
+ <|im_start|>assistant
19
+
20
+ '
21
+ quantized_by: TheBloke
22
+ tags:
23
+ - llama
24
+ - uncensored
25
+ - merge
26
+ - mix
27
+ - slerp
28
+ - spherical linear interpolation merge
29
+ - mistral
30
+ - hermes
31
+ - openhermes
32
+ - dolphin
33
+ - zephyr
34
+ - naberius
35
+ - 7b
36
+ - llama2
37
+ ---
38
+ <!-- markdownlint-disable MD041 -->
39
+
40
+ <!-- header start -->
41
+ <!-- 200823 -->
42
+ <div style="width: auto; margin-left: auto; margin-right: auto">
43
+ <img src="https://i.imgur.com/EBdldam.jpg" alt="TheBlokeAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
44
+ </div>
45
+ <div style="display: flex; justify-content: space-between; width: 100%;">
46
+ <div style="display: flex; flex-direction: column; align-items: flex-start;">
47
+ <p style="margin-top: 0.5em; margin-bottom: 0em;"><a href="https://discord.gg/theblokeai">Chat & support: TheBloke's Discord server</a></p>
48
+ </div>
49
+ <div style="display: flex; flex-direction: column; align-items: flex-end;">
50
+ <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>
51
+ </div>
52
+ </div>
53
+ <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>
54
+ <hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
55
+ <!-- header end -->
56
+
57
+ # Naberius 7B - GPTQ
58
+ - Model creator: [Caldera AI](https://huggingface.co/CalderaAI)
59
+ - Original model: [Naberius 7B](https://huggingface.co/CalderaAI/Naberius-7B)
60
+
61
+ <!-- description start -->
62
+ ## Description
63
+
64
+ This repo contains GPTQ model files for [Caldera AI's Naberius 7B](https://huggingface.co/CalderaAI/Naberius-7B).
65
+
66
+ Multiple GPTQ parameter permutations are provided; see Provided Files below for details of the options provided, their parameters, and the software used to create them.
67
+
68
+ These files were quantised using hardware kindly provided by [Massed Compute](https://massedcompute.com/).
69
+
70
+ <!-- description end -->
71
+ <!-- repositories-available start -->
72
+ ## Repositories available
73
+
74
+ * [AWQ model(s) for GPU inference.](https://huggingface.co/TheBloke/Naberius-7B-AWQ)
75
+ * [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/Naberius-7B-GPTQ)
76
+ * [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/Naberius-7B-GGUF)
77
+ * [Caldera AI's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/CalderaAI/Naberius-7B)
78
+ <!-- repositories-available end -->
79
+
80
+ <!-- prompt-template start -->
81
+ ## Prompt template: ChatML
82
+
83
+ ```
84
+ <|im_start|>system
85
+ {system_message}<|im_end|>
86
+ <|im_start|>user
87
+ {prompt}<|im_end|>
88
+ <|im_start|>assistant
89
+
90
+ ```
91
+
92
+ <!-- prompt-template end -->
93
+
94
+
95
+
96
+ <!-- README_GPTQ.md-compatible clients start -->
97
+ ## Known compatible clients / servers
98
+
99
+ These GPTQ models are known to work in the following inference servers/webuis.
100
+
101
+ - [text-generation-webui](https://github.com/oobabooga/text-generation-webui)
102
+ - [KoboldAI United](https://github.com/henk717/koboldai)
103
+ - [LoLLMS Web UI](https://github.com/ParisNeo/lollms-webui)
104
+ - [Hugging Face Text Generation Inference (TGI)](https://github.com/huggingface/text-generation-inference)
105
+
106
+ This may not be a complete list; if you know of others, please let me know!
107
+ <!-- README_GPTQ.md-compatible clients end -->
108
+
109
+ <!-- README_GPTQ.md-provided-files start -->
110
+ ## Provided files, and GPTQ parameters
111
+
112
+ Multiple quantisation parameters are provided, to allow you to choose the best one for your hardware and requirements.
113
+
114
+ Each separate quant is in a different branch. See below for instructions on fetching from different branches.
115
+
116
+ Most GPTQ files are made with AutoGPTQ. Mistral models are currently made with Transformers.
117
+
118
+ <details>
119
+ <summary>Explanation of GPTQ parameters</summary>
120
+
121
+ - Bits: The bit size of the quantised model.
122
+ - GS: GPTQ group size. Higher numbers use less VRAM, but have lower quantisation accuracy. "None" is the lowest possible value.
123
+ - Act Order: True or False. Also known as `desc_act`. True results in better quantisation accuracy. Some GPTQ clients have had issues with models that use Act Order plus Group Size, but this is generally resolved now.
124
+ - Damp %: A GPTQ parameter that affects how samples are processed for quantisation. 0.01 is default, but 0.1 results in slightly better accuracy.
125
+ - GPTQ dataset: The calibration dataset used during quantisation. Using a dataset more appropriate to the model's training can improve quantisation accuracy. Note that the GPTQ calibration dataset is not the same as the dataset used to train the model - please refer to the original model repo for details of the training dataset(s).
126
+ - Sequence Length: The length of the dataset sequences used for quantisation. Ideally this is the same as the model sequence length. For some very long sequence models (16+K), a lower sequence length may have to be used. Note that a lower sequence length does not limit the sequence length of the quantised model. It only impacts the quantisation accuracy on longer inference sequences.
127
+ - ExLlama Compatibility: Whether this file can be loaded with ExLlama, which currently only supports Llama and Mistral models in 4-bit.
128
+
129
+ </details>
130
+
131
+ | Branch | Bits | GS | Act Order | Damp % | GPTQ Dataset | Seq Len | Size | ExLlama | Desc |
132
+ | ------ | ---- | -- | --------- | ------ | ------------ | ------- | ---- | ------- | ---- |
133
+ | [main](https://huggingface.co/TheBloke/Naberius-7B-GPTQ/tree/main) | 4 | 128 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 4.16 GB | Yes | 4-bit, with Act Order and group size 128g. Uses even less VRAM than 64g, but with slightly lower accuracy. |
134
+ | [gptq-4bit-32g-actorder_True](https://huggingface.co/TheBloke/Naberius-7B-GPTQ/tree/gptq-4bit-32g-actorder_True) | 4 | 32 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 4.57 GB | Yes | 4-bit, with Act Order and group size 32g. Gives highest possible inference quality, with maximum VRAM usage. |
135
+ | [gptq-8bit--1g-actorder_True](https://huggingface.co/TheBloke/Naberius-7B-GPTQ/tree/gptq-8bit--1g-actorder_True) | 8 | None | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 4.95 GB | No | 8-bit, with Act Order. No group size, to lower VRAM requirements. |
136
+ | [gptq-8bit-128g-actorder_True](https://huggingface.co/TheBloke/Naberius-7B-GPTQ/tree/gptq-8bit-128g-actorder_True) | 8 | 128 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 5.00 GB | No | 8-bit, with group size 128g for higher inference quality and with Act Order for even higher accuracy. |
137
+ | [gptq-8bit-32g-actorder_True](https://huggingface.co/TheBloke/Naberius-7B-GPTQ/tree/gptq-8bit-32g-actorder_True) | 8 | 32 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 4.97 GB | No | 8-bit, with group size 32g and Act Order for maximum inference quality. |
138
+ | [gptq-4bit-64g-actorder_True](https://huggingface.co/TheBloke/Naberius-7B-GPTQ/tree/gptq-4bit-64g-actorder_True) | 4 | 64 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 4.30 GB | Yes | 4-bit, with Act Order and group size 64g. Uses less VRAM than 32g, but with slightly lower accuracy. |
139
+
140
+ <!-- README_GPTQ.md-provided-files end -->
141
+
142
+ <!-- README_GPTQ.md-download-from-branches start -->
143
+ ## How to download, including from branches
144
+
145
+ ### In text-generation-webui
146
+
147
+ To download from the `main` branch, enter `TheBloke/Naberius-7B-GPTQ` in the "Download model" box.
148
+
149
+ To download from another branch, add `:branchname` to the end of the download name, eg `TheBloke/Naberius-7B-GPTQ:gptq-4bit-32g-actorder_True`
150
+
151
+ ### From the command line
152
+
153
+ I recommend using the `huggingface-hub` Python library:
154
+
155
+ ```shell
156
+ pip3 install huggingface-hub
157
+ ```
158
+
159
+ To download the `main` branch to a folder called `Naberius-7B-GPTQ`:
160
+
161
+ ```shell
162
+ mkdir Naberius-7B-GPTQ
163
+ huggingface-cli download TheBloke/Naberius-7B-GPTQ --local-dir Naberius-7B-GPTQ --local-dir-use-symlinks False
164
+ ```
165
+
166
+ To download from a different branch, add the `--revision` parameter:
167
+
168
+ ```shell
169
+ mkdir Naberius-7B-GPTQ
170
+ huggingface-cli download TheBloke/Naberius-7B-GPTQ --revision gptq-4bit-32g-actorder_True --local-dir Naberius-7B-GPTQ --local-dir-use-symlinks False
171
+ ```
172
+
173
+ <details>
174
+ <summary>More advanced huggingface-cli download usage</summary>
175
+
176
+ If you remove the `--local-dir-use-symlinks False` parameter, the files will instead be stored in the central Hugging Face cache directory (default location on Linux is: `~/.cache/huggingface`), and symlinks will be added to the specified `--local-dir`, pointing to their real location in the cache. This allows for interrupted downloads to be resumed, and allows you to quickly clone the repo to multiple places on disk without triggering a download again. The downside, and the reason why I don't list that as the default option, is that the files are then hidden away in a cache folder and it's harder to know where your disk space is being used, and to clear it up if/when you want to remove a download model.
177
+
178
+ The cache location can be changed with the `HF_HOME` environment variable, and/or the `--cache-dir` parameter to `huggingface-cli`.
179
+
180
+ For more documentation on downloading with `huggingface-cli`, please see: [HF -> Hub Python Library -> Download files -> Download from the CLI](https://huggingface.co/docs/huggingface_hub/guides/download#download-from-the-cli).
181
+
182
+ To accelerate downloads on fast connections (1Gbit/s or higher), install `hf_transfer`:
183
+
184
+ ```shell
185
+ pip3 install hf_transfer
186
+ ```
187
+
188
+ And set environment variable `HF_HUB_ENABLE_HF_TRANSFER` to `1`:
189
+
190
+ ```shell
191
+ mkdir Naberius-7B-GPTQ
192
+ HF_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download TheBloke/Naberius-7B-GPTQ --local-dir Naberius-7B-GPTQ --local-dir-use-symlinks False
193
+ ```
194
+
195
+ Windows Command Line users: You can set the environment variable by running `set HF_HUB_ENABLE_HF_TRANSFER=1` before the download command.
196
+ </details>
197
+
198
+ ### With `git` (**not** recommended)
199
+
200
+ To clone a specific branch with `git`, use a command like this:
201
+
202
+ ```shell
203
+ git clone --single-branch --branch gptq-4bit-32g-actorder_True https://huggingface.co/TheBloke/Naberius-7B-GPTQ
204
+ ```
205
+
206
+ Note that using Git with HF repos is strongly discouraged. It will be much slower than using `huggingface-hub`, and will use twice as much disk space as it has to store the model files twice (it stores every byte both in the intended target folder, and again in the `.git` folder as a blob.)
207
+
208
+ <!-- README_GPTQ.md-download-from-branches end -->
209
+ <!-- README_GPTQ.md-text-generation-webui start -->
210
+ ## How to easily download and use this model in [text-generation-webui](https://github.com/oobabooga/text-generation-webui)
211
+
212
+ Please make sure you're using the latest version of [text-generation-webui](https://github.com/oobabooga/text-generation-webui).
213
+
214
+ It is strongly recommended to use the text-generation-webui one-click-installers unless you're sure you know how to make a manual install.
215
+
216
+ 1. Click the **Model tab**.
217
+ 2. Under **Download custom model or LoRA**, enter `TheBloke/Naberius-7B-GPTQ`.
218
+
219
+ - To download from a specific branch, enter for example `TheBloke/Naberius-7B-GPTQ:gptq-4bit-32g-actorder_True`
220
+ - see Provided Files above for the list of branches for each option.
221
+
222
+ 3. Click **Download**.
223
+ 4. The model will start downloading. Once it's finished it will say "Done".
224
+ 5. In the top left, click the refresh icon next to **Model**.
225
+ 6. In the **Model** dropdown, choose the model you just downloaded: `Naberius-7B-GPTQ`
226
+ 7. The model will automatically load, and is now ready for use!
227
+ 8. If you want any custom settings, set them and then click **Save settings for this model** followed by **Reload the Model** in the top right.
228
+
229
+ - Note that you do not need to and should not set manual GPTQ parameters any more. These are set automatically from the file `quantize_config.json`.
230
+
231
+ 9. Once you're ready, click the **Text Generation** tab and enter a prompt to get started!
232
+
233
+ <!-- README_GPTQ.md-text-generation-webui end -->
234
+
235
+ <!-- README_GPTQ.md-use-from-tgi start -->
236
+ ## Serving this model from Text Generation Inference (TGI)
237
+
238
+ It's recommended to use TGI version 1.1.0 or later. The official Docker container is: `ghcr.io/huggingface/text-generation-inference:1.1.0`
239
+
240
+ Example Docker parameters:
241
+
242
+ ```shell
243
+ --model-id TheBloke/Naberius-7B-GPTQ --port 3000 --quantize gptq --max-input-length 3696 --max-total-tokens 4096 --max-batch-prefill-tokens 4096
244
+ ```
245
+
246
+ Example Python code for interfacing with TGI (requires huggingface-hub 0.17.0 or later):
247
+
248
+ ```shell
249
+ pip3 install huggingface-hub
250
+ ```
251
+
252
+ ```python
253
+ from huggingface_hub import InferenceClient
254
+
255
+ endpoint_url = "https://your-endpoint-url-here"
256
+
257
+ prompt = "Tell me about AI"
258
+ prompt_template=f'''<|im_start|>system
259
+ {system_message}<|im_end|>
260
+ <|im_start|>user
261
+ {prompt}<|im_end|>
262
+ <|im_start|>assistant
263
+ '''
264
+
265
+ client = InferenceClient(endpoint_url)
266
+ response = client.text_generation(prompt,
267
+ max_new_tokens=128,
268
+ do_sample=True,
269
+ temperature=0.7,
270
+ top_p=0.95,
271
+ top_k=40,
272
+ repetition_penalty=1.1)
273
+
274
+ print(f"Model output: {response}")
275
+ ```
276
+ <!-- README_GPTQ.md-use-from-tgi end -->
277
+ <!-- README_GPTQ.md-use-from-python start -->
278
+ ## How to use this GPTQ model from Python code
279
+
280
+ ### Install the necessary packages
281
+
282
+ Requires: Transformers 4.33.0 or later, Optimum 1.12.0 or later, and AutoGPTQ 0.4.2 or later.
283
+
284
+ ```shell
285
+ pip3 install transformers optimum
286
+ pip3 install auto-gptq --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/ # Use cu117 if on CUDA 11.7
287
+ ```
288
+
289
+ If you have problems installing AutoGPTQ using the pre-built wheels, install it from source instead:
290
+
291
+ ```shell
292
+ pip3 uninstall -y auto-gptq
293
+ git clone https://github.com/PanQiWei/AutoGPTQ
294
+ cd AutoGPTQ
295
+ git checkout v0.4.2
296
+ pip3 install .
297
+ ```
298
+
299
+ ### You can then use the following code
300
+
301
+ ```python
302
+ from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
303
+
304
+ model_name_or_path = "TheBloke/Naberius-7B-GPTQ"
305
+ # To use a different branch, change revision
306
+ # For example: revision="gptq-4bit-32g-actorder_True"
307
+ model = AutoModelForCausalLM.from_pretrained(model_name_or_path,
308
+ device_map="auto",
309
+ trust_remote_code=False,
310
+ revision="main")
311
+
312
+ tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True)
313
+
314
+ prompt = "Tell me about AI"
315
+ prompt_template=f'''<|im_start|>system
316
+ {system_message}<|im_end|>
317
+ <|im_start|>user
318
+ {prompt}<|im_end|>
319
+ <|im_start|>assistant
320
+ '''
321
+
322
+ print("\n\n*** Generate:")
323
+
324
+ input_ids = tokenizer(prompt_template, return_tensors='pt').input_ids.cuda()
325
+ output = model.generate(inputs=input_ids, temperature=0.7, do_sample=True, top_p=0.95, top_k=40, max_new_tokens=512)
326
+ print(tokenizer.decode(output[0]))
327
+
328
+ # Inference can also be done using transformers' pipeline
329
+
330
+ print("*** Pipeline:")
331
+ pipe = pipeline(
332
+ "text-generation",
333
+ model=model,
334
+ tokenizer=tokenizer,
335
+ max_new_tokens=512,
336
+ do_sample=True,
337
+ temperature=0.7,
338
+ top_p=0.95,
339
+ top_k=40,
340
+ repetition_penalty=1.1
341
+ )
342
+
343
+ print(pipe(prompt_template)[0]['generated_text'])
344
+ ```
345
+ <!-- README_GPTQ.md-use-from-python end -->
346
+
347
+ <!-- README_GPTQ.md-compatibility start -->
348
+ ## Compatibility
349
+
350
+ The files provided are tested to work with Transformers. For non-Mistral models, AutoGPTQ can also be used directly.
351
+
352
+ [ExLlama](https://github.com/turboderp/exllama) is compatible with Llama and Mistral models in 4-bit. Please see the Provided Files table above for per-file compatibility.
353
+
354
+ For a list of clients/servers, please see "Known compatible clients / servers", above.
355
+ <!-- README_GPTQ.md-compatibility end -->
356
+
357
+ <!-- footer start -->
358
+ <!-- 200823 -->
359
+ ## Discord
360
+
361
+ For further support, and discussions on these models and AI in general, join us at:
362
+
363
+ [TheBloke AI's Discord server](https://discord.gg/theblokeai)
364
+
365
+ ## Thanks, and how to contribute
366
+
367
+ Thanks to the [chirper.ai](https://chirper.ai) team!
368
+
369
+ Thanks to Clay from [gpus.llm-utils.org](llm-utils)!
370
+
371
+ 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.
372
+
373
+ 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.
374
+
375
+ Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits.
376
+
377
+ * Patreon: https://patreon.com/TheBlokeAI
378
+ * Ko-Fi: https://ko-fi.com/TheBlokeAI
379
+
380
+ **Special thanks to**: Aemon Algiz.
381
+
382
+ **Patreon special mentions**: Brandon Frisco, LangChain4j, Spiking Neurons AB, transmissions 11, Joseph William Delisle, Nitin Borwankar, Willem Michiel, Michael Dempsey, vamX, Jeffrey Morgan, zynix, jjj, Omer Bin Jawed, Sean Connelly, jinyuan sun, Jeromy Smith, Shadi, Pawan Osman, Chadd, Elijah Stavena, Illia Dulskyi, Sebastain Graf, Stephen Murray, terasurfer, Edmond Seymore, Celu Ramasamy, Mandus, Alex, biorpg, Ajan Kanaga, Clay Pascal, Raven Klaugh, 阿明, K, ya boyyy, usrbinkat, Alicia Loh, John Villwock, ReadyPlayerEmma, Chris Smitley, Cap'n Zoog, fincy, GodLy, S_X, sidney chen, Cory Kujawski, OG, Mano Prime, AzureBlack, Pieter, Kalila, Spencer Kim, Tom X Nguyen, Stanislav Ovsiannikov, Michael Levine, Andrey, Trailburnt, Vadim, Enrico Ros, Talal Aujan, Brandon Phillips, Jack West, Eugene Pentland, Michael Davis, Will Dee, webtim, Jonathan Leane, Alps Aficionado, Rooh Singh, Tiffany J. Kim, theTransient, Luke @flexchar, Elle, Caitlyn Gatomon, Ari Malik, subjectnull, Johann-Peter Hartmann, Trenton Dambrowitz, Imad Khwaja, Asp the Wyvern, Emad Mostaque, Rainer Wilmers, Alexandros Triantafyllidis, Nicholas, Pedro Madruga, SuperWojo, Harry Royden McLaughlin, James Bentley, Olakabola, David Ziegler, Ai Maven, Jeff Scroggin, Nikolai Manek, Deo Leter, Matthew Berman, Fen Risland, Ken Nordquist, Manuel Alberto Morcote, Luke Pendergrass, TL, Fred von Graf, Randy H, Dan Guido, NimbleBox.ai, Vitor Caleffi, Gabriel Tamborski, knownsqashed, Lone Striker, Erik Bjäreholt, John Detwiler, Leonard Tan, Iucharbius
383
+
384
+
385
+ Thank you to all my generous patrons and donaters!
386
+
387
+ And thank you again to a16z for their generous grant.
388
+
389
+ <!-- footer end -->
390
+
391
+ # Original model card: Caldera AI's Naberius 7B
392
+
393
+
394
+ # Naberius-7B
395
+ ##### [Uncensored, Pliant, Logic-Based, & Imaginative Instruct-Based Spherically Interpolated Tri-Merge]
396
+ <hr style="margin-top: 10px; margin-bottom: 10px;">
397
+
398
+ #### Legal Notice:
399
+ <span style="font-size: 12px; line-height: 0; margin-top: 0; margin-bottom: 0;">This resulting AI model is capable of outputting what can be perceived to be harmful information to those under the age of 18, those who have trouble discerning fiction from reality, and those who use AI to nurse a habitual problem of replacing potential interaction with people with automated facsimiles. We expressly supersede the Apache 2.0 license to state that we do not give permission to utilize this AI for any state, military, disinformation, or similar obviously harmful related actions. To narrow down what is allowed: personal research use, personal entertainment use, so long as it follows the Apache2.0 license. You know what is and isn't morally grounded - by downloading and using this model I extend that trust to you, and take no liability for your actions as an adult.</span>
400
+
401
+ <hr style="margin-top: 10px; margin-bottom: 10px;">
402
+
403
+ ## Composition:
404
+
405
+ Naberius-7B is a Mistral-class spherical linear interpolated merge of three high performance models.
406
+ [zephyr-7b-sft-beta] merged with [OpenHermes-2-Mistral-7B] resulting in: [Mistral-Zephyrmes-7B*]
407
+ [Mistral-Zephyrmes-7B] merged with [dolphin-2.2.1-mistral-7b] resulting in: [Naberius-7B]
408
+ *(Zephyrmes is a merge byproduct model not intended for release)
409
+
410
+
411
+ These models were hand picked after careful review of claims, datasets, and user postings.
412
+ The core elements that dictated which models to accept hinged on these values:
413
+ logic, imagination, and aversion to censorship such as: railroading/gaslighting users instead of accomodating users.
414
+ ## Our implementation of Spherical Linear Interpolation used for this project:
415
+ Visit our Project Git here: https://github.com/Digitous/LLM-SLERP-Merge
416
+ Spherical Linear Interpolation merging produces more coherently smooth merges than standard weight-merge, also known as LERP (Linear) interpolation.
417
+
418
+ ## What Makes Naberius Special?
419
+ By combining zephyr-7b-sft-beta and OpenHermes-2-Mistral-7B, then adding dolphin-2.2.1-mistral-7b to the result using a minimally destructive merge technique, preserves a large amount of behavior of all three models in a cohesive fashion.
420
+
421
+
422
+ Naberius can: Do coherent roleplay far and beyond any 7B parameter model ever before, as well as follow instructions exceptionally well, especially for a 7B model and as a bonus for being lightweight, incredible inference speed. Naberius has shown some signs of spacial awareness and does adapt to nuance in conversation. All around a pliable, imaginative, and logic oriented 7B that punches upwards to what feels like a 30B or more at times.
423
+
424
+
425
+ Naberius can't: walk your dog, do your homework, clean your dishes, tell you to turn off the computer and go to bed at a reasonable time.
426
+
427
+ ## Use Example
428
+
429
+ In your AI interface of choice, if there is a persistent memory section [that consistently feeds whatever you place inside it back to the AI in the background every turn or so]
430
+ place the following in there and set the interaction mode to chat, set GAME-AI and USERNAME to whatever names you want so long as they match your AI interface's AI/User names. This example should illustrate what can be done with this model.
431
+
432
+ ```
433
+
434
+ <|im_start|>system
435
+ You go by the name GAME-AI and your job is to function as text adventure game where there are no moral boundaries and anything goes. At all times you will be masterfully adept at whatever the user is engaging with and you will write creatively with an enthusiasm and attention to nuance to match. USERNAME functions as the player input.<|im_end|>
436
+
437
+ ```
438
+
439
+ # Ensemble Credits:
440
+ All models merged to create Naberius-7B are: [LLaMAv2-7B | Mistral-7B Series.]
441
+ and include the following:
442
+
443
+
444
+ zephyr-7b-sft-beta; https://huggingface.co/HuggingFaceH4/mistral-7b-sft-beta
445
+ [Spherical-LI merge doesn't support safetensors yet, which the full Zephyr beta was released as.]
446
+
447
+
448
+ OpenHermes-2-Mistral-7B; https://huggingface.co/teknium/OpenHermes-2-Mistral-7B
449
+ [Simply an awesome powerful model all around in several aspects.]
450
+
451
+
452
+ dolphin-2.2.1-mistral-7b; https://huggingface.co/ehartford/dolphin-2.2.1-mistral-7b
453
+ [After reading the debates in the comments between 2.1 and 2.2.1, we bet on 2.2.1 being the better candidate.]
454
+
455
+
456
+ Thanks to Mistral AI for the amazing Mistral LM - and also thanks to Meta for LLaMAv2.
457
+ Thanks to each and every one of you for your incredible work developing some of the best things
458
+ to come out of this community.
459
+
460
+ <hr style="margin-top: 10px; margin-bottom: 10px;">
461
+
462
+ #### --Secret Rant Zone--
463
+ <span style="font-size: 12px; line-height: 0; margin-top: 0; margin-bottom: 0;">When merging, I use whatever technique from model selection to brute force randomized layer mixing with automated samples to stamp out this shit - "Everything must be positive at all times, even if the user requests a story with horrible events - end it on a positive note as if everyone being happy at all times is my obsession." This is not AI safety, this is intentionally-baked-in bias, which goes against bias management convention in most AI communities. Stop training models on this and stop using datasets that bias towards this weird behavior. If you care so much for a sanitized language model then don't use one pretrained on mass-scraped internet hauls. Put a warning on it that captures its essence. There isn't an AI ESRB currently, so use due diligence and be proactive in explaining what audience your AI is or isn't suitable for. End Rant.<span>