TheBloke commited on
Commit
e90102b
1 Parent(s): ab207d2

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +442 -0
README.md ADDED
@@ -0,0 +1,442 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: tavtav/Rose-20B
3
+ inference: false
4
+ language:
5
+ - en
6
+ license: llama2
7
+ model_creator: Tav
8
+ model_name: Rose 20B
9
+ model_type: llama
10
+ pipeline_tag: text-generation
11
+ prompt_template: 'Below is an instruction that describes a task. Write a response
12
+ that appropriately completes the request.
13
+
14
+
15
+ ### Instruction:
16
+
17
+ {prompt}
18
+
19
+
20
+ ### Response:
21
+
22
+ '
23
+ quantized_by: TheBloke
24
+ tags:
25
+ - text-generation-inference
26
+ - instruct
27
+ ---
28
+ <!-- markdownlint-disable MD041 -->
29
+
30
+ <!-- header start -->
31
+ <!-- 200823 -->
32
+ <div style="width: auto; margin-left: auto; margin-right: auto">
33
+ <img src="https://i.imgur.com/EBdldam.jpg" alt="TheBlokeAI" style="width: 100%; min-width: 400px; display: block; margin: auto;">
34
+ </div>
35
+ <div style="display: flex; justify-content: space-between; width: 100%;">
36
+ <div style="display: flex; flex-direction: column; align-items: flex-start;">
37
+ <p style="margin-top: 0.5em; margin-bottom: 0em;"><a href="https://discord.gg/theblokeai">Chat & support: TheBloke's Discord server</a></p>
38
+ </div>
39
+ <div style="display: flex; flex-direction: column; align-items: flex-end;">
40
+ <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>
41
+ </div>
42
+ </div>
43
+ <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>
44
+ <hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
45
+ <!-- header end -->
46
+
47
+ # Rose 20B - GPTQ
48
+ - Model creator: [Tav](https://huggingface.co/tavtav)
49
+ - Original model: [Rose 20B](https://huggingface.co/tavtav/Rose-20B)
50
+
51
+ <!-- description start -->
52
+ # Description
53
+
54
+ This repo contains GPTQ model files for [Tav's Rose 20B](https://huggingface.co/tavtav/Rose-20B).
55
+
56
+ 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.
57
+
58
+ These files were quantised using hardware kindly provided by [Massed Compute](https://massedcompute.com/).
59
+
60
+ <!-- description end -->
61
+ <!-- repositories-available start -->
62
+ ## Repositories available
63
+
64
+ * [AWQ model(s) for GPU inference.](https://huggingface.co/TheBloke/Rose-20B-AWQ)
65
+ * [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/Rose-20B-GPTQ)
66
+ * [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/Rose-20B-GGUF)
67
+ * [Tav's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/tavtav/Rose-20B)
68
+ <!-- repositories-available end -->
69
+
70
+ <!-- prompt-template start -->
71
+ ## Prompt template: Alpaca
72
+
73
+ ```
74
+ Below is an instruction that describes a task. Write a response that appropriately completes the request.
75
+
76
+ ### Instruction:
77
+ {prompt}
78
+
79
+ ### Response:
80
+
81
+ ```
82
+
83
+ <!-- prompt-template end -->
84
+
85
+
86
+
87
+ <!-- README_GPTQ.md-compatible clients start -->
88
+ ## Known compatible clients / servers
89
+
90
+ These GPTQ models are known to work in the following inference servers/webuis.
91
+
92
+ - [text-generation-webui](https://github.com/oobabooga/text-generation-webui)
93
+ - [KoboldAI United](https://github.com/henk717/koboldai)
94
+ - [LoLLMS Web UI](https://github.com/ParisNeo/lollms-webui)
95
+ - [Hugging Face Text Generation Inference (TGI)](https://github.com/huggingface/text-generation-inference)
96
+
97
+ This may not be a complete list; if you know of others, please let me know!
98
+ <!-- README_GPTQ.md-compatible clients end -->
99
+
100
+ <!-- README_GPTQ.md-provided-files start -->
101
+ ## Provided files, and GPTQ parameters
102
+
103
+ Multiple quantisation parameters are provided, to allow you to choose the best one for your hardware and requirements.
104
+
105
+ Each separate quant is in a different branch. See below for instructions on fetching from different branches.
106
+
107
+ Most GPTQ files are made with AutoGPTQ. Mistral models are currently made with Transformers.
108
+
109
+ <details>
110
+ <summary>Explanation of GPTQ parameters</summary>
111
+
112
+ - Bits: The bit size of the quantised model.
113
+ - GS: GPTQ group size. Higher numbers use less VRAM, but have lower quantisation accuracy. "None" is the lowest possible value.
114
+ - 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.
115
+ - Damp %: A GPTQ parameter that affects how samples are processed for quantisation. 0.01 is default, but 0.1 results in slightly better accuracy.
116
+ - 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).
117
+ - 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.
118
+ - ExLlama Compatibility: Whether this file can be loaded with ExLlama, which currently only supports Llama and Mistral models in 4-bit.
119
+
120
+ </details>
121
+
122
+ | Branch | Bits | GS | Act Order | Damp % | GPTQ Dataset | Seq Len | Size | ExLlama | Desc |
123
+ | ------ | ---- | -- | --------- | ------ | ------------ | ------- | ---- | ------- | ---- |
124
+ | [main](https://huggingface.co/TheBloke/Rose-20B-GPTQ/tree/main) | 4 | None | Yes | 0.1 | [VMware Open Instruct](https://huggingface.co/datasets/VMware/open-instruct/viewer/) | 4096 | 10.52 GB | Yes | 4-bit, with Act Order. No group size, to lower VRAM requirements. |
125
+ | [gptq-4bit-128g-actorder_True](https://huggingface.co/TheBloke/Rose-20B-GPTQ/tree/gptq-4bit-128g-actorder_True) | 4 | 128 | Yes | 0.1 | [VMware Open Instruct](https://huggingface.co/datasets/VMware/open-instruct/viewer/) | 4096 | 10.89 GB | Yes | 4-bit, with Act Order and group size 128g. Uses even less VRAM than 64g, but with slightly lower accuracy. |
126
+ | [gptq-4bit-32g-actorder_True](https://huggingface.co/TheBloke/Rose-20B-GPTQ/tree/gptq-4bit-32g-actorder_True) | 4 | 32 | Yes | 0.1 | [VMware Open Instruct](https://huggingface.co/datasets/VMware/open-instruct/viewer/) | 4096 | 12.04 GB | Yes | 4-bit, with Act Order and group size 32g. Gives highest possible inference quality, with maximum VRAM usage. |
127
+ | [gptq-3bit-128g-actorder_True](https://huggingface.co/TheBloke/Rose-20B-GPTQ/tree/gptq-3bit-128g-actorder_True) | 3 | 128 | Yes | 0.1 | [VMware Open Instruct](https://huggingface.co/datasets/VMware/open-instruct/viewer/) | 4096 | 8.41 GB | No | 3-bit, with group size 128g and act-order. Higher quality than 128g-False. |
128
+ | [gptq-8bit--1g-actorder_True](https://huggingface.co/TheBloke/Rose-20B-GPTQ/tree/gptq-8bit--1g-actorder_True) | 8 | None | Yes | 0.1 | [VMware Open Instruct](https://huggingface.co/datasets/VMware/open-instruct/viewer/) | 4096 | 20.35 GB | No | 8-bit, with Act Order. No group size, to lower VRAM requirements. |
129
+ | [gptq-3bit-32g-actorder_True](https://huggingface.co/TheBloke/Rose-20B-GPTQ/tree/gptq-3bit-32g-actorder_True) | 3 | 32 | Yes | 0.1 | [VMware Open Instruct](https://huggingface.co/datasets/VMware/open-instruct/viewer/) | 4096 | 9.51 GB | No | 3-bit, with group size 64g and act-order. Highest quality 3-bit option. |
130
+ | [gptq-8bit-128g-actorder_True](https://huggingface.co/TheBloke/Rose-20B-GPTQ/tree/gptq-8bit-128g-actorder_True) | 8 | 128 | Yes | 0.1 | [VMware Open Instruct](https://huggingface.co/datasets/VMware/open-instruct/viewer/) | 4096 | 20.80 GB | No | 8-bit, with group size 128g for higher inference quality and with Act Order for even higher accuracy. |
131
+
132
+ <!-- README_GPTQ.md-provided-files end -->
133
+
134
+ <!-- README_GPTQ.md-download-from-branches start -->
135
+ ## How to download, including from branches
136
+
137
+ ### In text-generation-webui
138
+
139
+ To download from the `main` branch, enter `TheBloke/Rose-20B-GPTQ` in the "Download model" box.
140
+
141
+ To download from another branch, add `:branchname` to the end of the download name, eg `TheBloke/Rose-20B-GPTQ:gptq-4bit-128g-actorder_True`
142
+
143
+ ### From the command line
144
+
145
+ I recommend using the `huggingface-hub` Python library:
146
+
147
+ ```shell
148
+ pip3 install huggingface-hub
149
+ ```
150
+
151
+ To download the `main` branch to a folder called `Rose-20B-GPTQ`:
152
+
153
+ ```shell
154
+ mkdir Rose-20B-GPTQ
155
+ huggingface-cli download TheBloke/Rose-20B-GPTQ --local-dir Rose-20B-GPTQ --local-dir-use-symlinks False
156
+ ```
157
+
158
+ To download from a different branch, add the `--revision` parameter:
159
+
160
+ ```shell
161
+ mkdir Rose-20B-GPTQ
162
+ huggingface-cli download TheBloke/Rose-20B-GPTQ --revision gptq-4bit-128g-actorder_True --local-dir Rose-20B-GPTQ --local-dir-use-symlinks False
163
+ ```
164
+
165
+ <details>
166
+ <summary>More advanced huggingface-cli download usage</summary>
167
+
168
+ 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.
169
+
170
+ The cache location can be changed with the `HF_HOME` environment variable, and/or the `--cache-dir` parameter to `huggingface-cli`.
171
+
172
+ 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).
173
+
174
+ To accelerate downloads on fast connections (1Gbit/s or higher), install `hf_transfer`:
175
+
176
+ ```shell
177
+ pip3 install hf_transfer
178
+ ```
179
+
180
+ And set environment variable `HF_HUB_ENABLE_HF_TRANSFER` to `1`:
181
+
182
+ ```shell
183
+ mkdir Rose-20B-GPTQ
184
+ HF_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download TheBloke/Rose-20B-GPTQ --local-dir Rose-20B-GPTQ --local-dir-use-symlinks False
185
+ ```
186
+
187
+ Windows Command Line users: You can set the environment variable by running `set HF_HUB_ENABLE_HF_TRANSFER=1` before the download command.
188
+ </details>
189
+
190
+ ### With `git` (**not** recommended)
191
+
192
+ To clone a specific branch with `git`, use a command like this:
193
+
194
+ ```shell
195
+ git clone --single-branch --branch gptq-4bit-128g-actorder_True https://huggingface.co/TheBloke/Rose-20B-GPTQ
196
+ ```
197
+
198
+ 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.)
199
+
200
+ <!-- README_GPTQ.md-download-from-branches end -->
201
+ <!-- README_GPTQ.md-text-generation-webui start -->
202
+ ## How to easily download and use this model in [text-generation-webui](https://github.com/oobabooga/text-generation-webui)
203
+
204
+ Please make sure you're using the latest version of [text-generation-webui](https://github.com/oobabooga/text-generation-webui).
205
+
206
+ 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.
207
+
208
+ 1. Click the **Model tab**.
209
+ 2. Under **Download custom model or LoRA**, enter `TheBloke/Rose-20B-GPTQ`.
210
+
211
+ - To download from a specific branch, enter for example `TheBloke/Rose-20B-GPTQ:gptq-4bit-128g-actorder_True`
212
+ - see Provided Files above for the list of branches for each option.
213
+
214
+ 3. Click **Download**.
215
+ 4. The model will start downloading. Once it's finished it will say "Done".
216
+ 5. In the top left, click the refresh icon next to **Model**.
217
+ 6. In the **Model** dropdown, choose the model you just downloaded: `Rose-20B-GPTQ`
218
+ 7. The model will automatically load, and is now ready for use!
219
+ 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.
220
+
221
+ - 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`.
222
+
223
+ 9. Once you're ready, click the **Text Generation** tab and enter a prompt to get started!
224
+
225
+ <!-- README_GPTQ.md-text-generation-webui end -->
226
+
227
+ <!-- README_GPTQ.md-use-from-tgi start -->
228
+ ## Serving this model from Text Generation Inference (TGI)
229
+
230
+ 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`
231
+
232
+ Example Docker parameters:
233
+
234
+ ```shell
235
+ --model-id TheBloke/Rose-20B-GPTQ --port 3000 --quantize gptq --max-input-length 3696 --max-total-tokens 4096 --max-batch-prefill-tokens 4096
236
+ ```
237
+
238
+ Example Python code for interfacing with TGI (requires huggingface-hub 0.17.0 or later):
239
+
240
+ ```shell
241
+ pip3 install huggingface-hub
242
+ ```
243
+
244
+ ```python
245
+ from huggingface_hub import InferenceClient
246
+
247
+ endpoint_url = "https://your-endpoint-url-here"
248
+
249
+ prompt = "Tell me about AI"
250
+ prompt_template=f'''Below is an instruction that describes a task. Write a response that appropriately completes the request.
251
+
252
+ ### Instruction:
253
+ {prompt}
254
+
255
+ ### Response:
256
+ '''
257
+
258
+ client = InferenceClient(endpoint_url)
259
+ response = client.text_generation(prompt,
260
+ max_new_tokens=128,
261
+ do_sample=True,
262
+ temperature=0.7,
263
+ top_p=0.95,
264
+ top_k=40,
265
+ repetition_penalty=1.1)
266
+
267
+ print(f"Model output: {response}")
268
+ ```
269
+ <!-- README_GPTQ.md-use-from-tgi end -->
270
+ <!-- README_GPTQ.md-use-from-python start -->
271
+ ## Python code example: inference from this GPTQ model
272
+
273
+ ### Install the necessary packages
274
+
275
+ Requires: Transformers 4.33.0 or later, Optimum 1.12.0 or later, and AutoGPTQ 0.4.2 or later.
276
+
277
+ ```shell
278
+ pip3 install --upgrade transformers optimum
279
+ # If using PyTorch 2.1 + CUDA 12.x:
280
+ pip3 install --upgrade auto-gptq
281
+ # or, if using PyTorch 2.1 + CUDA 11.x:
282
+ pip3 install --upgrade auto-gptq --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/
283
+ ```
284
+
285
+ If you are using PyTorch 2.0, you will need to install AutoGPTQ from source. Likewise if you have problems with the pre-built wheels, you should try building from source:
286
+
287
+ ```shell
288
+ pip3 uninstall -y auto-gptq
289
+ git clone https://github.com/PanQiWei/AutoGPTQ
290
+ cd AutoGPTQ
291
+ git checkout v0.5.1
292
+ pip3 install .
293
+ ```
294
+
295
+ ### Example Python code
296
+
297
+ ```python
298
+ from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
299
+
300
+ model_name_or_path = "TheBloke/Rose-20B-GPTQ"
301
+ # To use a different branch, change revision
302
+ # For example: revision="gptq-4bit-128g-actorder_True"
303
+ model = AutoModelForCausalLM.from_pretrained(model_name_or_path,
304
+ device_map="auto",
305
+ trust_remote_code=False,
306
+ revision="main")
307
+
308
+ tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True)
309
+
310
+ prompt = "Tell me about AI"
311
+ prompt_template=f'''Below is an instruction that describes a task. Write a response that appropriately completes the request.
312
+
313
+ ### Instruction:
314
+ {prompt}
315
+
316
+ ### Response:
317
+ '''
318
+
319
+ print("\n\n*** Generate:")
320
+
321
+ input_ids = tokenizer(prompt_template, return_tensors='pt').input_ids.cuda()
322
+ output = model.generate(inputs=input_ids, temperature=0.7, do_sample=True, top_p=0.95, top_k=40, max_new_tokens=512)
323
+ print(tokenizer.decode(output[0]))
324
+
325
+ # Inference can also be done using transformers' pipeline
326
+
327
+ print("*** Pipeline:")
328
+ pipe = pipeline(
329
+ "text-generation",
330
+ model=model,
331
+ tokenizer=tokenizer,
332
+ max_new_tokens=512,
333
+ do_sample=True,
334
+ temperature=0.7,
335
+ top_p=0.95,
336
+ top_k=40,
337
+ repetition_penalty=1.1
338
+ )
339
+
340
+ print(pipe(prompt_template)[0]['generated_text'])
341
+ ```
342
+ <!-- README_GPTQ.md-use-from-python end -->
343
+
344
+ <!-- README_GPTQ.md-compatibility start -->
345
+ ## Compatibility
346
+
347
+ The files provided are tested to work with Transformers. For non-Mistral models, AutoGPTQ can also be used directly.
348
+
349
+ [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.
350
+
351
+ For a list of clients/servers, please see "Known compatible clients / servers", above.
352
+ <!-- README_GPTQ.md-compatibility end -->
353
+
354
+ <!-- footer start -->
355
+ <!-- 200823 -->
356
+ ## Discord
357
+
358
+ For further support, and discussions on these models and AI in general, join us at:
359
+
360
+ [TheBloke AI's Discord server](https://discord.gg/theblokeai)
361
+
362
+ ## Thanks, and how to contribute
363
+
364
+ Thanks to the [chirper.ai](https://chirper.ai) team!
365
+
366
+ Thanks to Clay from [gpus.llm-utils.org](llm-utils)!
367
+
368
+ 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.
369
+
370
+ 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.
371
+
372
+ Donaters will get priority support on any and all AI/LLM/model questions and requests, access to a private Discord room, plus other benefits.
373
+
374
+ * Patreon: https://patreon.com/TheBlokeAI
375
+ * Ko-Fi: https://ko-fi.com/TheBlokeAI
376
+
377
+ **Special thanks to**: Aemon Algiz.
378
+
379
+ **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
380
+
381
+
382
+ Thank you to all my generous patrons and donaters!
383
+
384
+ And thank you again to a16z for their generous grant.
385
+
386
+ <!-- footer end -->
387
+
388
+ # Original model card: Tav's Rose 20B
389
+
390
+ <h1 style="text-align: center">Ross-20B</h1>
391
+ <center><img src="https://files.catbox.moe/rze9c9.png" alt="roseimage" width="350" height="350"></center>
392
+ <center><i>Image sourced by Shinon</i></center>
393
+ <h2 style="text-align: center">Experimental Frankenmerge Model</h2>
394
+
395
+
396
+ ## GGUF
397
+ [GGUF version here](https://huggingface.co/tavtav/Rose-20B-GGUF)
398
+
399
+ ## Model Details
400
+ A Frankenmerge with [Thorns-13B](https://huggingface.co/CalderaAI/13B-Thorns-l2) by CalderaAI and [Noromaid-13-v0.1.1](https://huggingface.co/NeverSleep/Noromaid-13b-v0.1.1) by NeverSleep (IkariDev and Undi). This recipe was proposed by Trappu and the layer distribution recipe was made by Undi. I thank them for sharing their knowledge with me. This model should be very good at any roleplay scenarios. I called the model "Rose" because it was a fitting name for a "thorny maid".
401
+
402
+ The recommended format to use is Alpaca.
403
+ ```
404
+ Below is an instruction that describes a task. Write a response that appropriately completes the request.
405
+
406
+ ### Instruction:
407
+ {prompt}
408
+
409
+ ### Response:
410
+ ```
411
+
412
+ Feel free to share any other prompts that works. This model is very robust.
413
+
414
+ ## Justification for its Existence
415
+ Potential base model for finetune experiments using our dataset to create Pygmalion-20B. Due to the already high capabilities, adding our dataset will mesh well with how the model performs.
416
+ Potential experimentation with merging with other 20B Frankenmerge models.
417
+
418
+ ## Model Recipe
419
+ ```
420
+ slices:
421
+ - sources:
422
+ - model: Thorns-13B
423
+ layer_range: [0, 16]
424
+ - sources:
425
+ - model: Noromaid-13B
426
+ layer_range: [8, 24]
427
+ - sources:
428
+ - model: Thorns-13B
429
+ layer_range: [17, 32]
430
+ - sources:
431
+ - model: Noromaid-13B
432
+ layer_range: [25, 40]
433
+ merge_method: passthrough
434
+ dtype: float16
435
+ ```
436
+ Again, credits to [Undi](https://huggingface.co/Undi95) for the recipe.
437
+
438
+ ## Reception
439
+ The model was given to a handful of members in the PygmalionAI Discord community for testing. A strong majority really enjoyed the model with only a couple giving the model a passing grade. Since our community has high standards for roleplaying models, I was surprised at the positive reception.
440
+
441
+ ## Contact
442
+ Send a message to tav (tav) on Discord if you want to talk about the model to me. I'm always open to receive comments.