TheBloke commited on
Commit
17a5ed2
1 Parent(s): 1d6cadf

Upload README.md

Browse files
Files changed (1) hide show
  1. README.md +39 -16
README.md CHANGED
@@ -1,4 +1,5 @@
1
  ---
 
2
  datasets:
3
  - PygmalionAI/PIPPA
4
  - Open-Orca/OpenOrca
@@ -10,10 +11,21 @@ language:
10
  - en
11
  license: llama2
12
  model_creator: PygmalionAI
13
- model_link: https://huggingface.co/PygmalionAI/mythalion-13b
14
  model_name: Mythalion 13B
15
  model_type: llama
16
  pipeline_tag: text-generation
 
 
 
 
 
 
 
 
 
 
 
 
17
  quantized_by: TheBloke
18
  tags:
19
  - text generation
@@ -53,6 +65,7 @@ Multiple GPTQ parameter permutations are provided; see Provided Files below for
53
  <!-- repositories-available start -->
54
  ## Repositories available
55
 
 
56
  * [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/Mythalion-13B-GPTQ)
57
  * [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/Mythalion-13B-GGUF)
58
  * [PygmalionAI's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/PygmalionAI/mythalion-13b)
@@ -73,6 +86,7 @@ Below is an instruction that describes a task. Write a response that appropriate
73
 
74
  <!-- prompt-template end -->
75
 
 
76
  <!-- README_GPTQ.md-provided-files start -->
77
  ## Provided files and GPTQ parameters
78
 
@@ -97,22 +111,22 @@ All recent GPTQ files are made with AutoGPTQ, and all files in non-main branches
97
 
98
  | Branch | Bits | GS | Act Order | Damp % | GPTQ Dataset | Seq Len | Size | ExLlama | Desc |
99
  | ------ | ---- | -- | --------- | ------ | ------------ | ------- | ---- | ------- | ---- |
100
- | [main](https://huggingface.co/TheBloke/Mythalion-13B-GPTQ/tree/main) | 4 | 128 | No | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 7.26 GB | Yes | Most compatible option. Good inference speed in AutoGPTQ and GPTQ-for-LLaMa. Lower inference quality than other options. |
101
- | [gptq-4bit-32g-actorder_True](https://huggingface.co/TheBloke/Mythalion-13B-GPTQ/tree/gptq-4bit-32g-actorder_True) | 4 | 32 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 8.00 GB | Yes | 4-bit, with Act Order and group size 32g. Gives highest possible inference quality, with maximum VRAM usage. Poor AutoGPTQ CUDA speed. |
102
- | [gptq-4bit-64g-actorder_True](https://huggingface.co/TheBloke/Mythalion-13B-GPTQ/tree/gptq-4bit-64g-actorder_True) | 4 | 64 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 7.51 GB | Yes | 4-bit, with Act Order and group size 64g. Uses less VRAM than 32g, but with slightly lower accuracy. Poor AutoGPTQ CUDA speed. |
103
- | [gptq-4bit-128g-actorder_True](https://huggingface.co/TheBloke/Mythalion-13B-GPTQ/tree/gptq-4bit-128g-actorder_True) | 4 | 128 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 7.26 GB | Yes | 4-bit, with Act Order and group size 128g. Uses even less VRAM than 64g, but with slightly lower accuracy. Poor AutoGPTQ CUDA speed. |
104
- | [gptq-8bit--1g-actorder_True](https://huggingface.co/TheBloke/Mythalion-13B-GPTQ/tree/gptq-8bit--1g-actorder_True) | 8 | None | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 13.36 GB | No | 8-bit, with Act Order. No group size, to lower VRAM requirements and to improve AutoGPTQ speed. |
105
- | [gptq-8bit-128g-actorder_True](https://huggingface.co/TheBloke/Mythalion-13B-GPTQ/tree/gptq-8bit-128g-actorder_True) | 8 | 128 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 13.65 GB | No | 8-bit, with group size 128g for higher inference quality and with Act Order for even higher accuracy. Poor AutoGPTQ CUDA speed. |
106
 
107
  <!-- README_GPTQ.md-provided-files end -->
108
 
109
  <!-- README_GPTQ.md-download-from-branches start -->
110
  ## How to download from branches
111
 
112
- - In text-generation-webui, you can add `:branch` to the end of the download name, eg `TheBloke/Mythalion-13B-GPTQ:gptq-4bit-32g-actorder_True`
113
  - With Git, you can clone a branch with:
114
  ```
115
- git clone --single-branch --branch gptq-4bit-32g-actorder_True https://huggingface.co/TheBloke/Mythalion-13B-GPTQ
116
  ```
117
  - In Python Transformers code, the branch is the `revision` parameter; see below.
118
  <!-- README_GPTQ.md-download-from-branches end -->
@@ -125,7 +139,7 @@ It is strongly recommended to use the text-generation-webui one-click-installers
125
 
126
  1. Click the **Model tab**.
127
  2. Under **Download custom model or LoRA**, enter `TheBloke/Mythalion-13B-GPTQ`.
128
- - To download from a specific branch, enter for example `TheBloke/Mythalion-13B-GPTQ:gptq-4bit-32g-actorder_True`
129
  - see Provided Files above for the list of branches for each option.
130
  3. Click **Download**.
131
  4. The model will start downloading. Once it's finished it will say "Done".
@@ -173,10 +187,10 @@ from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
173
 
174
  model_name_or_path = "TheBloke/Mythalion-13B-GPTQ"
175
  # To use a different branch, change revision
176
- # For example: revision="gptq-4bit-32g-actorder_True"
177
  model = AutoModelForCausalLM.from_pretrained(model_name_or_path,
178
- torch_dtype=torch.float16,
179
  device_map="auto",
 
180
  revision="main")
181
 
182
  tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True)
@@ -194,7 +208,7 @@ prompt_template=f'''Below is an instruction that describes a task. Write a respo
194
  print("\n\n*** Generate:")
195
 
196
  input_ids = tokenizer(prompt_template, return_tensors='pt').input_ids.cuda()
197
- output = model.generate(inputs=input_ids, temperature=0.7, max_new_tokens=512)
198
  print(tokenizer.decode(output[0]))
199
 
200
  # Inference can also be done using transformers' pipeline
@@ -205,9 +219,11 @@ pipe = pipeline(
205
  model=model,
206
  tokenizer=tokenizer,
207
  max_new_tokens=512,
 
208
  temperature=0.7,
209
  top_p=0.95,
210
- repetition_penalty=1.15
 
211
  )
212
 
213
  print(pipe(prompt_template)[0]['generated_text'])
@@ -232,10 +248,12 @@ For further support, and discussions on these models and AI in general, join us
232
 
233
  [TheBloke AI's Discord server](https://discord.gg/theblokeai)
234
 
235
- ## Thanks, and how to contribute.
236
 
237
  Thanks to the [chirper.ai](https://chirper.ai) team!
238
 
 
 
239
  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.
240
 
241
  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.
@@ -247,7 +265,7 @@ Donaters will get priority support on any and all AI/LLM/model questions and req
247
 
248
  **Special thanks to**: Aemon Algiz.
249
 
250
- **Patreon special mentions**: Russ Johnson, J, alfie_i, Alex, NimbleBox.ai, Chadd, Mandus, Nikolai Manek, Ken Nordquist, ya boyyy, Illia Dulskyi, Viktor Bowallius, vamX, Iucharbius, zynix, Magnesian, Clay Pascal, Pierre Kircher, Enrico Ros, Tony Hughes, Elle, Andrey, knownsqashed, Deep Realms, Jerry Meng, Lone Striker, Derek Yates, Pyrater, Mesiah Bishop, James Bentley, Femi Adebogun, Brandon Frisco, SuperWojo, Alps Aficionado, Michael Dempsey, Vitor Caleffi, Will Dee, Edmond Seymore, usrbinkat, LangChain4j, Kacper Wikieł, Luke Pendergrass, John Detwiler, theTransient, Nathan LeClaire, Tiffany J. Kim, biorpg, Eugene Pentland, Stanislav Ovsiannikov, Fred von Graf, terasurfer, Kalila, Dan Guido, Nitin Borwankar, 阿明, Ai Maven, John Villwock, Gabriel Puliatti, Stephen Murray, Asp the Wyvern, danny, Chris Smitley, ReadyPlayerEmma, S_X, Daniel P. Andersen, Olakabola, Jeffrey Morgan, Imad Khwaja, Caitlyn Gatomon, webtim, Alicia Loh, Trenton Dambrowitz, Swaroop Kallakuri, Erik Bjäreholt, Leonard Tan, Spiking Neurons AB, Luke @flexchar, Ajan Kanaga, Thomas Belote, Deo Leter, RoA, Willem Michiel, transmissions 11, subjectnull, Matthew Berman, Joseph William Delisle, David Ziegler, Michael Davis, Johann-Peter Hartmann, Talal Aujan, senxiiz, Artur Olbinski, Rainer Wilmers, Spencer Kim, Fen Risland, Cap'n Zoog, Rishabh Srivastava, Michael Levine, Geoffrey Montalvo, Sean Connelly, Alexandros Triantafyllidis, Pieter, Gabriel Tamborski, Sam, Subspace Studios, Junyu Yang, Pedro Madruga, Vadim, Cory Kujawski, K, Raven Klaugh, Randy H, Mano Prime, Sebastain Graf, Space Cruiser
251
 
252
 
253
  Thank you to all my generous patrons and donaters!
@@ -310,3 +328,8 @@ form a conversation history.
310
  The intended use-case for this model is fictional writing for entertainment purposes. Any other sort of usage is out of scope.
311
 
312
  As such, it was **not** fine-tuned to be safe and harmless: the base model _and_ this fine-tune have been trained on data known to contain profanity and texts that are lewd or otherwise offensive. It may produce socially unacceptable or undesirable text, even if the prompt itself does not include anything explicitly offensive. Outputs might often be factually wrong or misleading.
 
 
 
 
 
 
1
  ---
2
+ base_model: https://huggingface.co/PygmalionAI/mythalion-13b
3
  datasets:
4
  - PygmalionAI/PIPPA
5
  - Open-Orca/OpenOrca
 
11
  - en
12
  license: llama2
13
  model_creator: PygmalionAI
 
14
  model_name: Mythalion 13B
15
  model_type: llama
16
  pipeline_tag: text-generation
17
+ prompt_template: 'Below is an instruction that describes a task. Write a response
18
+ that appropriately completes the request.
19
+
20
+
21
+ ### Instruction:
22
+
23
+ {prompt}
24
+
25
+
26
+ ### Response:
27
+
28
+ '
29
  quantized_by: TheBloke
30
  tags:
31
  - text generation
 
65
  <!-- repositories-available start -->
66
  ## Repositories available
67
 
68
+ * [AWQ model(s) for GPU inference.](https://huggingface.co/TheBloke/Mythalion-13B-AWQ)
69
  * [GPTQ models for GPU inference, with multiple quantisation parameter options.](https://huggingface.co/TheBloke/Mythalion-13B-GPTQ)
70
  * [2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference](https://huggingface.co/TheBloke/Mythalion-13B-GGUF)
71
  * [PygmalionAI's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions](https://huggingface.co/PygmalionAI/mythalion-13b)
 
86
 
87
  <!-- prompt-template end -->
88
 
89
+
90
  <!-- README_GPTQ.md-provided-files start -->
91
  ## Provided files and GPTQ parameters
92
 
 
111
 
112
  | Branch | Bits | GS | Act Order | Damp % | GPTQ Dataset | Seq Len | Size | ExLlama | Desc |
113
  | ------ | ---- | -- | --------- | ------ | ------------ | ------- | ---- | ------- | ---- |
114
+ | [main](https://huggingface.co/TheBloke/Mythalion-13B-GPTQ/tree/main) | 4 | 128 | No | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 7.26 GB | Yes | 4-bit, without Act Order and group size 128g. |
115
+ | [gptq-4bit-32g-actorder_True](https://huggingface.co/TheBloke/Mythalion-13B-GPTQ/tree/gptq-4bit-32g-actorder_True) | 4 | 32 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 8.00 GB | Yes | 4-bit, with Act Order and group size 32g. Gives highest possible inference quality, with maximum VRAM usage. |
116
+ | [gptq-4bit-64g-actorder_True](https://huggingface.co/TheBloke/Mythalion-13B-GPTQ/tree/gptq-4bit-64g-actorder_True) | 4 | 64 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 7.51 GB | Yes | 4-bit, with Act Order and group size 64g. Uses less VRAM than 32g, but with slightly lower accuracy. |
117
+ | [gptq-4bit-128g-actorder_True](https://huggingface.co/TheBloke/Mythalion-13B-GPTQ/tree/gptq-4bit-128g-actorder_True) | 4 | 128 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 7.26 GB | Yes | 4-bit, with Act Order and group size 128g. Uses even less VRAM than 64g, but with slightly lower accuracy. |
118
+ | [gptq-8bit--1g-actorder_True](https://huggingface.co/TheBloke/Mythalion-13B-GPTQ/tree/gptq-8bit--1g-actorder_True) | 8 | None | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 13.36 GB | No | 8-bit, with Act Order. No group size, to lower VRAM requirements. |
119
+ | [gptq-8bit-128g-actorder_True](https://huggingface.co/TheBloke/Mythalion-13B-GPTQ/tree/gptq-8bit-128g-actorder_True) | 8 | 128 | Yes | 0.1 | [wikitext](https://huggingface.co/datasets/wikitext/viewer/wikitext-2-v1/test) | 4096 | 13.65 GB | No | 8-bit, with group size 128g for higher inference quality and with Act Order for even higher accuracy. |
120
 
121
  <!-- README_GPTQ.md-provided-files end -->
122
 
123
  <!-- README_GPTQ.md-download-from-branches start -->
124
  ## How to download from branches
125
 
126
+ - In text-generation-webui, you can add `:branch` to the end of the download name, eg `TheBloke/Mythalion-13B-GPTQ:main`
127
  - With Git, you can clone a branch with:
128
  ```
129
+ git clone --single-branch --branch main https://huggingface.co/TheBloke/Mythalion-13B-GPTQ
130
  ```
131
  - In Python Transformers code, the branch is the `revision` parameter; see below.
132
  <!-- README_GPTQ.md-download-from-branches end -->
 
139
 
140
  1. Click the **Model tab**.
141
  2. Under **Download custom model or LoRA**, enter `TheBloke/Mythalion-13B-GPTQ`.
142
+ - To download from a specific branch, enter for example `TheBloke/Mythalion-13B-GPTQ:main`
143
  - see Provided Files above for the list of branches for each option.
144
  3. Click **Download**.
145
  4. The model will start downloading. Once it's finished it will say "Done".
 
187
 
188
  model_name_or_path = "TheBloke/Mythalion-13B-GPTQ"
189
  # To use a different branch, change revision
190
+ # For example: revision="main"
191
  model = AutoModelForCausalLM.from_pretrained(model_name_or_path,
 
192
  device_map="auto",
193
+ trust_remote_code=False,
194
  revision="main")
195
 
196
  tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True)
 
208
  print("\n\n*** Generate:")
209
 
210
  input_ids = tokenizer(prompt_template, return_tensors='pt').input_ids.cuda()
211
+ output = model.generate(inputs=input_ids, temperature=0.7, do_sample=True, top_p=0.95, top_k=40, max_new_tokens=512)
212
  print(tokenizer.decode(output[0]))
213
 
214
  # Inference can also be done using transformers' pipeline
 
219
  model=model,
220
  tokenizer=tokenizer,
221
  max_new_tokens=512,
222
+ do_sample=True,
223
  temperature=0.7,
224
  top_p=0.95,
225
+ top_k=40,
226
+ repetition_penalty=1.1
227
  )
228
 
229
  print(pipe(prompt_template)[0]['generated_text'])
 
248
 
249
  [TheBloke AI's Discord server](https://discord.gg/theblokeai)
250
 
251
+ ## Thanks, and how to contribute
252
 
253
  Thanks to the [chirper.ai](https://chirper.ai) team!
254
 
255
+ Thanks to Clay from [gpus.llm-utils.org](llm-utils)!
256
+
257
  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.
258
 
259
  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.
 
265
 
266
  **Special thanks to**: Aemon Algiz.
267
 
268
+ **Patreon special mentions**: Alicia Loh, Stephen Murray, K, Ajan Kanaga, RoA, Magnesian, Deo Leter, Olakabola, Eugene Pentland, zynix, Deep Realms, Raymond Fosdick, Elijah Stavena, Iucharbius, Erik Bjäreholt, Luis Javier Navarrete Lozano, Nicholas, theTransient, John Detwiler, alfie_i, knownsqashed, Mano Prime, Willem Michiel, Enrico Ros, LangChain4j, OG, Michael Dempsey, Pierre Kircher, Pedro Madruga, James Bentley, Thomas Belote, Luke @flexchar, Leonard Tan, Johann-Peter Hartmann, Illia Dulskyi, Fen Risland, Chadd, S_X, Jeff Scroggin, Ken Nordquist, Sean Connelly, Artur Olbinski, Swaroop Kallakuri, Jack West, Ai Maven, David Ziegler, Russ Johnson, transmissions 11, John Villwock, Alps Aficionado, Clay Pascal, Viktor Bowallius, Subspace Studios, Rainer Wilmers, Trenton Dambrowitz, vamX, Michael Levine, 준교 김, Brandon Frisco, Kalila, Trailburnt, Randy H, Talal Aujan, Nathan Dryer, Vadim, 阿明, ReadyPlayerEmma, Tiffany J. Kim, George Stoitzev, Spencer Kim, Jerry Meng, Gabriel Tamborski, Cory Kujawski, Jeffrey Morgan, Spiking Neurons AB, Edmond Seymore, Alexandros Triantafyllidis, Lone Striker, Cap'n Zoog, Nikolai Manek, danny, ya boyyy, Derek Yates, usrbinkat, Mandus, TL, Nathan LeClaire, subjectnull, Imad Khwaja, webtim, Raven Klaugh, Asp the Wyvern, Gabriel Puliatti, Caitlyn Gatomon, Joseph William Delisle, Jonathan Leane, Luke Pendergrass, SuperWojo, Sebastain Graf, Will Dee, Fred von Graf, Andrey, Dan Guido, Daniel P. Andersen, Nitin Borwankar, Elle, Vitor Caleffi, biorpg, jjj, NimbleBox.ai, Pieter, Matthew Berman, terasurfer, Michael Davis, Alex, Stanislav Ovsiannikov
269
 
270
 
271
  Thank you to all my generous patrons and donaters!
 
328
  The intended use-case for this model is fictional writing for entertainment purposes. Any other sort of usage is out of scope.
329
 
330
  As such, it was **not** fine-tuned to be safe and harmless: the base model _and_ this fine-tune have been trained on data known to contain profanity and texts that are lewd or otherwise offensive. It may produce socially unacceptable or undesirable text, even if the prompt itself does not include anything explicitly offensive. Outputs might often be factually wrong or misleading.
331
+
332
+ ## Acknowledgements
333
+ We would like to thank [SpicyChat](https://spicychat.ai/) for sponsoring the training for the [Pygmalion-2 13B](https://huggingface.co/PygmalionAI/pygmalion-2-13b) model.
334
+
335
+ [<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)