TheBloke's LLM work is generously supported by a grant from andreessen horowitz (a16z)
Karen TheEditor V2 Strict Mistral 7B - GPTQ
- Model creator: FPHam
- Original model: Karen TheEditor V2 Strict Mistral 7B
Description
This repo contains GPTQ model files for FPHam's Karen TheEditor V2 Strict Mistral 7B.
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.
These files were quantised using hardware kindly provided by Massed Compute.
Repositories available
- AWQ model(s) for GPU inference.
- GPTQ models for GPU inference, with multiple quantisation parameter options.
- 2, 3, 4, 5, 6 and 8-bit GGUF models for CPU+GPU inference
- FPHam's original unquantised fp16 model in pytorch format, for GPU inference and for further conversions
Prompt template: ChatML
<|im_start|>system
{system_message}<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant
Known compatible clients / servers
These GPTQ models are known to work in the following inference servers/webuis.
This may not be a complete list; if you know of others, please let me know!
Provided files, and GPTQ parameters
Multiple quantisation parameters are provided, to allow you to choose the best one for your hardware and requirements.
Each separate quant is in a different branch. See below for instructions on fetching from different branches.
Most GPTQ files are made with AutoGPTQ. Mistral models are currently made with Transformers.
Explanation of GPTQ parameters
- Bits: The bit size of the quantised model.
- GS: GPTQ group size. Higher numbers use less VRAM, but have lower quantisation accuracy. "None" is the lowest possible value.
- 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. - Damp %: A GPTQ parameter that affects how samples are processed for quantisation. 0.01 is default, but 0.1 results in slightly better accuracy.
- 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).
- 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.
- ExLlama Compatibility: Whether this file can be loaded with ExLlama, which currently only supports Llama and Mistral models in 4-bit.
| Branch | Bits | GS | Act Order | Damp % | GPTQ Dataset | Seq Len | Size | ExLlama | Desc |
|---|---|---|---|---|---|---|---|---|---|
| main | 4 | 128 | Yes | 0.1 | wikitext | 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. |
| gptq-4bit-32g-actorder_True | 4 | 32 | Yes | 0.1 | wikitext | 4096 | 4.57 GB | Yes | 4-bit, with Act Order and group size 32g. Gives highest possible inference quality, with maximum VRAM usage. |
| gptq-8bit--1g-actorder_True | 8 | None | Yes | 0.1 | wikitext | 4096 | 7.52 GB | No | 8-bit, with Act Order. No group size, to lower VRAM requirements. |
| gptq-8bit-128g-actorder_True | 8 | 128 | Yes | 0.1 | wikitext | 4096 | 7.68 GB | No | 8-bit, with group size 128g for higher inference quality and with Act Order for even higher accuracy. |
| gptq-8bit-32g-actorder_True | 8 | 32 | Yes | 0.1 | wikitext | 4096 | 8.17 GB | No | 8-bit, with group size 32g and Act Order for maximum inference quality. |
| gptq-4bit-64g-actorder_True | 4 | 64 | Yes | 0.1 | wikitext | 4096 | 4.30 GB | Yes | 4-bit, with Act Order and group size 64g. Uses less VRAM than 32g, but with slightly lower accuracy. |
How to download, including from branches
In text-generation-webui
To download from the main branch, enter TheBloke/Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ in the "Download model" box.
To download from another branch, add :branchname to the end of the download name, eg TheBloke/Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ:gptq-4bit-32g-actorder_True
From the command line
I recommend using the huggingface-hub Python library:
pip3 install huggingface-hub
To download the main branch to a folder called Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ:
mkdir Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ
huggingface-cli download TheBloke/Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ --local-dir Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ --local-dir-use-symlinks False
To download from a different branch, add the --revision parameter:
mkdir Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ
huggingface-cli download TheBloke/Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ --revision gptq-4bit-32g-actorder_True --local-dir Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ --local-dir-use-symlinks False
More advanced huggingface-cli download usage
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.
The cache location can be changed with the HF_HOME environment variable, and/or the --cache-dir parameter to huggingface-cli.
For more documentation on downloading with huggingface-cli, please see: HF -> Hub Python Library -> Download files -> Download from the CLI.
To accelerate downloads on fast connections (1Gbit/s or higher), install hf_transfer:
pip3 install hf_transfer
And set environment variable HF_HUB_ENABLE_HF_TRANSFER to 1:
mkdir Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ
HF_HUB_ENABLE_HF_TRANSFER=1 huggingface-cli download TheBloke/Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ --local-dir Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ --local-dir-use-symlinks False
Windows Command Line users: You can set the environment variable by running set HF_HUB_ENABLE_HF_TRANSFER=1 before the download command.
With git (not recommended)
To clone a specific branch with git, use a command like this:
git clone --single-branch --branch gptq-4bit-32g-actorder_True https://huggingface.co/TheBloke/Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ
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.)
How to easily download and use this model in text-generation-webui
Please make sure you're using the latest version of text-generation-webui.
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.
Click the Model tab.
Under Download custom model or LoRA, enter
TheBloke/Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ.- To download from a specific branch, enter for example
TheBloke/Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ:gptq-4bit-32g-actorder_True - see Provided Files above for the list of branches for each option.
- To download from a specific branch, enter for example
Click Download.
The model will start downloading. Once it's finished it will say "Done".
In the top left, click the refresh icon next to Model.
In the Model dropdown, choose the model you just downloaded:
Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQThe model will automatically load, and is now ready for use!
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.
- 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.
- Note that you do not need to and should not set manual GPTQ parameters any more. These are set automatically from the file
Once you're ready, click the Text Generation tab and enter a prompt to get started!
Serving this model from Text Generation Inference (TGI)
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
Example Docker parameters:
--model-id TheBloke/Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ --port 3000 --quantize gptq --max-input-length 3696 --max-total-tokens 4096 --max-batch-prefill-tokens 4096
Example Python code for interfacing with TGI (requires huggingface-hub 0.17.0 or later):
pip3 install huggingface-hub
from huggingface_hub import InferenceClient
endpoint_url = "https://your-endpoint-url-here"
prompt = "Tell me about AI"
prompt_template=f'''<|im_start|>system
{system_message}<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant
'''
client = InferenceClient(endpoint_url)
response = client.text_generation(prompt,
max_new_tokens=128,
do_sample=True,
temperature=0.7,
top_p=0.95,
top_k=40,
repetition_penalty=1.1)
print(f"Model output: {response}")
Python code example: inference from this GPTQ model
Install the necessary packages
Requires: Transformers 4.33.0 or later, Optimum 1.12.0 or later, and AutoGPTQ 0.4.2 or later.
pip3 install --upgrade transformers optimum
# If using PyTorch 2.1 + CUDA 12.x:
pip3 install --upgrade auto-gptq
# or, if using PyTorch 2.1 + CUDA 11.x:
pip3 install --upgrade auto-gptq --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/
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:
pip3 uninstall -y auto-gptq
git clone https://github.com/PanQiWei/AutoGPTQ
cd AutoGPTQ
git checkout v0.5.1
pip3 install .
Example Python code
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
model_name_or_path = "TheBloke/Karen_TheEditor_V2_STRICT_Mistral_7B-GPTQ"
# To use a different branch, change revision
# For example: revision="gptq-4bit-32g-actorder_True"
model = AutoModelForCausalLM.from_pretrained(model_name_or_path,
device_map="auto",
trust_remote_code=False,
revision="main")
tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True)
prompt = "Tell me about AI"
prompt_template=f'''<|im_start|>system
{system_message}<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant
'''
print("\n\n*** Generate:")
input_ids = tokenizer(prompt_template, return_tensors='pt').input_ids.cuda()
output = model.generate(inputs=input_ids, temperature=0.7, do_sample=True, top_p=0.95, top_k=40, max_new_tokens=512)
print(tokenizer.decode(output[0]))
# Inference can also be done using transformers' pipeline
print("*** Pipeline:")
pipe = pipeline(
"text-generation",
model=model,
tokenizer=tokenizer,
max_new_tokens=512,
do_sample=True,
temperature=0.7,
top_p=0.95,
top_k=40,
repetition_penalty=1.1
)
print(pipe(prompt_template)[0]['generated_text'])
Compatibility
The files provided are tested to work with Transformers. For non-Mistral models, AutoGPTQ can also be used directly.
ExLlama is compatible with Llama and Mistral models in 4-bit. Please see the Provided Files table above for per-file compatibility.
For a list of clients/servers, please see "Known compatible clients / servers", above.
Discord
For further support, and discussions on these models and AI in general, join us at:
Thanks, and how to contribute
Thanks to the chirper.ai team!
Thanks to Clay from gpus.llm-utils.org!
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: 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
Thank you to all my generous patrons and donaters!
And thank you again to a16z for their generous grant.
Original model card: FPHam's Karen TheEditor V2 Strict Mistral 7B
Karen is an editor for your text. (v.2) STRICT edition
Ah, Karen, a true peach among grammatical cucumbers! She yearns to rectify the missteps and linguistic tangles that infest your horribly written fiction. Yet, unlike those ChatGPT kaboodles that morph into self-absorbed, constipated gurus of self-help style, Karen remains steadfastly grounded in grammatical wisdom but respectfull of your style.
Info
Karen, Version 2, uses a completely different data set and base model than the previous Karen.
There are two versions of Karen V2
- Strict (this one), in which Karen will try not to make too many changes to your original text, mostly fixing grammar and spelling, assuming that you know what you are doing.
- Creative (to be uploaded), in which Karen may suggest slight contextual improvements or rephrasing where necessary. It's Karen, after a glass of wine.
Goals
Karen's primary goal is to rectify grammatical and spelling errors in US English without altering the style of the text. She is adept at identifying and correcting common ESL errors.
Verb Tense Errors:
Incorrect use of verb tenses, such as using present tense when past tense is required and vice versa.
Confusion between continuous and simple tenses.
Subject-Verb Agreement:
Lack of agreement between the subject and verb in number, e.g., using a singular verb with a plural subject or vice versa.
Articles (a, an, the):
Incorrect use or omission of articles, such as using "a" instead of "an" or vice versa.
Overuse or omission of the definite article "the."
Prepositions:
Misuse of prepositions, such as using "in" instead of "on" or "at," or omitting prepositions where they are needed.
Word Order:
Incorrect word order in sentences, especially in questions and negative sentences.
Misplacement of adverbs or adjectives.
Pluralization:
Incorrect plural forms of nouns, such as failing to add "-s" or "-es" when necessary.
Pronoun Errors:
Confusion between subject and object pronouns.
Incorrect use of possessive pronouns.
Double Negatives:
Using double negatives, which is grammatically incorrect in standard English.
Modal Verbs:
Misuse of modal verbs like can, could, will, would, should, etc.
Confusing Similar Words:
Confusing words that sound similar but have different meanings and spellings (e.g., "their," "there," and "they're").
Lack of Plural/Singular Agreement:
Mistakes in matching singular and plural nouns and verbs in a sentence.
Future Goals
Use bigger model, add grammar cases that the model misses. Better datasets. Use larger datasets.
Training
It was reversely trained on fict/non-fiction US text where errors were intentionally inserted by another Llama model (Darth Karen) and Python script.
Usage
It should be used by submitting a paragraph or block of text at a time.
Model uses ChatML
<|im_start|>system
<|im_end|>
<|im_start|>user
Edit the following text for spelling and grammar mistakes: {paragraph of text} <|im_end|>
<|im_start|>assistant
Note the pretext: Edit the following text for spelling and grammar mistakes: before the actual text. This way Karen wouldn't start talking ABOUT the text.
Recomended settings
- Temperature: 0.7
- top_p: 0.1
- top_k: 40
- repetition penalty: 1.18
Karen could also be used in chatting. But if Karen is involved in a conversation where the text is lengthy, she might interpret your message as needing proofreading assistance rather than simply acknowledging that you are using her in casual chat.
Example:
Edit the following text for spelling and grammar mistakes: I see future writing more about you writing the story any way you can and LLM editing it and making it into a proper fiction. That seems to be more satisfying way where you could feel you actually had done something creATIVE. So human is the creative, LLM is the burger flipping intern. The opposite is what many people are trying rn. Make LLM to invent story (from premise, to creating outline to writing actual chapters) and the human to somehow assemble it from those pieces. So LLM is the creative man, and human is the floor polisher. I can't believe a normal person would find it satisfying long term. So in my view, the job of editor would be replaced by Ai, not the job of author.
(Grammarly Score: 48)
Output STRICT:
I see the future of writing being more about you writing the story any way you can and having an AI like LLM edit it and make it into a proper fiction. That seems to be a more satisfying way where you could feel you actually did something creative. So humans are the creative ones, while LLMs are the burger-flipping interns.
On the other hand, many people are currently trying to do the exact opposite. They want to make LLMs create stories (from premises, to outlining, to writing entire chapters). And then they want humans to somehow piece them together from these parts. In this scenario, LLMs become the creative individuals, and humans become the floor polishers.
I can't imagine a regular person finding that satisfying over the long run.
In my opinion, the role of an editor will eventually be taken over by AI, not the role of an author.
(Grammarly Score: 93)
Output CREATIVE:
I see the future of writing being more focused on you writing the story any way you can and having an AI like GPT-3 edit it and turn it into a proper piece of fiction. This seems to be a more satisfying approach where you can truly feel that you have done something creative. In this scenario, humans remain the creative force, while the AI acts as the burger flipping intern.
On the other hand, there are those who argue that we should let the AI generate stories from scratch - from coming up with a premise, to outlining the plot, to writing individual chapters. They believe that the role of the human should then become one of assembling these pieces together. In this case, the AI becomes the creative mind, and the human takes on the role of a floor polisher.
I cannot imagine how someone could find this satisfying over the long term. Therefore, in my opinion, the job of the editor will eventually be taken over by AI, not the job of the author.
(Grammarly Score: 83)
Conclusion
After probably 10 different versions with subsequent changes, I can now say that the current model works reasonably well, with occasional (but often debatable) grammar misses. The limitations seem to be related to the 7B parameters. It appears that the size isn't sufficient to have a fine-grained understanding of various nuances of the input. This correlates with my other findings - the Mistral model performs quite well when generating its own text, but its comprehension is less than perfect, again related to only 7B parameters.
The goal was to create a model that wouldn't change the style of the text. Often, LLM models, when asked to edit text, will attempt to rewrite the text even if the text is already fine. This proved to be quite challenging for such a small model where the main task was to determine the right balance between fixing the text (and not changing its style) and copying it verbatim.
The strict model assumes that you're already a good writer that doesn't need hand-holding and that every word you've written you've meant.
- Downloads last month
- 7