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.
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.
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.
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.)
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.
shell
1pip3 install --upgrade transformers optimum
2# If using PyTorch 2.1 + CUDA 12.x:3pip3 install --upgrade auto-gptq
4# or, if using PyTorch 2.1 + CUDA 11.x:5pip3 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:
1from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
23model_name_or_path ="TheBloke/medicine-chat-GPTQ"4# To use a different branch, change revision5# For example: revision="gptq-4bit-32g-actorder_True"6model = AutoModelForCausalLM.from_pretrained(model_name_or_path,7 device_map="auto",8 trust_remote_code=False,9 revision="main")1011tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True)1213prompt ="Write a story about llamas"14system_message ="You are a story writing assistant"15prompt_template=f'''[INST] <<SYS>>
16{system_message}17<</SYS>>
18{prompt} [/INST]
19'''2021print("\n\n*** Generate:")2223input_ids = tokenizer(prompt_template, return_tensors='pt').input_ids.cuda()24output = model.generate(inputs=input_ids, temperature=0.7, do_sample=True, top_p=0.95, top_k=40, max_new_tokens=512)25print(tokenizer.decode(output[0]))2627# Inference can also be done using transformers' pipeline2829print("*** Pipeline:")30pipe = pipeline(31"text-generation",32 model=model,33 tokenizer=tokenizer,34 max_new_tokens=512,35 do_sample=True,36 temperature=0.7,37 top_p=0.95,38 top_k=40,39 repetition_penalty=1.140)4142print(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 architecture models (including Mistral, Yi, DeepSeek, SOLAR, etc) 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:
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 special mentions: Michael Levine, 阿明, Trailburnt, Nikolai Manek, John Detwiler, Randy H, Will Dee, Sebastain Graf, NimbleBox.ai, Eugene Pentland, Emad Mostaque, Ai Maven, Jim Angel, Jeff Scroggin, Michael Davis, Manuel Alberto Morcote, Stephen Murray, Robert, Justin Joy, Luke @flexchar, Brandon Frisco, Elijah Stavena, S_X, Dan Guido, Undi ., Komninos Chatzipapas, Shadi, theTransient, Lone Striker, Raven Klaugh, jjj, Cap'n Zoog, Michel-Marie MAUDET (LINAGORA), Matthew Berman, David, Fen Risland, Omer Bin Jawed, Luke Pendergrass, Kalila, OG, Erik Bjäreholt, Rooh Singh, Joseph William Delisle, Dan Lewis, TL, John Villwock, AzureBlack, Brad, Pedro Madruga, Caitlyn Gatomon, K, jinyuan sun, Mano Prime, Alex, Jeffrey Morgan, Alicia Loh, Illia Dulskyi, Chadd, transmissions 11, fincy, Rainer Wilmers, ReadyPlayerEmma, knownsqashed, Mandus, biorpg, Deo Leter, Brandon Phillips, SuperWojo, Sean Connelly, Iucharbius, Jack West, Harry Royden McLaughlin, Nicholas, terasurfer, Vitor Caleffi, Duane Dunston, Johann-Peter Hartmann, David Ziegler, Olakabola, Ken Nordquist, Trenton Dambrowitz, Tom X Nguyen, Vadim, Ajan Kanaga, Leonard Tan, Clay Pascal, Alexandros Triantafyllidis, JM33133, Xule, vamX, ya boyyy, subjectnull, Talal Aujan, Alps Aficionado, wassieverse, Ari Malik, James Bentley, Woland, Spencer Kim, Michael Dempsey, Fred von Graf, Elle, zynix, William Richards, Stanislav Ovsiannikov, Edmond Seymore, Jonathan Leane, Martin Kemka, usrbinkat, Enrico Ros
Thank you to all my generous patrons and donaters!
And thank you again to a16z for their generous grant.
We explore continued pre-training on domain-specific corpora for large language models. While this approach enriches LLMs with domain knowledge, it significantly hurts their prompting ability for question answering. Inspired by human learning via reading comprehension, we propose a simple method to transform large-scale pre-training corpora into reading comprehension texts, consistently improving prompting performance across tasks in biomedicine, finance, and law domains. Our 7B model competes with much larger domain-specific models like BloombergGPT-50B.
🤗 We are currently working hard on developing models across different domains, scales and architectures! Please stay tuned! 🤗
In our paper, we develop three domain-specific models from LLaMA-1-7B, which are also available in Huggingface: Biomedicine-LLM, Finance-LLM and Law-LLM, the performances of our AdaptLLM compared to other domain-specific LLMs are:
LLaMA-1-13B
Moreover, we scale up our base model to LLaMA-1-13B to see if our method is similarly effective for larger-scale models, and the results are consistently positive too: Biomedicine-LLM-13B, Finance-LLM-13B and Law-LLM-13B.
Domain-Specific LLaMA-2-Chat
Our method is also effective for aligned models! LLaMA-2-Chat requires a specific data format, and our reading comprehension can perfectly fit the data format by transforming the reading comprehension into a multi-turn conversation. We have also open-sourced chat models in different domains: Biomedicine-Chat, Finance-Chat and Law-Chat
For example, to chat with the biomedicine-chat model:
python
1from transformers import AutoModelForCausalLM, AutoTokenizer
23model = AutoModelForCausalLM.from_pretrained("AdaptLLM/medicine-chat")4tokenizer = AutoTokenizer.from_pretrained("AdaptLLM/medicine-chat")56# Put your input here:7user_input ='''Question: Which of the following is an example of monosomy?
8Options:
9- 46,XX
10- 47,XXX
11- 69,XYY
12- 45,X
1314Please provide your choice first and then provide explanations if possible.'''1516# Apply the prompt template and system prompt of LLaMA-2-Chat demo for chat models (NOTE: NO prompt template is required for base models!)17our_system_prompt ="\nYou are a helpful, respectful and honest assistant. Always answer as helpfully as possible, while being safe. Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure that your responses are socially unbiased and positive in nature.\n\nIf a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don't know the answer to a question, please don't share false information.\n"# Please do NOT change this18prompt =f"<s>[INST] <<SYS>>{our_system_prompt}<</SYS>>\n\n{user_input} [/INST]"1920# # NOTE:21# # If you want to apply your own system prompt, please integrate it into the instruction part following our system prompt like this:22# your_system_prompt = "Please, answer this question faithfully."23# prompt = f"<s>[INST] <<SYS>>{our_system_prompt}<</SYS>>\n\n{your_system_prompt}\n{user_input} [/INST]"2425inputs = tokenizer(prompt, return_tensors="pt", add_special_tokens=False).input_ids.to(model.device)26outputs = model.generate(input_ids=inputs, max_length=4096)[0]2728answer_start =int(inputs.shape[-1])29pred = tokenizer.decode(outputs[answer_start:], skip_special_tokens=True)3031print(f'### User Input:\n{user_input}\n\n### Assistant Output:\n{pred}')
Domain-Specific Tasks
To easily reproduce our results, we have uploaded the filled-in zero/few-shot input instructions and output completions of each domain-specific task: biomedicine-tasks, finance-tasks, and law-tasks.
Note: those filled-in instructions are specifically tailored for models before alignment and do NOT fit for the specific data format required for chat models.
Citation
If you find our work helpful, please cite us:
bibtex
1@article{adaptllm,
2 title = {Adapting Large Language Models via Reading Comprehension},
3 author = {Daixuan Cheng and Shaohan Huang and Furu Wei},
4 journal = {CoRR},
5 volume = {abs/2309.09530},
6 year = {2023}
7}
