These are SuperHOT GGMLs with an increased context length. SuperHOT is a new system that employs RoPE to expand context beyond what was originally possible for a model. It was discovered and developed by kaiokendev.
In order to use the increased context length, you can presently use:
Support is also expected to come to llama.cpp, however work is still being done to find the optimal implementation.
To use the increased context with KoboldCpp, simply use --contextsize to set the desired context, eg --contextsize 4096 or --contextsize 8192.
NOTE: Increased context length is an area seeing rapid developments and improvements. It is quite possible that these models may be superseded by new developments in the coming days. If that's the case, I will remove them, or update this README as appropriate.
These GGMLs will work with any llama.cpp-compatible GGML client that supports k-quants.
However the increased context length won't work without specific support. See the note in the introduction for details on using increased context.
Explanation of the new k-quant methods
The new methods available are:
GGML_TYPE_Q2_K - "type-1" 2-bit quantization in super-blocks containing 16 blocks, each block having 16 weight. Block scales and mins are quantized with 4 bits. This ends up effectively using 2.5625 bits per weight (bpw)
GGML_TYPE_Q3_K - "type-0" 3-bit quantization in super-blocks containing 16 blocks, each block having 16 weights. Scales are quantized with 6 bits. This end up using 3.4375 bpw.
GGML_TYPE_Q4_K - "type-1" 4-bit quantization in super-blocks containing 8 blocks, each block having 32 weights. Scales and mins are quantized with 6 bits. This ends up using 4.5 bpw.
GGML_TYPE_Q5_K - "type-1" 5-bit quantization. Same super-block structure as GGML_TYPE_Q4_K resulting in 5.5 bpw
GGML_TYPE_Q6_K - "type-0" 6-bit quantization. Super-blocks with 16 blocks, each block having 16 weights. Scales are quantized with 8 bits. This ends up using 6.5625 bpw
GGML_TYPE_Q8_K - "type-0" 8-bit quantization. Only used for quantizing intermediate results. The difference to the existing Q8_0 is that the block size is 256. All 2-6 bit dot products are implemented for this quantization type.
Refer to the Provided Files table below to see what files use which methods, and how.
Provided files
Name
Quant method
Bits
Size
Max RAM required
Use case
selfee-13b-superhot-8k.ggmlv3.q4_0.bin
q4_0
4
7.32 GB
9.82 GB
Original llama.cpp quant method, 4-bit.
selfee-13b-superhot-8k.ggmlv3.q4_1.bin
q4_1
4
8.14 GB
10.64 GB
Original llama.cpp quant method, 4-bit. Higher accuracy than q4_0 but not as high as q5_0. However has quicker inference than q5 models.
selfee-13b-superhot-8k.ggmlv3.q5_0.bin
q5_0
5
8.95 GB
11.45 GB
Original llama.cpp quant method, 5-bit. Higher accuracy, higher resource usage and slower inference.
selfee-13b-superhot-8k.ggmlv3.q5_1.bin
q5_1
5
9.76 GB
12.26 GB
Original llama.cpp quant method, 5-bit. Even higher accuracy, resource usage and slower inference.
selfee-13b-superhot-8k.ggmlv3.q8_0.bin
q8_0
8
13.83 GB
16.33 GB
Original llama.cpp quant method, 8-bit. Almost indistinguishable from float16. High resource use and slow. Not recommended for most users.
Note: the above RAM figures assume no GPU offloading. If layers are offloaded to the GPU, this will reduce RAM usage and use VRAM instead.
How to run in koboldcpp
On Linux I use the following command line to launch the KoboldCpp UI with OpenCL aceleration and a context size of 4096:
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.
Special thanks to: Luke from CarbonQuill, Aemon Algiz.
Patreon special mentions: RoA, Lone Striker, Gabriel Puliatti, Derek Yates, Randy H, Jonathan Leane, Eugene Pentland, Karl Bernard, Viktor Bowallius, senxiiz, Daniel P. Andersen, Pierre Kircher, Deep Realms, Cory Kujawski, Oscar Rangel, Fen Risland, Ajan Kanaga, LangChain4j, webtim, Nikolai Manek, Trenton Dambrowitz, Raven Klaugh, Kalila, Khalefa Al-Ahmad, Chris McCloskey, Luke @flexchar, Ai Maven, Dave, Asp the Wyvern, Sean Connelly, Imad Khwaja, Space Cruiser, Rainer Wilmers, subjectnull, Alps Aficionado, Willian Hasse, Fred von Graf, Artur Olbinski, Johann-Peter Hartmann, WelcomeToTheClub, Willem Michiel, Michael Levine, Iucharbius , Spiking Neurons AB, K, biorpg, John Villwock, Pyrater, Greatston Gnanesh, Mano Prime, Junyu Yang, Stephen Murray, John Detwiler, Luke Pendergrass, terasurfer , Pieter, zynix , Edmond Seymore, theTransient, Nathan LeClaire, vamX, Kevin Schuppel, Preetika Verma, ya boyyy, Alex , SuperWojo, Ghost , Joseph William Delisle, Matthew Berman, Talal Aujan, chris gileta, Illia Dulskyi.
Thank you to all my generous patrons and donaters!
Original model card: Kaio Ken's SuperHOT 8K
SuperHOT Prototype 2 w/ 8K Context
This is a second prototype of SuperHOT, a NSFW focused LoRA, this time 7B with 8K context and no RLHF, using the same technique described in the github blog.
Looking for Merged & Quantized Models?
Make some please :)
Using the monkey-patch?
You will NEED to apply the monkeypatch or, if you are already using the monkeypatch, change the scaling factor to 0.25 and the maximum sequence length to 8192
The monkeypatch is only necessary if you are using a front-end/back-end that does not already support scaling and said front-end/back-end is Python-based (i.e. Huggingface Transformers). To apply the patch, you will need to copy the llama_rope_scaled_monkey_patch.py into your working directory and call the exported function replace_llama_rope_with_scaled_rope at the very start of your Python program. It will modify the Transformers library's implementation of RoPE to properly apply the scaling factor.
Using Oobabooga with Exllama?
Switch your loader to exllama or exllama_hf Add the arguments max_seq_len 8192 and compress_pos_emb 4. While the model may work well with compress_pos_emb 2, it was trained on 4, so that is what I advocate for you to use
In the UI, you will see the loader option in the Models tab. Once you select either exllama or exllama_hf, the max_seq_len and compress_pos_emb settings will appear.
Training Details
I trained the LoRA with the following configuration:
1200 samples (~400 samples over 2048 sequence length)
learning rate of 3e-4
3 epochs
The exported modules are:
q_proj
k_proj
v_proj
o_proj
no bias
Rank = 4
Alpha = 8
no dropout
weight decay of 0.1
AdamW beta1 of 0.9 and beta2 0.99, epsilon of 1e-5
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.
Special thanks to: Luke from CarbonQuill, Aemon Algiz, Dmitriy Samsonov.
Patreon special mentions: Derek Yates, Sean Connelly, Luke, Nathan LeClaire, Trenton Dambrowitz, Mano Prime, David Flickinger, vamX, Nikolai Manek, senxiiz, Khalefa Al-Ahmad, Illia Dulskyi, trip7s trip, Jonathan Leane, Talal Aujan, Artur Olbinski, Cory Kujawski, Joseph William Delisle, Pyrater, Oscar Rangel, Lone Striker, Luke Pendergrass, Eugene Pentland, Johann-Peter Hartmann.
Thank you to all my generous patrons and donaters!
[May 31, 2023] Initial release: We released the first version of SelFee! Check out the blog post for more details.
Overview
This is the repository for the KAIST SelFee project, which aims to build and share an instruction-following LLaMA model. This repo mainly has five contents:
The selection process of the 178K training data for SelFee (detail, code).
The generation process for the training data and its result. (detail, code).
The training process for the model (detail, code).
The inference process for the model (detail, code).
This repository is based on the Stanford-Alpaca and Vicuna repository. Thanks to all the contributors for these awesome repositories!! 🙌
We highly recommend you read our blog post for more details about the model.
Data Release
For data collection, we collected datasets from five different fields. These are the Stanford Alpaca dataset, math collection, code collection, Flan collection, and ShareGPT. We provide code that we used to make a dataset for training. We also provide code how we preprocessed ShareGPT. For ShareGPT, we only use the first (question, answer) pair from human and GPT, respectively. We only use instances which are classified as english,and filter instance which is not a form of question.
For other datsets, we do not need special data collection method.
Data Generation Process
To train our model with high-quality instructions and answer pairs, we utilized data augmentation using OpenAI API calls. The process involved three steps.
Firstly, we collected various instructions from multiple fields and fed them to ChatGPT to generate answers.
Secondly, we gathered feedback on the generated answer by querying ChatGPT again and asked it to determine if the initial answer required any revision.
Thirdly, if a revision was necessary, we passed the instruction, initial answer, and feedback pair to ChatGPT to generate a revised answer and its feedback pair.
We repeated the process until we received feedback that required no further revision or hit the maximum iteration. However, due to the token limitation of the ChatGPT API, we had to truncate some instances that needed more than 4096 tokens while augmenting.
You can see the details with command here.
*We provide the whole dataset after collection and augmentation using huggingface(code), so you can either use the code or follow our data merging step to replicate the training dataset. Feel free to use any of them!
Training
We utilize FastChat to train the model. Given the instruction, we fine-tune the model to generate the answer and feedback chain (including the revisions).
To reproduce the training procedure, here are the steps.
The hyperparameters are as follows, following Vicuna and Alpaca.
Hyperparameter
Global Batch Size
Learning rate
Epochs
Max length
Weight decay
SelFee (7B, 13B)
128
2e-5
3
2048
0
Inference
Restoring checkpoint using diff
We provide diff weight and code which can restore the same model with SelFee. To restore the original SelFee weight, you first need to convert the Meta's original LLAMA checkpoint into huggingface format into your local machine. Once you are done, you can restore the same checkpoint of our model by using the following command
Because SelFee is trained to generate iterative feedback and revisions until the response is satisfying, it automatically generates iterative feedback and revisions on a single forward pass. The model autonomously decides when to stop generating revisions based on the feedback. If the feedback chain ends with sequences like Revision is not needed., the model autonomously terminates generation.
Revision Enforce Inference Mode
We observed that increasing the minimum number of required revisions corresponds to a corresponding increase in performance. To enforce revisions, we automatically replace sequences such as Revision is not needed. into Revision is needed. during self-feedback generation. Because SelFee is trained to generate Revision {index}: after the sequence of Revision is needed., the model would continually revise the answer.
For revision enforce inference mode, use the max-num-revision argument.
Following evaluation setting of Vicuna, we evaluate on 80 diverse queries and utilize GPT-4 language model as the evaluator, scoring a model's response relative to ChatGPT's response. One of the difference with Vicuna evaluation is that due to positional bias of GPT-4, we employ a bidirectional evaluation setting. This means that each evaluation instance is inferred twice, depending on its position.
We release the inference result of SelFee in the folder of evaluation/answer and also the scores generated by GPT-4 in the folder of evaluation/review.
GPT-4 Automatic Evaluation
First, you need to get your API key to get access to the GPT-4 API.
export OPENAI_API_KEYS={personal_key}
To compare the performance of a generation result (for example, located on evaluation/answer/file_A.jsonl) with another generation result (located on evaluation/anwer/file_B.jsonl),
To mitigate the positional bias of GPT-4 model, we apply a bidirectional evaluation setting. Therefore, automatic evaluation with opposite position is also needed.
Similar to other LLaMA-finetuned models, SelFee also make some mistakes especially for math, reasoning, factuality, and coding tasks. Although our performance outperforms ChatGPT on Vicuna setting, the evaluation setting contains some limitations in terms of comprehension (limited to 80 queries), inconsistency, and unreliability. Therefore, further research for a better evaluation setting is needed. Please take these claims with a grain of salt.
We have released the SelFee-7B and SelFee-13B model diff weights, which can be found with instructions here. Moreover, the training instances used to train SelFee is released on huggingface.
License
The research preview online demo is only for non-commercial use and is subject to various licenses and terms of use, including the LLaMA model License, OpenAI's Terms of Use for the generated data, and ShareGPT's Privacy Practices. If you suspect any violations, please reach out to us.
Citation
Please cite if you use the data or code in this repo.
@misc{selfee2023,
author = {Ye, Seonghyeon and Jo, Yongrae and Kim, Doyoung and Kim, Sungdong and Hwang, Hyeonbin and Seo, Minjoon},
title = {SelFee: Iterative Self-Revising LLM Empowered by Self-Feedback Generation},
url = {https://kaistai.github.io/SelFee/},
month = {May},
year = {2023},
howpublished = {Blog post}
}
