This repo contains fp16 pytorch format model files for Kaist AI's Selfee 13B.

It is the result of merging the diff at the above repo with base Llama 13B, then converting fp32 to fp16.

For further support, and discussions on these models and AI in general, join us at:

Thanks to the chirper.ai team!

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: Aemon Algiz.

Patreon special mentions: Sam, theTransient, Jonathan Leane, Steven Wood, webtim, Johann-Peter Hartmann, Geoffrey Montalvo, Gabriel Tamborski, Willem Michiel, John Villwock, Derek Yates, Mesiah Bishop, Eugene Pentland, Pieter, Chadd, Stephen Murray, Daniel P. Andersen, terasurfer, Brandon Frisco, Thomas Belote, Sid, Nathan LeClaire, Magnesian, Alps Aficionado, Stanislav Ovsiannikov, Alex, Joseph William Delisle, Nikolai Manek, Michael Davis, Junyu Yang, K, J, Spencer Kim, Stefan Sabev, Olusegun Samson, transmissions 11, Michael Levine, Cory Kujawski, Rainer Wilmers, zynix, Kalila, Luke @flexchar, Ajan Kanaga, Mandus, vamX, Ai Maven, Mano Prime, Matthew Berman, subjectnull, Vitor Caleffi, Clay Pascal, biorpg, alfie_i, 阿明, Jeffrey Morgan, ya boyyy, Raymond Fosdick, knownsqashed, Olakabola, Leonard Tan, ReadyPlayerEmma, Enrico Ros, Dave, Talal Aujan, Illia Dulskyi, Sean Connelly, senxiiz, Artur Olbinski, Elle, Raven Klaugh, Fen Risland, Deep Realms, Imad Khwaja, Fred von Graf, Will Dee, usrbinkat, SuperWojo, Alexandros Triantafyllidis, Swaroop Kallakuri, Dan Guido, John Detwiler, Pedro Madruga, Iucharbius, Viktor Bowallius, Asp the Wyvern, Edmond Seymore, Trenton Dambrowitz, Space Cruiser, Spiking Neurons AB, Pyrater, LangChain4j, Tony Hughes, Kacper Wikieł, Rishabh Srivastava, David Ziegler, Luke Pendergrass, Andrey, Gabriel Puliatti, Lone Striker, Sebastain Graf, Pierre Kircher, Randy H, NimbleBox.ai, Vadim, danny, Deo Leter

Thank you to all my generous patrons and donaters!

And thank you again to a16z for their generous grant.

[May 31, 2023] Initial release: We released the first version of SelFee! Check out the blog post for more details.

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:

This repository is based on the Stanford-Alpaca and Vicuna repository. Thanks to all the contributors for these awesome repositories!! 🙌

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.

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!

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.

torchrun --nproc_per_node=4 train/train_mem.py \
    --model_name_or_path llama-7b \
    --data_path outputs/feedback_gpt_3.5_turbo_merged_whole.json \
    --bf16 True \
    --output_dir ckpt/selfee-7b \
    --num_train_epochs 3 \
    --per_device_train_batch_size 16 \
    --per_device_eval_batch_size 16 \
    --gradient_accumulation_steps 2 \
    --evaluation_strategy "no" \
    --save_strategy "steps" \
    --save_steps 5000 \
    --save_total_limit 1 \
    --learning_rate 2e-5 \
    --weight_decay 0. \
    --warmup_ratio 0.03 \
    --lr_scheduler_type "cosine" \
    --logging_steps 1 \
    --fsdp "shard_grad_op auto_wrap" \
    --fsdp_transformer_layer_cls_to_wrap 'LlamaDecoderLayer' \
    --tf32 True \
    --model_max_length 2048 \
    --gradient_checkpointing True \
    --lazy_preprocess True \
    --training_objective full \

The hyperparameters are as follows, following Vicuna and Alpaca.

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.

For autonomous inference mode,

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.

First, you need to get your API key to get access to the GPT-4 API.

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.

Check out the demo!

To serve the web demo yourself, run the following commands:

You can find the serving code here.

Seonghyeon Ye*, Yongrae Jo*, Doyoung Kim*, Sungdong Kim, Hyeonbin Hwang, and Minjoon Seo.
(* denotes equal contribution)

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.

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.

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}
}