Skip to content
/ Lion Public

Code for "Lion: Adversarial Distillation of Proprietary Large Language Models (EMNLP 2023)"

License

Notifications You must be signed in to change notification settings

YJiangcm/Lion

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Lion: Adversarial Distillation of Proprietary Large Language Models (EMNLP 2023)

Lion

[📄 Paper] | [🤗 Lion Weights]


Code License Data License Weight Diff License Python Pytorch

News

  • [October 8, 2023] Our paper has been accepted to EMNLP 2023.
  • [June 10, 2023] We released insturctions for addressing OOM during fine-tuning, check it in Training Process.
  • [May 26, 2023] We released the model weights. Check out the 7B model!
  • [May 25, 2023] We released an online demo, try our model here!
  • [May 23, 2023] We released the code for training and inference.

Contents

  1. Overview
  1. Recovering Lion weights

  2. Inference

  3. Training Process

  4. Evaluation

  5. Citation

  6. Disclaimer

Overview

The high-level overview of our adversarial distillation framework, where we craft a compact Student LLM based on a superior closed-source LLM that serves three roles: the Teacher, the Referee, and the Generator. From left to right, there are three stages in an iteration:

  1. an imitation stage to align the student’s response with the teacher’s response;
  2. a discrimination stage to identify hard samples;
  3. a generation stage to produce new hard samples for escalating the challenges presented to the student model.

Recovering Lion weights

We release Lion weights as delta weights to comply with the LLaMA model license.

You can add our delta to the original LLaMA weights to obtain the Lion weights. Instructions:

  1. Get the original LLaMA weights in the huggingface format by following the instructions here
  2. Please download our delta model from Hugging Face
  3. Use the following scripts to get Lion weights by applying our delta:
python src/weight_diff.py recover --path_raw huggyllama/llama-7b --path_diff YuxinJiang/lion-7b --path_tuned <path_to_store_recovered_weights>

Inference

For inference and training of Lion, please first install the requirements:

pip install -r requirements.txt

We provide the decoding script for Lion, which reads a input file and generates corresponding responses for each sample, and finally consolidates them into an output file. It can be run on a single machine with 16GB GPU.

python src/lion_inference.py \
    --model_dir <path_to_hf_converted_lion_ckpt_and_tokenizer> \
    --data_dir <path_to_input_json_file> \
    --output_dir <path_to_output_json_file> \
    --num_gpus 1

Training Process

Below shows one iteration of our adversarial distillation framework.

1. Imitation Stage

1.1 Acquire the teacher's response on the Train Pool

python src/chatgpt_inference.py \
    -q <path_to_json_file_for_the_Train_Pool> \
    -o <path_to_chatgpt_inference_for_the_Train_Pool> \
    --api_key <your_openai_api_key>

1.2 Instruction-tuning the student based on the teacher’s response on the Train Pool

Fine-tuning was conducted on a machine with 8 A100 80G GPUs.

torchrun --nproc_per_node=8 --master_port=<your_random_port> src/train.py \
    --model_name_or_path <path_to_hf_converted_ckpt_and_tokenizer> \
    --data_path <path_to_chatgpt_inference_for_the_Train_Pool> \
    --bf16 True \
    --output_dir result \
    --num_train_epochs 3 \
    --model_max_length 1024 \
    --per_device_train_batch_size 2 \
    --per_device_eval_batch_size 2 \
    --gradient_accumulation_steps 8 \
    --evaluation_strategy "no" \
    --save_strategy "steps" \
    --save_steps 600 \
    --save_total_limit 1 \
    --learning_rate 2e-5 \
    --weight_decay 0. \
    --warmup_ratio 0.03 \
    --lr_scheduler_type "cosine" \
    --logging_steps 1 \
    --fsdp "full_shard auto_wrap" \
    --fsdp_transformer_layer_cls_to_wrap 'LlamaDecoderLayer' \
    --tf32 True

Addressing OOM

Naively, fine-tuning a 7B model requires about 7 x 8 x 2 = 112 GB of VRAM. Commands given above enable parameter sharding, so no redundant model copy is stored on any GPU. If you'd like to further reduce the memory footprint, here are some options:

  • Turn on CPU offload for FSDP with --fsdp "full_shard auto_wrap offload". This saves VRAM at the cost of longer runtime.

  • In our experience, DeepSpeed stage-3 (with offload) can at times be more memory efficient than FSDP with offload. Here's an example to use DeepSpeed stage-3 with 8 GPUs with both parameter and optimizer offload:

    deepspeed src/train_deepspeed.py \
        --model_name_or_path <path_to_hf_converted_ckpt_and_tokenizer> \
        --data_path <path_to_chatgpt_inference_for_the_Train_Pool> \
        --output_dir result \
        --num_train_epochs 3 \
        --model_max_length 1024 \
        --per_device_train_batch_size 16 \
        --per_device_eval_batch_size 1 \
        --gradient_accumulation_steps 1 \
        --evaluation_strategy "no" \
        --save_strategy "steps" \
        --save_steps 600 \
        --save_total_limit 1 \
        --learning_rate 2e-5 \
        --warmup_ratio 0.03 \
        --logging_steps 1 \
        --lr_scheduler_type "cosine" \
        --report_to "tensorboard" \
        --gradient_checkpointing True \
        --deepspeed srcs/configs/deepspeed_config.json \
        --fp16 True
    • The DeepSpeed library also provides some helpful functions to estimate memory usage.
  • LoRA fine-tunes low-rank slices of the query, key, and value embedding heads. This can reduce the total memory footprint from 112GB to about 7x4=28GB. We may release our re-implemention of this in the future, but for now the peft codebase can be a useful resource.

2. Discrimination Stage

2.1 Acquire the teacher's response on the Cache Pool

python src/chatgpt_inference.py \
    -q <path_to_json_file_for_the_Cache_Pool> \
    -o <path_to_chatgpt_inference_for_the_Cache_Pool> \
    --api_key <your_openai_api_key>

2.2 Acquire the student's response on the Cache Pool

python src/lion_inference.py \
    --model_dir <path_to_hf_converted_lion_ckpt_and_tokenizer> \
    --data_dir <path_to_json_file_for_the_Cache_Pool> \
    --output_dir <path_to_lion_inference_for_the_Cache_Pool> \
    --num_gpus 8

2.3 Ask the referee to output two scores according to the respose quality of the teacher and the student

To mitigate the position bias of the LLM referee, we conduct two runs by exchanging the positions of the teacher's response and the student's response.

python src/chatgpt_referee.py \
    -a <path_to_chatgpt_inference_for_the_Cache_Pool> <path_to_lion_inference_for_the_Cache_Pool> \
    -o <path_to_output_review_chatgpt_lion_file> \
    --api_key <your_openai_api_key>
python src/chatgpt_referee.py \
    -a <path_to_lion_inference_for_the_Cache_Pool> <path_to_chatgpt_inference_for_the_Cache_Pool> \
    -o <path_to_output_review_lion_chatgpt_file> \
    --api_key <your_openai_api_key>

2.4 Discriminate hard instructions and easy instructions

python src/discrimination.py \
    --review12_path <path_to_output_review_chatgpt_lion_file> \
    --review21_path <path_to_output_review_lion_chatgpt_file> \
    --chatgpt_inference_path <path_to_chatgpt_inference_for_the_Cache_Pool> \
    --lion_inference_path <path_to_lion_inference_for_the_Cache_Pool> \
    --hard_save_path <path_to_identified_hard_instructions> \
    --easy_save_path <path_to_identified_easy_instructions>

3. Generation Stage

3.1 Generate new hard instructions

python -m src/generate_hard_instruction generate_instruction_following_data \
    --seed_tasks_path <path_to_identified_hard_instructions> \
    --all_tasks_path <path_to_json_file_for_the_Cache_Pool> \
    --output_dir <path_to_generated_hard_instructions> \
    --num_instructions_to_generate 3000 \
    --api_key <your_openai_api_key>

3.2 Generate new easy instructions

python -m src/generate_easy_instruction generate_instruction_following_data \
    --seed_tasks_path <path_to_identified_easy_instructions> \
    --all_tasks_path <path_to_json_file_for_the_Cache_Pool> \
    --output_dir <path_to_generated_easy_instructions> \
    --num_instructions_to_generate 3000 \
    --api_key <your_openai_api_key>

Evaluation

Results for Open-ended Generation Dataset

We leverage GPT-4 to automatically assess the quality of responses (rated on a scale of 1 to 10) between a reference model (ChatGPT) and a candidate model. Subsequently, we calculate the candidate model’s performance as the percentage of the total score it achieves compared to the reference model.

Results for Reasoning Dataset

Citation

Please cite our paper if you use the code in this repo.

@inproceedings{jiang-etal-2023-lion,
    title = "Lion: Adversarial Distillation of Proprietary Large Language Models",
    author = "Jiang, Yuxin  and
      Chan, Chunkit  and
      Chen, Mingyang  and
      Wang, Wei",
    editor = "Bouamor, Houda  and
      Pino, Juan  and
      Bali, Kalika",
    booktitle = "Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing",
    month = dec,
    year = "2023",
    address = "Singapore",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.emnlp-main.189",
    doi = "10.18653/v1/2023.emnlp-main.189",
    pages = "3134--3154",
}

Disclaimer

⚠️ Lion is intended and licensed for research use ONLY. Commercial use is strictly prohibited. The content produced by any version of Lion is influenced by uncontrollable variables such as randomness, and therefore, the accuracy of the output cannot be guaranteed by this project. This project does not accept any legal liability for the content of the model output, nor does it assume responsibility for any losses incurred due to the use of associated resources and output results.

About

Code for "Lion: Adversarial Distillation of Proprietary Large Language Models (EMNLP 2023)"

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages