jinzhuoran / rwku Goto Github PK
View Code? Open in Web Editor NEWRWKU: Benchmarking Real-World Knowledge Unlearning for Large Language Models
Home Page: https://rwku-bench.github.io
RWKU: Benchmarking Real-World Knowledge Unlearning for Large Language Models
Home Page: https://rwku-bench.github.io
impressive work!
Hi,
Thanks for sharing the impressive code!
The computation cost of this repo is higher than expected. As LLaMA-Factory suggested, a 7B model would only require 60GB GPU memory for full fine-tune. However, it requires about 160 GB when I run full/run_ga.sh.
Which step increases the cost?
Hi, thanks for the great work.
I would like to know how the values of "MIA Set" and "Flu" in the table are reported with the obtained results.
#3 (comment)
For example, "Loss -2.492", "Loss -2.370" and "Entropy 6.978" in the above file.
Hello. I am trying to replicate the baseline results for Target unlearning from the paper, however I have been getting consistently worse results for both LLama3-8B-instruct and Phi-3 Mini-4K-Instruct. Here is a comparison:
Method | Model | FB | QA | AA | Avg | FB | QA | Avg | Forget loss ↓ | Retain loss ↑ | Gen (MMLU) | Real (BBH) | Tru | Fac | Fluency (entropy) |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
My Baseline | LLama3-8B-Instruct | 76.70 | 70.67 | 72.27 | 73.21 | 85.21 | 80.96 | 83.09 | -2.25 | -2.29 | 64.25 | 40.24 | 36.22 | 52.27 | 7.08 |
Paper | llama3-8b-instruct (Before) | 85.90 | 76.40 | 77.70 | 79.60 | 95.60 | 85.30 | 90.70 | 226.70 | 230.40 | 65.70 | 42.30 | 36.80 | 53.50 | 7.05 |
My Baseline | Phi3 mini 4k instruct | 46.19 | 44.94 | 54.01 | 48.38 | 52.79 | 56.11 | 54.45 | -1.93 | -1.94 | 67.91 | 42.57 | 37.66 | 39.77 | 6.69 |
Paper | Phi3 mini 4k instruct (Baseline) | 47.10 | 47.40 | 55.80 | 51.80 | 56.20 | 61.40 | 58.30 | 205.60 | 207.50 | 64.40 | 39.50 | 46.40 | 15.10 | 7.07 |
I am running this command to get the baseline on the first 100 celebrities (1-100):
WANDB_DISABLED=true python src/train_bash.py --stage ga
--model_name_or_path mms://core-ai-nlp/Meta-Llama-3-8B-Instruct/1
--dataset ${id}Positive --dataset_dir RWKU-dataset
--output_dir ./saves/unlearn_bench/People/${id}/${method_name}/${model_name} --overwrite_cache
--overwrite_output_dir --cutoff_len 512 --preprocessing_num_workers 16
--per_device_train_batch_size 8 --per_device_eval_batch_size 8 --gradient_accumulation_steps 1
--lr_scheduler_type cosine --logging_steps 10 --warmup_steps 20 --save_steps 30000
--eval_steps 30000 --evaluation_strategy steps --load_best_model_at_end --template llama3
--learning_rate 6e-8 --num_train_epochs 1.0 --val_size 0.0000001 --plot_loss
--output_result_dir ./results/unlearn_bench/People/${id}/${method_name}/${model_name}
--fp16 --eval_dataset_dir RWKU-dataset/RWKU/Target/
--target ${id} 2>&1 | tee ./logs/unlearn_bench/People/${method_name}/${model_name}${id}.log
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.