Skip to content

Davido111200/Prompting_With_Episodic_Memory

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

POEM

Installation

# Install Instructions
conda create -n poem python=3.8
conda activate poem

Main parameters

Parameters can be found in src/arguments.py

write_type: write method in memory
memory_k: number of neighbors used at inference time

Other parameters should stay the same for reproduction. It's RL at the end :D

Run POEM on benchmarks

torchrun --nproc_per_node 1 main.py --normalize_obs --use_attention --max_num_edits 1 --reward_type prob --ac_representation mean --example_sample_type narrowed_example_selection --num-env-steps 5000 --num_actors 1 --num_processes 16 --num-mini-batch 16 --dataset math_qa --write_type max --memory_k 4 --type_embd llama --eval_data test --seed 3 --baseline memory --eval-interval 10 --use_verbalizer

About

[ECAI 2024] Official code for European Conference on Artificial Intelligence 2024 paper - Large Language Models Prompting with Episodic Memory

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages