MC-SMoE
[ICLR‘24 Spotlight] Code for the paper "Merge, Then Compress: Demystify Efficient SMoE with Hints from Its Routing Policy"
File Explorer
Download Latest Version (.zip)- __init__.py
- calculate_pipline.py
- flops_counter.py
- pytorch_ops.py
- utils.py
- __init__.py
- data_collator.py
- evaluation.py
- preprocessors.py
- __init__.py
- lm_eval.py
- minipile.py
- __init__.py
- fisher.py
- grouping.py
- grouping_fsgpt.py
- grouping_mixtral.py
- permutation.py
- utils.py
- __init__.py
- fsgpt.py
- fsgpt_moe.py
- modeling_outputs.py
- sparse.py
- switch_transformers.py
- __init__.py
- losparse.py
- task_specific.py
- task_specific_fsgpt.py
- __init__.py
- constants.py
- dummy_test.py
- sparsity.py
- training_utils.py
- __init__.py
- evaluate-expert-usage.py
- evaluate-fsgpt-expert-usage.py
- evaluate-fsgpt-zero-shot.py
- finetune-switch-transformers.py
- fsgpt-zero-shot-losparse.py
- fsgpt-zero-shot-with-task-specific-pruning.py
- losparse-downstream.py
- merge-and-kd-for-recover-downstream.py
- merge-by-averaging.py
- merge-fsgpt-by-usage-frequency-weighted.py
- msmoe-merging-mixtral.py
- msmoe-merging.py
- permute-fsgpt-model.py
- permute-model.py
- profile-flops.py
- profile-fsgpt-flops.py
- profile-latency.py
- prune-fsgpt-non-core-experts.py
- prune-non-core-experts-and-kd-downstream.py
- random-merge-fsgpt.py
- random-merge.py
- switch-finetune-with-task-specific-pruning.py
- t5-finetune-on-downstream.py
- evaluate-dense-zero-shot.sh
- evaluate-merged-moe-zero-shot.sh
- evaluate-moe-zero-shot.sh
- merge-by-usage-frequency-weighted.sh
- permute-moe.sh
- post-merging-losparse.sh
- profile-flops.sh
- prune-non-core.sh
- random-merge.sh
- task-specific.sh
- finetune-switch.sh
- finetune-t5.sh
- losparse-only.sh
- merge-by-averaging.sh
- merge-by-usage-frequency-weighted.sh
- permute-switch.sh
- post-merging-losparse.sh
- profile-flops.sh
- profile-latency.sh
- prune-non-core.sh
- random-merge.sh
- task-specific.sh
- evaluation_config.yaml
- finetune_config.yaml
- pipeline.jpg
- .gitignore
- LICENSE
- README.md
- requirements.txt
# Installation Guide
1. Get the code
git clone https://github.com/UNITES-Lab/MC-SMoE
Downloads the entire project code from GitHub to your computer.
cd MC-SMoE
Moves into the project folder you just downloaded.
2. Python
Easy RecommendedPrerequisites
pip install -r requirements.txt
Installs the Python libraries listed in requirements.txt (or similar).
If it runs without errors and prints output in the terminal, it worked.
Pulled directly from this repo's README.
// repository documentation
Was this content helpful?
(0 ratings)
