https://github.com/wellecks/multiset

Loss Functions for Multiset Prediction
https://github.com/wellecks/multiset

Last synced: 17 days ago
JSON representation

Loss Functions for Multiset Prediction

• Host: GitHub
• URL: https://github.com/wellecks/multiset
• Owner: wellecks
• Created: 2017-11-24T14:34:38.000Z (about 7 years ago)
• Default Branch: master
• Last Pushed: 2017-12-10T02:58:08.000Z (about 7 years ago)
• Last Synced: 2024-10-29T20:13:00.749Z (2 months ago)
• Language: Python
• Size: 12.7 KB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# multiset

Loss Functions for Multiset Prediction

### Running

1. Generate an MNIST-multi dataset

2. Run `train.py` with suitable cmd line arguments e.g:

```bash

python train.py --dataset-path data/mnist_multi_70000_min20_max50_4 --mnist-multi \

                --max-objects 4 --loss multiset_loss --use-cuda

```

Run `train.py -h` for cmd line argument details. 

Note that `--dataset-path` and `--max-objects` vary based on the MNIST Multi dataset used.

Choose the loss with `--loss`.

When using the sequential loss (`--loss ce_loss`), choose an ordering strategy with `--label-order`.