Skip to content

Latest commit

 

History

History
49 lines (36 loc) · 1.69 KB

GETTING_STARTED.md

File metadata and controls

49 lines (36 loc) · 1.69 KB

Getting Started with Compositor

This document provides a brief intro of the usage of Compositor.

Please see Getting Started with Detectron2 for full usage.

Training & Evaluation in Command Line

We provide script train_net.py, that is made to train all the configs provided in Compositor.

Train Compositor w/ MaskFormer

To train a Compositor w/ MaskFormer model with "train_net.py", first setup the corresponding datasets following DATASETS.md, and change directory to Compositor_Mask2Former, then run:

python train_net.py --num-gpus 8 \
  --config-file configs/partimagenet/semantic-segmentation/compositor_R50_bs16_90k.yaml

The configs are made for 8-GPU training. Since we use ADAMW optimizer, it is not clear how to scale learning rate with batch size. To train on 1 GPU, you need to figure out learning rate and batch size by yourself:

python train_net.py \
  --config-file configs/partimagenet/semantic-segmentation/compositor_R50_bs16_90k.yaml \
  --num-gpus 1 SOLVER.IMS_PER_BATCH SET_TO_SOME_REASONABLE_VALUE SOLVER.BASE_LR SET_TO_SOME_REASONABLE_VALUE

Train Compositor w/ kMaX-DeepLab

To train a Compositor w/ kMaX-DeepLab model with "train_net.py", change directory to Compositor_kMaX-DeepLab, then run

python train_net.py --num-gpus 8 \
  --config-file configs/partimagenet/semantic_segmentation/compositor_r50.yaml \

Evaluate Compositor w/ Mask2Former or kMaX-DeepLab

To evaluate a model's performance, use

python train_net.py \
  --config-file /path/to/config \
  --eval-only MODEL.WEIGHTS /path/to/checkpoint_file \

For more options, see python train_net.py -h.