cwm/eval/Segmentation/archive/linear_probing.py

import os
import argparse
import sys
import torch
import warnings
warnings.filterwarnings("ignore")
torch.multiprocessing.set_sharing_strategy('file_system')
# Set environment variables
# os.environ['CUDA_VISIBLE_DEVICES'] = '2,3,4,5,6'
os.environ['OMP_NUM_THREADS'] = '1'
os.environ['DETECTRON2_DATASETS'] = '/ccn2/u/honglinc/datasets'

# Add necessary path
MASK2FORMER_PATH = '/ccn2/u/honglinc/Mask2Former'
BBNET_PATH = '/home/honglinc/BBNet'
sys.path.append(os.path.join(BBNET_PATH, 'bbnet/models/VideoMAE-main/'))
sys.path.append(BBNET_PATH)
sys.path.append(MASK2FORMER_PATH)

# BBNet import
import modeling_pretrain as vmae_tranformers
from evaluate_segmentation_readout_helper_v2 import CWMSegmentPredictorV2

import detectron2.utils.comm as comm
from detectron2.evaluation import verify_results
from train_net import setup, Trainer, DetectionCheckpointer
from detectron2.engine import default_argument_parser, launch

def main(args):
    cfg = setup(args)

    if args.eval_only:
        model = Trainer.build_model(cfg)
        DetectionCheckpointer(model, save_dir=cfg.OUTPUT_DIR).resume_or_load(
            cfg.MODEL.WEIGHTS, resume=args.resume
        )
        res = Trainer.test(cfg, model)
        if cfg.TEST.AUG.ENABLED:
            res.update(Trainer.test_with_TTA(cfg, model))
        if comm.is_main_process():
            verify_results(cfg, res)
        return res

    trainer = Trainer(cfg)
    trainer.resume_or_load(resume=args.resume)
    return trainer.train()


if __name__ == "__main__":
    args = default_argument_parser().parse_args()
    print("Command Line Args:", args)
    launch(
        main,
        args.num_gpus,
        num_machines=args.num_machines,
        machine_rank=args.machine_rank,
        dist_url=args.dist_url,
        args=(args,),
    )