MCPcopy Create free account
hub / github.com/Zhiyuan-R/Tiger-Diffusion / main

Function main

test_generation.py:485–530  ·  view source on GitHub ↗
(opt)

Source from the content-addressed store, hash-verified

483
484
485def main(opt):
486
487 if opt.category == 'airplane':
488 opt.beta_start = 1e-5
489 opt.beta_end = 0.008
490 opt.schedule_type = 'warm0.1'
491
492 exp_id = os.path.splitext(os.path.basename(__file__))[0]
493 dir_id = os.path.dirname(__file__)
494 output_dir = get_output_dir(dir_id, exp_id)
495 copy_source(__file__, output_dir)
496 logger = setup_logging(output_dir)
497
498 outf_syn, = setup_output_subdirs(output_dir, 'syn')
499
500 betas = get_betas(opt.schedule_type, opt.beta_start, opt.beta_end, opt.time_num)
501 model = Model(opt, betas, opt.loss_type, opt.model_mean_type, opt.model_var_type)
502
503 if opt.cuda:
504 model.cuda()
505
506 def _transform_(m):
507 return nn.parallel.DataParallel(m)
508
509 model = model.cuda()
510 model.multi_gpu_wrapper(_transform_)
511
512 model.eval()
513
514 with torch.no_grad():
515
516 logger.info("Resume Path:%s" % opt.model)
517
518 resumed_param = torch.load(opt.model)
519 model.load_state_dict(resumed_param['model_state'])
520
521
522 ref = None
523 if opt.generate:
524 opt.eval_path = os.path.join(outf_syn, 'samples.pth')
525 Path(opt.eval_path).parent.mkdir(parents=True, exist_ok=True)
526 ref=generate(model, opt)
527
528 if opt.eval_gen:
529 # Evaluate generation
530 evaluate_gen(opt, ref, logger)
531
532
533def parse_args():

Callers 1

test_generation.pyFile · 0.70

Calls 10

multi_gpu_wrapperMethod · 0.95
evalMethod · 0.95
get_output_dirFunction · 0.85
copy_sourceFunction · 0.85
setup_loggingFunction · 0.85
setup_output_subdirsFunction · 0.85
generateFunction · 0.85
evaluate_genFunction · 0.85
get_betasFunction · 0.70
ModelClass · 0.70

Tested by

no test coverage detected