(alg, args, inner=False, alias=True, isteacher=False)
| 3 | |
| 4 | |
| 5 | def get_params(alg, args, inner=False, alias=True, isteacher=False): |
| 6 | if args.schuse: |
| 7 | if args.schusech == 'cos': |
| 8 | initlr = args.lr |
| 9 | else: |
| 10 | initlr = 1.0 |
| 11 | else: |
| 12 | if inner: |
| 13 | initlr = args.inner_lr |
| 14 | else: |
| 15 | initlr = args.lr |
| 16 | if isteacher: |
| 17 | params = [ |
| 18 | {'params': alg[0].parameters(), 'lr': args.lr_decay1 * initlr}, |
| 19 | {'params': alg[1].parameters(), 'lr': args.lr_decay2 * initlr}, |
| 20 | {'params': alg[2].parameters(), 'lr': args.lr_decay2 * initlr} |
| 21 | ] |
| 22 | return params |
| 23 | if inner: |
| 24 | params = [ |
| 25 | {'params': alg[0].parameters(), 'lr': args.lr_decay1 * |
| 26 | initlr}, |
| 27 | {'params': alg[1].parameters(), 'lr': args.lr_decay2 * |
| 28 | initlr} |
| 29 | ] |
| 30 | elif alias: |
| 31 | params = [ |
| 32 | {'params': alg.featurizer.parameters(), 'lr': args.lr_decay1 * initlr}, |
| 33 | {'params': alg.classifier.parameters(), 'lr': args.lr_decay2 * initlr} |
| 34 | ] |
| 35 | else: |
| 36 | params = [ |
| 37 | {'params': alg[0].parameters(), 'lr': args.lr_decay1 * initlr}, |
| 38 | {'params': alg[1].parameters(), 'lr': args.lr_decay2 * initlr} |
| 39 | ] |
| 40 | if ('DANN' in args.algorithm) or ('CDANN' in args.algorithm): |
| 41 | params.append({'params': alg.discriminator.parameters(), |
| 42 | 'lr': args.lr_decay2 * initlr}) |
| 43 | if ('CDANN' in args.algorithm): |
| 44 | params.append({'params': alg.class_embeddings.parameters(), |
| 45 | 'lr': args.lr_decay2 * initlr}) |
| 46 | return params |
| 47 | |
| 48 | |
| 49 | def get_optimizer(alg, args, inner=False, alias=True, isteacher=False): |
no test coverage detected