]) starting_model_number = 280000 weights = torch.load( f'{Paths.default.models()}/cardio_part_{str(starting_model_number).zfill(6)}.pt', map_location="cpu" ) enc_dec = StyleGanAutoEncoder().load_state_dict(weights, style=False).cuda() discriminator_img = Discriminator(image_size) discriminator_img.load_state_dict(weights['di']) discriminator_img = discriminator_img.cuda() heatmapper = ToGaussHeatMap(256, 4) hg = HG_heatmap(heatmapper, num_classes=200) # hg.load_state_dict(weights['gh']) hg = hg.cuda() hm_discriminator = Discriminator(image_size, input_nc=1, channel_multiplier=1) hm_discriminator.load_state_dict(weights["dh"]) hm_discriminator = hm_discriminator.cuda() gan_model_tuda = StyleGanModel[HeatmapToImage](enc_dec.generator, StyleGANLoss(discriminator_img), (0.001/4, 0.0015/4)) gan_model_obratno = StyleGanModel[HG_skeleton](hg, StyleGANLoss(hm_discriminator, r1=3), (2e-5, 0.0015/4)) style_opt = optim.Adam(enc_dec.style_encoder.parameters(), lr=1e-5) print(f"board path: {Paths.default.board()}/cardio{int(time.time())}") writer = SummaryWriter(f"{Paths.default.board()}/cardio{int(time.time())}") WR.writer = writer
ToTensor(device), ]) starting_model_number = 280000 weights = torch.load( f'{Paths.default.models()}/cardio_brule_unsup_{str(starting_model_number).zfill(6)}.pt', map_location="cpu") enc_dec = StyleGanAutoEncoder().load_state_dict(weights, style=False).cuda() discriminator_img = Discriminator(image_size) discriminator_img.load_state_dict(weights['di']) discriminator_img = discriminator_img.cuda() heatmapper = ToGaussHeatMap(256, 4) hg = HG_heatmap(heatmapper, num_blocks=1, num_classes=200) hg.load_state_dict(weights['gh']) hg = hg.cuda() cont_opt = optim.Adam(hg.parameters(), lr=2e-5, betas=(0, 0.8)) gan_model_tuda = StyleGanModel[HeatmapToImage](enc_dec.generator, StyleGANLoss(discriminator_img), (0.001 / 4, 0.0015 / 4)) style_opt = optim.Adam(enc_dec.style_encoder.parameters(), lr=1e-5) writer = SummaryWriter( f"{Paths.default.board()}/brule1_cardio_{int(time.time())}") WR.writer = writer batch = next(LazyLoader.cardio().loader_train_inf)
for k in vars(args): print(f"{k}: {vars(args)[k]}") device = torch.device(f"cuda:{args.cuda}" if torch.cuda.is_available() else "cpu") torch.cuda.set_device(device) Cardio.batch_size = batch_size starting_model_number = 264000 weights = torch.load( f'{Paths.default.models()}/cardio_brule_sup_{str(starting_model_number).zfill(6)}.pt', map_location="cpu" ) heatmapper = ToGaussHeatMap(256, 4) hg = HG_heatmap(heatmapper, num_blocks=1, num_classes=200) hg.load_state_dict(weights['gh']) hg = hg.cuda() cont_opt = optim.Adam(hg.parameters(), lr=2e-5, betas=(0, 0.8)) # gan_model_tuda = StyleGanModel[HeatmapToImage](enc_dec.generator, StyleGANLoss(discriminator_img), (0.001/4, 0.0015/4)) # style_opt = optim.Adam(enc_dec.style_encoder.parameters(), lr=1e-5) writer = SummaryWriter(f"{Paths.default.board()}/brule1_cardio_{int(time.time())}") WR.writer = writer batch = next(iter(LazyLoader.cardio().test_loader)) test_img = batch["image"].cuda()
starting_model_number = 90000 + 210000 weights = torch.load( f'{Paths.default.models()}/human_{str(starting_model_number).zfill(6)}.pt', map_location="cpu") enc_dec = StyleGanAutoEncoder( hm_nc=measure_size, image_size=image_size).load_state_dict(weights).cuda() discriminator_img = Discriminator(image_size) discriminator_img.load_state_dict(weights['di']) discriminator_img = discriminator_img.cuda() heatmapper = ToGaussHeatMap(image_size, 2) hg = HG_heatmap(heatmapper, num_classes=measure_size, image_size=image_size, num_blocks=4) hg.load_state_dict(weights['gh']) hg = hg.cuda() hm_discriminator = Discriminator(image_size, input_nc=measure_size, channel_multiplier=1) hm_discriminator.load_state_dict(weights["dh"]) hm_discriminator = hm_discriminator.cuda() gan_model_tuda = StyleGanModel[HeatmapToImage](enc_dec.generator, StyleGANLoss(discriminator_img), (0.001 / 4, 0.0015 / 4)) gan_model_obratno = StyleGanModel[HG_skeleton](hg, StyleGANLoss(hm_discriminator), (2e-5, 0.0015 / 4))
for k in vars(args): print(f"{k}: {vars(args)[k]}") device = torch.device( f"cuda:{args.cuda}" if torch.cuda.is_available() else "cpu") torch.cuda.set_device(device) HumanLoader.batch_size = batch_size starting_model_number = 90000 + 120000 weights = torch.load( f'{Paths.default.models()}/human_{str(starting_model_number).zfill(6)}.pt', map_location="cpu") heatmapper = ToGaussHeatMap(image_size, 2) hg = HG_heatmap(heatmapper, num_classes=32, image_size=image_size, num_blocks=4) hg.load_state_dict(weights['gh']) hg = hg.cuda() requires_grad(hg, False) enc_dec = StyleGanAutoEncoder( hm_nc=measure_size, image_size=image_size).load_state_dict(weights).cuda() discriminator_img = Discriminator(image_size) discriminator_img.load_state_dict(weights['di']) discriminator_img = discriminator_img.cuda() gan_model_tuda = StyleGanModel[HeatmapToImage](enc_dec.generator, StyleGANLoss(discriminator_img), (0.001, 0.0015))