def cfg(data_config): model_config = { # Load pretrained model for testing "model_name": "DenseDepth", "model_params": { "existing": os.path.join("models", "nyu.h5"), }, "model_state_dict_fn": None } ckpt_file = None # Keep as None save_outputs = True seed = 95290421 # changing seed does not impact evaluation small_run = 0 dataset_type = "test" entry = None # print(data_config.keys()) output_dir = os.path.join( "results", data_config["data_name"], # e.g. nyu_depth_v2 "{}_{}".format(dataset_type, small_run), model_config["model_name"]) # e.g. DORN_nyu_nohints safe_makedir(output_dir) ex.observers.append( FileStorageObserver.create(os.path.join(output_dir, "runs"))) cuda_device = "0" # The gpu index to run on. Should be a string os.environ["CUDA_VISIBLE_DEVICES"] = cuda_device # print("after: {}".format(os.environ["CUDA_VISIBLE_DEVICES"])) if ckpt_file is not None: model_update, _, _ = load_checkpoint(ckpt_file) model_config.update(model_update) del model_update, _ # So sacred doesn't collect them.
def cfg(data_config, spad_config): model_config = { # Load pretrained model for testing "model_name": "SinkhornOpt", "model_params": { "sgd_iters": 300, "sinkhorn_iters": 40, "sigma": 0.5, "lam": 2e1, "kde_eps": 1e-4, "sinkhorn_eps": 1e-7, "dc_eps": 1e-5, "remove_dc": spad_config["dc_count"] > 0., "use_intensity": spad_config["use_intensity"], "use_squared_falloff": spad_config["use_squared_falloff"], "lr": 1e5, "sid_bins": data_config["sid_bins"], "offset": data_config["offset"], "min_depth": data_config["min_depth"], "max_depth": data_config["max_depth"], "alpha": data_config["alpha"], "beta": data_config["beta"], }, "model_state_dict_fn": None # Keep as None } ckpt_file = None # Keep as None save_outputs = True seed = 95290421 small_run = 0 entry = None pdict = model_config["model_params"] comment = "_".join(["sgd_iters_{}".format(pdict["sgd_iters"]), "sinkhorn_iters_{}".format(pdict["sinkhorn_iters"]), "sigma_{}".format(pdict["sigma"]), "lam_{}".format(pdict["lam"]), "kde_eps_{}".format(pdict["kde_eps"]), "sinkhorn_eps_{}".format(pdict["sinkhorn_eps"]), ]) del pdict # print(data_config.keys()) fullcomment = comment + "_" + spad_config["spad_comment"] output_dir = os.path.join("results", data_config["data_name"], # e.g. nyu_depth_v2 "test_{}".format(small_run), model_config["model_name"]) # e.g. DORN_nyu_nohints if fullcomment is not "": output_dir = os.path.join(output_dir, fullcomment) safe_makedir(output_dir) ex.observers.append(FileStorageObserver.create(os.path.join(output_dir, "runs"))) # Devices are for pytorch. cuda_device = "0" # The gpu index to run on. Should be a string os.environ["CUDA_VISIBLE_DEVICES"] = tf_cuda_device device = torch.device("cuda" if torch.cuda.is_available() else "cpu") print("using device: {} (CUDA_VISIBLE_DEVICES = {})".format(device, os.environ["CUDA_VISIBLE_DEVICES"])) if ckpt_file is not None: model_update, _, _ = load_checkpoint(ckpt_file) model_config.update(model_update) del model_update, _ # So sacred doesn't collect them.
def cfg(data_config): model_config = { # Load pretrained model for testing "model_name": "DORN_nyu_hints_Unet", "model_params": { "hints_len": 68, "spad_weight": 1., "in_channels": 3, "in_height": 257, "in_width": 353, "sid_bins": data_config["sid_bins"], "offset": data_config["offset"], "min_depth": data_config["min_depth"], "max_depth": data_config["max_depth"], "alpha": data_config["alpha"], "beta": data_config["beta"], "frozen": True, "pretrained": True, "state_dict_file": os.path.join("models", "torch_params_nyuv2_BGR.pth.tar"), }, "model_state_dict_fn": None } ckpt_file = "checkpoints/Mar15/04-10-54_DORN_nyu_hints_nyu_depth_v2/checkpoint_epoch_9_name_fixed.pth.tar" # ckpt_file = None # Bayesian hints eval dataset_type = "val" save_outputs = True output_dir = os.path.join( "results", data_config["data_name"], # e.g. nyu_depth_v2 model_config["model_name"], # e.g. DORN_nyu_nohints dataset_type) seed = 95290421 small_run = False cuda_device = "0" # The gpu index to run on. Should be a string os.environ["CUDA_VISIBLE_DEVICES"] = cuda_device # print("after: {}".format(os.environ["CUDA_VISIBLE_DEVICES"])) device = torch.device("cuda" if torch.cuda.is_available() else "cpu") print("using device: {} (CUDA_VISIBLE_DEVICES = {})".format( device, os.environ["CUDA_VISIBLE_DEVICES"])) if ckpt_file is not None: model_update, _, _ = load_checkpoint(ckpt_file) model_config.update(model_update) del model_update, _ # So sacred doesn't collect them.
def cfg(data_config): model_config = { # Load pretrained model for testing "model_name": "DenseDepthHistogramMatchingWasserstein", "model_params": { "sgd_iters": 100, "sinkhorn_iters": 40, "sigma": 0.5, "lam": 1e1, "kde_eps": 1e-4, "sinkhorn_eps": 1e-7, "dc_eps": 1e-5, "lr": 1e5, "min_depth": 0., "max_depth": 10., "sid_bins": 68, "offset": 0., "alpha": 0.6569154266167957, "beta": 9.972175646365525, "existing": os.path.join("models", "nyu.h5"), }, "model_state_dict_fn": None } ckpt_file = None # Keep as None save_outputs = True seed = 95290421 # changing seed does not impact evaluation small_run = 0 dataset_type = "test" entry = None # print(data_config.keys()) output_dir = os.path.join( "results", data_config["data_name"], # e.g. nyu_depth_v2 "{}_{}".format(dataset_type, small_run), model_config["model_name"]) # e.g. DORN_nyu_nohints safe_makedir(output_dir) ex.observers.append( FileStorageObserver.create(os.path.join(output_dir, "runs"))) cuda_device = "0,1" # The visible gpus. First one is the tensorflow, second one is pytorch. os.environ["CUDA_VISIBLE_DEVICES"] = cuda_device # print("after: {}".format(os.environ["CUDA_VISIBLE_DEVICES"])) device = torch.device("cuda:1" if torch.cuda.is_available() else "cpu") # print("after: {}".format(os.environ["CUDA_VISIBLE_DEVICES"])) if ckpt_file is not None: model_update, _, _ = load_checkpoint(ckpt_file) model_config.update(model_update) del model_update, _ # So sacred doesn't collect them.
def cfg(data_config): model_config = { # Load pretrained model for testing "model_name": "DORN_nyu_nohints", "model_params": { "in_channels": 3, "in_height": 257, "in_width": 353, "frozen": True, "pretrained": True, "state_dict_file": os.path.join("models", "torch_params_nyuv2_BGR.pth.tar"), }, "model_state_dict_fn": None } ckpt_file = None # Keep as None save_outputs = True seed = 95290421 small_run = 0 entry = None # hyperparams = ["sgd_iters", "sinkhorn_iters", "sigma", "lam", "kde_eps", "sinkhorn_eps"] pdict = model_config["model_params"] del pdict # print(data_config.keys()) output_dir = os.path.join( "results", data_config["data_name"], # e.g. nyu_depth_v2 "{}_{}".format("test", small_run), model_config["model_name"]) # e.g. DORN_nyu_nohints safe_makedir(output_dir) ex.observers.append( FileStorageObserver.create(os.path.join(output_dir, "runs"))) cuda_device = "0" # The gpu index to run on. Should be a string os.environ["CUDA_VISIBLE_DEVICES"] = cuda_device # print("after: {}".format(os.environ["CUDA_VISIBLE_DEVICES"])) device = torch.device("cuda" if torch.cuda.is_available() else "cpu") print("using device: {} (CUDA_VISIBLE_DEVICES = {})".format( device, os.environ["CUDA_VISIBLE_DEVICES"])) if ckpt_file is not None: model_update, _, _ = load_checkpoint(ckpt_file) model_config.update(model_update) del model_update, _ # So sacred doesn't collect them.
def cfg(data_config): model_config = { # Load pretrained model for testing "model_name": "DORN_median_matching", "model_params": { "in_channels": 3, "in_height": 257, "in_width": 353, "sid_bins": 68, "offset": 0., "min_depth": 0., "max_depth": 10., "alpha": 0.6569154266167957, "beta": 9.972175646365525, "frozen": True, "pretrained": True, "state_dict_file": os.path.join("models", "torch_params_nyuv2_BGR.pth.tar"), }, "model_state_dict_fn": None } ckpt_file = None # Keep as None dataset_type = "test" save_outputs = True seed = 95290421 small_run = 0 # print(data_config.keys()) output_dir = os.path.join("results", data_config["data_name"], # e.g. nyu_depth_v2 "{}_{}".format(dataset_type, small_run), model_config["model_name"]) # e.g. DORN_nyu_nohints safe_makedir(output_dir) ex.observers.append(FileStorageObserver.create(os.path.join(output_dir, "runs"))) cuda_device = "0" # The gpu index to run on. Should be a string os.environ["CUDA_VISIBLE_DEVICES"] = cuda_device # print("after: {}".format(os.environ["CUDA_VISIBLE_DEVICES"])) device = torch.device("cuda" if torch.cuda.is_available() else "cpu") print("using device: {} (CUDA_VISIBLE_DEVICES = {})".format(device, os.environ["CUDA_VISIBLE_DEVICES"])) if ckpt_file is not None: model_update, _, _ = load_checkpoint(ckpt_file) model_config.update(model_update) del model_update, _ # So sacred doesn't collect them.
def cfg(data_config): model_config = {} # To be loaded from the checkpoint file. ckpt_file = "checkpoints/Mar07/02-18-30_DenoisingUnetModel_cifar10/checkpoint_epoch_0.pth.tar" eval_config = { "dataset": "val", # {val, test} "mode": "save_outputs", # {save_outputs, evaluate_metrics} "output_dir": "cifar10_eval" } seed = 95290421 cuda_device = "0" # The gpu index to run on. Should be a string os.environ["CUDA_VISIBLE_DEVICES"] = cuda_device # print("after: {}".format(os.environ["CUDA_VISIBLE_DEVICES"])) device = torch.device("cuda" if torch.cuda.is_available() else "cpu") print("using device: {} (CUDA_VISIBLE_DEVICES = {})".format( device, os.environ["CUDA_VISIBLE_DEVICES"])) if ckpt_file is not None: model_update, _, _ = load_checkpoint(ckpt_file) model_config.update(model_update) del model_update, _ # So sacred doesn't collect them.
def cfg(data_config): model_config = { # Load pretrained model for testing "model_name": "DORN_bayesian_opt", "model_params": { "sgd_iters": 20, "lr": 1e-3, "hints_len": 68, "spad_weight": 1., "in_channels": 3, "in_height": 257, "in_width": 353, "sid_bins": data_config["sid_bins"], "offset": data_config["offset"], "min_depth": data_config["min_depth"], "max_depth": data_config["max_depth"], "alpha": data_config["alpha"], "beta": data_config["beta"], "frozen": True, "pretrained": True, "state_dict_file": os.path.join("models", "torch_params_nyuv2_BGR.pth.tar"), }, "model_state_dict_fn": None # Keep as None } ckpt_file = None # Keep as None dataset_type = "val" eval_config = { "save_outputs": True, "evaluate_metrics": True, "output_dir": os.path.join("data", "results", model_config["model_name"], dataset_type), "entry": None # If we want to evaluate on a single entry } seed = 95290421 small_run = False cuda_device = "0" # The gpu index to run on. Should be a string os.environ["CUDA_VISIBLE_DEVICES"] = cuda_device # print("after: {}".format(os.environ["CUDA_VISIBLE_DEVICES"])) device = torch.device("cuda" if torch.cuda.is_available() else "cpu") print("using device: {} (CUDA_VISIBLE_DEVICES = {})".format( device, os.environ["CUDA_VISIBLE_DEVICES"])) if ckpt_file is not None: model_update, _, _ = load_checkpoint(ckpt_file) model_config.update(model_update) del model_update, _ # So sacred doesn't collect them.