config.cols_to_use = [0,1,2,3,4] # may want to change: feat_size, layer_dims, etc. config.model_kwargs = {"in_channels": len(config.cols_to_use), "num_classes": 3} config.data_path = "/fast_scratch/WatChMaL/data/pointnet/pointnet_trainval.h5" config.indices_file = "/fast_scratch/WatChMaL/data/pointnet/pointnet_trainval_idxs.npz" #make sure to change this config.dump_path = "/home/dgreen/training_outputs/pointnet/time/adam/" config.num_data_workers = 0 # Sometime crashes if we do multiprocessing config.device = 'cuda:7' config.optimizer = "Adam" config.optimizer_kwargs = {"lr":1e-3, "betas": (0.9, 0.999)} config.use_scheduler = False config.scheduler_kwargs = {"mode":"min", "min_lr":1e-6, "patience":1, "verbose":True} config.scheduler_step = 190 config.batch_size = 32 config.epochs = 20 config.report_interval = 200 config.num_val_batches = 512 config.valid_interval = 1000 config.validate_batch_size = 32 config.validate_dump_interval = 256
config = EasyDict() config.cols_to_use = [0,1,2,3] # may want to change: feat_size, layer_dims, etc. config.model_kwargs = {"num_features":1, "num_classes": 3} config.data_path = "/fast_scratch/WatChMaL/data/pointnet/pointnet_trainval.h5" config.indices_file = "/fast_scratch/WatChMaL/data/pointnet/pointnet_trainval_idxs.npz" #make sure to change this config.dump_path = "/home/dgreen/training_outputs/pointnet2/no_time/adam/" config.num_data_workers = 0 # Sometime crashes if we do multiprocessing config.device = 'cuda:6' config.optimizer = "Adam" config.optimizer_kwargs = {"lr":1e-3, "betas": (0.9, 0.999)} config.use_scheduler = False config.scheduler_kwargs = {"mode":"min", "min_lr":1e-6, "patience":1, "verbose":True} config.scheduler_step = 190 config.batch_size = 32 config.epochs = 20 config.report_interval = 200 config.num_val_batches = 256 config.valid_interval = 1000 config.validate_batch_size = 32 config.validate_dump_interval = 256
config.num_data_workers = 0 # Sometime crashes if we do multiprocessing config.device = 'gpu' config.gpu_list = [7] config.optimizer = "SGD" config.optimizer_kwargs = { "lr": 0.01, "weight_decay": 1e-3, "momentum": 0.9, "nesterov": True } config.scheduler_kwargs = { "mode": "min", "min_lr": 1e-6, "patience": 1, "verbose": True } config.scheduler_step = 190 config.batch_size = 32 config.epochs = 6 config.report_interval = 50 config.num_val_batches = 128 config.valid_interval = 200 config.validate_batch_size = 32 config.validate_dump_interval = 256