# Dataset Configuration Dataset: num_days: 73 # Test with sequence of 1 day - should be the same as in Test - batch_size: 64 num_workers: 7 valid_ratio: 0.2 max_num_samples: None #1000 _DEFAULT_TRAIN_FILEPATH: "/mounts/Datasets3/2022-ChallengePlankton/sub_2CMEMS-MEDSEA-2010-2016-training.nc.bin" _DEFAULT_TEST_FILEPATH: "/mounts/Datasets3/2022-ChallengePlankton/sub_2CMEMS-MEDSEA-2017-testing.nc.bin" _ENCODING_LINEAR: "I" _ENCODING_INDEX: "I" # h(short) with 2 bytes should be sufficient _ENCODING_OFFSET_FORMAT: "" _ENCODING_ENDIAN: "<" # Data Transformation ApproximativeStats: False ApproximativeMean: "torch.tensor([ 4.2457e+01, 7.4651e+00, 1.6738e+02, 1.3576e+09, 2.3628e+00, 4.6839e+01, 2.3855e-01, 3.6535e+00, 1.9776e+00, 2.2628e+02, 8.1003e+00, 1.8691e-01, 3.8384e+01, 2.6626e+00, 1.4315e+01, -4.1419e-03, 6.0274e-03, -5.1017e-01])" ApproximativeSTD: "torch.tensor([5.8939e-01, 8.1625e-01, 1.4535e+02, 5.4952e+07, 1.7543e-02, 1.3846e+02, 2.1302e-01, 1.9558e+00, 4.1455e+00, 1.2408e+01, 2.2938e-02, 9.9070e-02, 1.9490e-01, 9.2847e-03, 2.2575e+00, 8.5310e-02, 7.8280e-02, 8.6237e-02])" ApproximativeMaxi: "torch.tensor([ 4.3479e+01, 9.0000e+00, 4.9267e+02, 1.4528e+09, 2.4088e+00, 2.7824e+03, 1.5576e+00, 6.2457e+00, 2.5120e+02, 2.7188e+02, 8.1683e+00, 3.2447e-01, 3.9041e+01, 2.7162e+00, 2.9419e+01, 8.6284e-01, 7.6471e-01, -7.7745e-02])" ApproximativeMini: "torch.tensor([ 4.1479e+01, 6.0000e+00, 1.0182e+00, 1.2623e+09, 2.2433e+00, 1.0910e+01, 1.0000e-11, 1.0000e-11, -1.1467e+01, 1.9718e+02, 7.9218e+00, 1.0000e-11, 3.7171e+01, 2.5584e+00, 1.2075e+01, -1.2436e+00, -9.9256e-01, -8.8131e-01])" #Optimizer selection Optimizer: Adam # in {Adam} #Training parameters Training: Epochs: 20 #Model selection Model: Name: RNN #Model parameters selection LinearRegression: # Bias in {True, False} Bias: True HiddenSize: 35 Initialization: init_he BidirectionalLSTM: HiddenSize: 70 NumLayers: 4 Initialization: None RNN: HiddenSize: 35 NumLayers: 4 Initialization: None #Name of directory containing logs LogDir: ./logs/ #Visualization Wandb: log_freq: 100 #log gradients and parameters every log_freq batches log_interval: 10 # log the train_loss every log_interval batches