corrected minor errors

README.md

@@ -9,7 +9,7 @@ tags:
 - vegetation
 - amazon
 model-index:
-- name: MultiTask ConvLSTM
+- name: MultiTask ConvLSTM w/veg inputs
   results:
   - task:
       type: time-series-forecasting
@@ -19,25 +19,25 @@ model-index:
       type: reanalysis
     metrics:
     - type: mean_squared_error
-      value: 0.448
+      value: 0.28
     - type: spearman_correlation
-      value: 0.8412
+      value: 0.87
     - type: pearson_correlation
-      value: 0.8302
+      value: 0.79
     - type: kendall_tau
-      value: 0.6870
+      value: 0.70
     - type: nash_sutcliffe_efficiency
-      value: 0.6890
+      value: 0.62
     - type: f1
-      value: 0.9260
+      value: 0.82
     - type: accuracy 
-      value: 0.9159
+      value: 0.90
     - type: precision
-      value: 0.9186
+      value: 0.90
     - type: ROC-AUC
-      value: 0.9754
+      value: 0.97
     - type: recall
-      value: 0.9335
+      value: 0.75
 ---
 
 # MultiTask ConvLSTM for Precipitation Prediction

no_veg/data/normalized_test_data2_no_veg_input.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5b76c968c3a80b260f8db30d5e9c219c241ac26fd44856c3c87008394dac8e9
+size 1644632048

no_veg/data/normalized_test_data_no_veg_input.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e5b76c968c3a80b260f8db30d5e9c219c241ac26fd44856c3c87008394dac8e9
-size 1644632048
+oid sha256:b190904dceead14eb69a6e0cc7d25456a9fb99fb352821a4bb9a1a2189bcb0db
+size 1644634018

no_veg/example_inference.py

@@ -3,7 +3,7 @@ import torch
 from MultiTaskConvLSTM import ConvLSTMNetwork
 from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, roc_auc_score
 import torch
-import toch.nn as nn
+import torch.nn as nn
 from tqdm.auto import tqdm
 from utils import (
     mse, mae, nash_sutcliffe_efficiency, r2_score, pearson_correlation,
@@ -24,7 +24,7 @@ set_forecast_horizon = 1
 #Define variables for evaluation
 batch_size = 16
 time_steps_out = set_forecast_horizon
-channels = 9
+channels = 8
 
 #Variable names
 #Variable names
@@ -32,8 +32,8 @@ variable_names = ['10 metre U wind component', '10 metre V wind component', '2 m
 
 # Adjust input_dim and output_channels according to your data specifics
 model = ConvLSTMNetwork(
-    input_dim=9 * set_lookback, 
-    hidden_dims=[9, 32, 64], 
+    input_dim=8 * set_lookback, 
+    hidden_dims=[8, 32, 64], 
     kernel_size=(3,3), 
     num_layers=3, 
     output_channels=64 * set_forecast_horizon, 
@@ -46,7 +46,7 @@ bce_loss_fn = nn.BCELoss()       # For classification output
 
 optimizer = optim.AdamW(model.parameters(), lr = 0.005) 
 
-checkpoint = torch.load("MultiTaskConvLSTM_no_veg_variables")
+checkpoint = torch.load("MultiTaskConvLSTM_no_veg_variables.pth", map_location = device)
 model.load_state_dict(checkpoint['model_state_dict'])
 
 # If you want to move the model to the GPU (optional, depending on your setup)

veg/MultiTaskConvLSTM_veg_variables.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b7e8b5e8b33db227257dd794dd6ba5ff10d4759aeed1f758550d4f3acb69cc26
-size 1383333
+oid sha256:5e14b94f02be8e5ea077091c165cedc5b140fe935b5d30f33aaa2722883b3618
+size 1383508

veg/data/normalized_test_data2_veg_input.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:075c60cf83a2d7ef68720b72c77f41a034d6418b67085e016fff6a5bdfef878c
+size 2631223074

veg/data/normalized_test_data_veg_input.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:075c60cf83a2d7ef68720b72c77f41a034d6418b67085e016fff6a5bdfef878c
-size 2631223074
+oid sha256:5d6345f35427703eae9bd718ec420566dd253c5dab6b17a3a9c99dbf6b4f18b2
+size 2631222089

veg/example_inference.py

@@ -3,7 +3,7 @@ import torch
 from MultiTaskConvLSTM import ConvLSTMNetwork
 from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, roc_auc_score
 import torch
-import toch.nn as nn
+import torch.nn as nn
 from tqdm.auto import tqdm
 from utils import (
     mse, mae, nash_sutcliffe_efficiency, r2_score, pearson_correlation,
@@ -45,7 +45,7 @@ bce_loss_fn = nn.BCELoss()       # For classification output
 
 optimizer = optim.AdamW(model.parameters(), lr = 0.005) 
 
-checkpoint = torch.load("MultiTaskConvLSTM_veg_variables")
+checkpoint = torch.load("MultiTaskConvLSTM_veg_variables.pth", map_location = device)
 model.load_state_dict(checkpoint['model_state_dict'])
 
 # If you want to move the model to the GPU (optional, depending on your setup)
@@ -146,7 +146,6 @@ def evaluate(model, test_loader, reg_loss_fn, class_loss_fn, device, variable_na
         "Pearson": pearson_correlation(y_true_reg_flat, y_pred_reg_flat),
         "Spearman": spearman_correlation(y_true_reg_flat, y_pred_reg_flat),
         "NSE": nash_sutcliffe_efficiency(y_true_reg_flat, y_pred_reg_flat),
-        "Percentage Error": percentage_error(y_true_reg_flat, y_pred_reg_flat),
         "Percentage Bias": percentage_bias(y_true_reg_flat, y_pred_reg_flat),
         "Kendall Tau": kendall_tau(y_true_reg_flat, y_pred_reg_flat),
         "Spatial Correlation": spatial_correlation(y_true_reg_flat, y_pred_reg_flat)}