Add gmadl loss function

configs/experiments/temporal-fusion-btcusdt-gmadl.yaml

@@ -0,0 +1,67 @@
+future_window:
+  value: 1
+past_window:
+  value: 24
+batch_size:
+  value: 64
+max_epochs:
+  value: 30
+data:
+  value:
+    dataset: "btc-usdt-5m:latest"
+    sliding_window: 4
+    validation: 0.2
+    fields:
+      time_index: "time_index"
+      target: "returns"
+      group_ids: ["group_id"]
+      dynamic_unknown_real:
+        - "high_price"
+        - "low_price"
+        - "open_price"
+        - "close_price"
+        - "volume"
+        - "open_to_close_price"
+        - "high_to_close_price"
+        - "low_to_close_price"
+        - "high_to_low_price"
+        - "returns"
+        - "log_returns"
+        - "vol_1h"
+        - "macd"
+        - "macd_signal"
+        - "rsi"
+        - "low_bband_to_close_price"
+        - "up_bband_to_close_price"
+        - "mid_bband_to_close_price"
+        - "sma_1h_to_close_price"
+        - "sma_1d_to_close_price"
+        - "sma_7d_to_close_price"
+        - "ema_1h_to_close_price"
+        - "ema_1d_to_close_price"
+      dynamic_unknown_cat: []
+      dynamic_known_real: []
+      dynamic_known_cat:
+        - "hour"
+      static_real:
+        - "effective_rates"
+        - "vix_close_price"
+        - "fear_greed_index"
+        - "vol_1d"
+        - "vol_7d"
+      static_cat:
+        - "weekday"
+loss:
+  value:
+   name: "GMADL"
+   a: 1000
+   b: 2
+model:
+  value:
+    name: "TemporalFusionTransformer"
+    hidden_size: 64
+    dropout: 0.1
+    attention_head_size: 2
+    hidden_continuous_size: 8
+    learning_rate: 0.001
+    optimizer: "Adam"

configs/sweeps/temporal-fusion-btcusdt-gmadl.yaml

@@ -0,0 +1,38 @@
+program: ./scripts/train.py
+project: wne-masters-thesis-testing
+command:
+  - ${env}
+  - ${interpreter}
+  - ${program}
+  - "./configs/experiments/temporal-fusion-btcusdt-gmadl.yaml"
+  - "--patience"
+  - "10"
+method: random
+metric:
+  goal: minimize
+  name: val_loss
+parameters:
+  past_window:
+    distribution: int_uniform
+    min: 20
+    max: 100
+  batch_size:
+    values: [64, 128, 256]
+  model:
+    parameters:
+      name:
+        value: "TemporalFusionTransformer"
+      share_single_variable_networks:
+        value: false
+      hidden_size:
+        values: [128, 256, 512, 1024]
+      dropout:
+        values: [0.0, 0.1, 0.2, 0.3]
+      attention_head_size:
+        values: [1, 2, 4, 6]
+      hidden_continuous_size: 
+        values: [4, 8, 16, 32]
+      learning_rate:
+        value: 0.001
+      optimizer:
+        value: "Adam"

scripts/train.py

@@ -15,6 +15,8 @@ from pytorch_forecasting.metrics import MAE, RMSE
 from pytorch_forecasting import QuantileLoss
 from pytorch_forecasting.models.temporal_fusion_transformer import TemporalFusionTransformer
 
+from ml.loss import GMADL
+
 
 def get_args():
     parser = argparse.ArgumentParser(
@@ -130,6 +132,12 @@ def get_loss(config):
     if loss_name == 'Quantile':
         return QuantileLoss(config['loss']['quantiles'])
 
+    if loss_name == 'GMADL':
+        return GMADL(
+            a=config['loss']['a'],
+            b=config['loss']['b']
+        )
+
     raise ValueError("Unknown loss")