chore: use hydra config parser

Author: JulesBelveze

examples/forecasting/config.yaml

@@ -1,28 +1,35 @@
 data:
+  _target_: tsa.dataset.TimeSeriesDataset
+  batch_size: 16
   categorical_cols: [ ]
-  label_col: ["AH"]
   index_col: "Date_Time"
+  target_col: [ "AH" ]
+  data_path: "../data/AirQualityUCI.csv"
   prediction_window: 1
-  seq_len: 3
+  seq_length: 3
+  task:
+    _target_: tsa.dataset.Tasks
+    value: prediction
 
 training:
-  num_epochs: 100
-  batch_size: 16
+  batch_size: ${data.batch_size}
+  denoising: False
+  directions: 1
+  gradient_accumulation_steps: 1
+  hidden_size_encoder: 64
+  hidden_size_decoder: 64
+  input_att: True
   lr: 1e-5
+  lrs_step_size: 5000
+  max_grad_norm: 0.1
+  num_epochs: 100
+  output_size: 1
   reg1: True
   reg2: False
   reg_factor1: 1e-4
   reg_factor2: 1e-4
-  hidden_size_encoder: 64
-  hidden_size_decoder: 64
-  output_size: 1
-  input_att: True
+  seq_len: ${data.seq_length}
   temporal_att: True
-  denoising: False
-  directions: 1
-  max_grad_norm: 0.1
-  gradient_accumulation_steps: 1
-  lrs_step_size: 5000
 
 general:
   do_eval: True

examples/forecasting/run_forecasting.py

@@ -1,36 +1,28 @@
 import hydra
-import pandas as pd
 import torch
 import torch.nn as nn
+from hydra.utils import instantiate
 
-from tsa import TimeSeriesDataset, AutoEncForecast, train, evaluate
+from tsa import AutoEncForecast, train, evaluate
 from tsa.utils import load_checkpoint
 
 device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 
 
 @hydra.main(config_path="./", config_name="config")
-def run(args):
-    df = pd.read_csv("data/AirQualityUCI.csv", index_col=args.data.index_col)
-
-    ts = TimeSeriesDataset(
-        data=df,
-        categorical_cols=args.data.categorical_cols,
-        target_col=args.data.label_col,
-        seq_length=args.data.seq_len,
-        prediction_window=args.data.prediction_window
-    )
-    train_iter, test_iter, nb_features = ts.get_loaders(batch_size=args.data.batch_size)
-
-    model = AutoEncForecast(args.training, input_size=nb_features).to(device)
+def run(cfg):
+    ts = instantiate(cfg.data)
+    train_iter, test_iter, nb_features = ts.get_loaders()
+
+    model = AutoEncForecast(cfg.training, input_size=nb_features).to(device)
     criterion = nn.MSELoss()
-    optimizer = torch.optim.Adam(model.parameters(), lr=args.training.lr)
+    optimizer = torch.optim.Adam(model.parameters(), lr=cfg.training.lr)
 
-    if args.general.do_eval and args.general.ckpt:
-        model, _, loss, epoch = load_checkpoint(args.general.ckpt, model, optimizer, device)
-        evaluate(test_iter, loss, model, args, ts)
-    elif args.general.do_train:
-        train(train_iter, test_iter, model, criterion, optimizer, args, ts)
+    if cfg.general.do_train:
+        train(train_iter, test_iter, model, criterion, optimizer, cfg, ts)
+    if cfg.general.do_eval and cfg.general.get("ckpt", False):
+        model, _, loss, epoch = load_checkpoint(cfg.general.ckpt, model, optimizer, device)
+        evaluate(test_iter, loss, model, cfg, ts)
 
 
 if __name__ == "__main__":

examples/reconstruction/config.yaml

@@ -1,28 +1,34 @@
 data:
+  _target_: tsa.dataset.TimeSeriesDataset
+  batch_size: 16
   categorical_cols: [ ]
-  label_col: []
   index_col: "Date_Time"
+  target_col: [ ]
+  data_path: "../data/AirQualityUCI.csv"
   prediction_window: 1
-  seq_len: 3
-  batch_size: 16
+  seq_length: 3
+  task:
+    _target_: tsa.dataset.Tasks
+    value: reconstruction
 
 training:
-  num_epochs: 100
+  denoising: False
+  directions: 1
+  gradient_accumulation_steps: 1
+  hidden_size_encoder: 64
+  hidden_size_decoder: 64
+  input_att: True
   lr: 1e-5
+  lrs_step_size: 5000
+  max_grad_norm: 0.1
+  num_epochs: 100
+  output_size: 13
   reg1: True
   reg2: False
   reg_factor1: 1e-4
   reg_factor2: 1e-4
-  hidden_size_encoder: 64
-  hidden_size_decoder: 64
-  input_att: True
+  seq_len: ${data.seq_length}
   temporal_att: True
-  output_size: 1
-  denoising: False
-  max_grad_norm: 0.1
-  lrs_step_size: 5000
-  gradient_accumulation_steps: 1
-  directions: 1
 
 general:
   do_eval: True

examples/reconstruction/run_reconstruction.py

@@ -1,36 +1,28 @@
 import hydra
-import pandas as pd
 import torch
 import torch.nn as nn
+from hydra.utils import instantiate
 
+from tsa import AutoEncForecast, train, evaluate
 from tsa.utils import load_checkpoint
-from tsa import TimeSeriesDataset, AutoEncForecast, train, evaluate
 
 device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
 
 
 @hydra.main(config_path="./", config_name="config")
-def run(args):
-    df = pd.read_csv("data/AirQualityUCI.csv", index_col=args.data.index_col)
-
-    ts = TimeSeriesDataset(
-        data=df,
-        categorical_cols=args.data.categorical_cols,
-        target_col=args.data.label_col,
-        seq_length=args.data.seq_len,
-        prediction_window=args.data.prediction_window
-    )
-    train_iter, test_iter, nb_features = ts.get_loaders(batch_size=args.data.batch_size)
-
-    model = AutoEncForecast(args, input_size=nb_features).to(device)
+def run(cfg):
+    ts = instantiate(cfg.data)
+    train_iter, test_iter, nb_features = ts.get_loaders()
+
+    model = AutoEncForecast(cfg.training, input_size=nb_features).to(device)
     criterion = nn.MSELoss()
-    optimizer = torch.optim.Adam(model.parameters(), lr=args.training.lr)
+    optimizer = torch.optim.Adam(model.parameters(), lr=cfg.training.lr)
 
-    if args.general.do_eval and args.general.ckpt:
-        model, _, loss, epoch = load_checkpoint(args.general.ckpt, model, optimizer, device)
-        evaluate(test_iter, loss, model, args, ts)
-    elif args.general.do_train:
-        train(train_iter, test_iter, model, criterion, optimizer, args, ts)
+    if cfg.general.do_eval and cfg.general.get("ckpt", False):
+        model, _, loss, epoch = load_checkpoint(cfg.general.ckpt, model, optimizer, device)
+        evaluate(test_iter, loss, model, cfg, ts)
+    elif cfg.general.do_train:
+        train(train_iter, test_iter, model, criterion, optimizer, cfg, ts)
 
 
 if __name__ == "__main__":

poetry.lock

@@ -12,7 +12,7 @@ sklearn = "^0.0"
 matplotlib = "^3.3.4"
 tensorboardX = "^2.1"
 tqdm = "^4.59.0"
-hydra-core = "^1.1.2"
+hydra-core = "^1.2.0"
 
 [tool.poetry.dev-dependencies]
 

pyproject.toml

@@ -1,4 +1,3 @@
-from .config import config
 from .dataset import TimeSeriesDataset
 from .eval import evaluate
 from .train import train