try with stock data

data/datasets.py

@@ -147,4 +147,4 @@ class ForecastDataset(Dataset):
         return x, y, x_time, y_time
 
     def inverse_transform(self, data):
-        return self.scaler.inverse_transform(data)
+        return self.scaler.inverse_transform(data)

experiments/base.py

@@ -62,9 +62,12 @@ class Experiment(ABC):
             # write command file
             command_file = os.path.join(instance_path, 'command')
             with open(command_file, 'w') as cmd:
+                # cmd.write(f'python -m {module} '
+                #           f'--config_path={instance_config_path} '
+                #           f'run >> {instance_path}/instance.log 2>&1')
                 cmd.write(f'python -m {module} '
-                          f'--config_path={instance_config_path} '
-                          f'run >> {instance_path}/instance.log 2>&1')
+                                        f'--config_path={instance_config_path} '
+                                        f'run 2>&1 | tee -a {instance_path}/instance.log')
 
     @abstractmethod
     def instance(self):
@@ -102,4 +105,4 @@ class Experiment(ABC):
     def build_experiment(self):
         if EXPERIMENTS_PATH in str(self.root):
             raise Exception('Cannot build ensemble from ensemble member configuration.')
-        self.build()
+        self.build()

experiments/configs/Stocks/96M.gin

@@ -0,0 +1,37 @@
+build.experiment_name = 'Stocks/96M'
+build.module = 'experiments.forecast'
+build.repeat = 1
+build.variables_dict = {
+}
+
+instance.model_type = 'deeptime'
+instance.save_vals = False
+
+get_optimizer.lr = 1e-3
+get_optimizer.lambda_lr = 1.
+get_optimizer.weight_decay = 0.
+
+get_scheduler.warmup_epochs = 5
+
+get_data.batch_size = 256
+
+train.loss_name = 'mse'
+train.epochs = 50
+train.clip = 10.
+
+Checkpoint.patience = 7
+
+deeptime.layer_size = 256
+deeptime.inr_layers = 5
+deeptime.n_fourier_feats = 4096
+deeptime.scales = [0.01, 0.1, 1, 5, 10, 20, 50, 100]
+
+ForecastDataset.data_path = 'stocks/OXY_2019.csv.gz'
+ForecastDataset.target = 'RSMKs_18_144_72'
+ForecastDataset.scale = True
+ForecastDataset.cross_learn = False
+ForecastDataset.time_features = []
+ForecastDataset.normalise_time_features = True
+ForecastDataset.features = 'M'
+ForecastDataset.horizon_len = 96
+ForecastDataset.lookback_mult = 1

experiments/configs/Stocks/96S.gin

@@ -0,0 +1,37 @@
+build.experiment_name = 'Stocks/96S'
+build.module = 'experiments.forecast'
+build.repeat = 1
+build.variables_dict = {
+}
+
+instance.model_type = 'deeptime'
+instance.save_vals = False
+
+get_optimizer.lr = 1e-3
+get_optimizer.lambda_lr = 1.
+get_optimizer.weight_decay = 0.
+
+get_scheduler.warmup_epochs = 5
+
+get_data.batch_size = 256
+
+train.loss_name = 'mse'
+train.epochs = 50
+train.clip = 10.
+
+Checkpoint.patience = 7
+
+deeptime.layer_size = 256
+deeptime.inr_layers = 5
+deeptime.n_fourier_feats = 4096
+deeptime.scales = [0.01, 0.1, 1, 5, 10, 20, 50, 100]
+
+ForecastDataset.data_path = 'stocks/OXY_2019.csv.gz'
+ForecastDataset.target = 'RSMKs_18_144_72'
+ForecastDataset.scale = True
+ForecastDataset.cross_learn = False
+ForecastDataset.time_features = []
+ForecastDataset.normalise_time_features = True
+ForecastDataset.features = 'S'
+ForecastDataset.horizon_len = 96
+ForecastDataset.lookback_mult = 3

experiments/configs/Stocks/96Splus.gin

@@ -0,0 +1,37 @@
+build.experiment_name = 'Stocks/96Splus'
+build.module = 'experiments.forecast'
+build.repeat = 1
+build.variables_dict = {
+}
+
+instance.model_type = 'deeptime2'
+instance.save_vals = False
+
+get_optimizer.lr = 1e-3
+get_optimizer.lambda_lr = 1.
+get_optimizer.weight_decay = 0.
+
+get_scheduler.warmup_epochs = 5
+
+get_data.batch_size = 256
+
+train.loss_name = 'mse'
+train.epochs = 50
+train.clip = 10.
+
+Checkpoint.patience = 7
+
+deeptime2.layer_size = 256
+deeptime2.inr_layers = 5
+deeptime2.n_fourier_feats = 4096
+deeptime2.scales = [0.01, 0.1, 1, 5, 10, 20, 50, 100]
+
+ForecastDataset.data_path = 'stocks/OXY_2019.csv.gz'
+ForecastDataset.target = 'RSMKs_18_144_72'
+ForecastDataset.scale = True
+ForecastDataset.cross_learn = False
+ForecastDataset.time_features = []
+ForecastDataset.normalise_time_features = True
+ForecastDataset.features = 'S'
+ForecastDataset.horizon_len = 96
+ForecastDataset.lookback_mult = 3

experiments/configs/Stocks/96Splusshort.gin

@@ -0,0 +1,37 @@
+build.experiment_name = 'Stocks/96Splusshort'
+build.module = 'experiments.forecast'
+build.repeat = 1
+build.variables_dict = {
+}
+
+instance.model_type = 'deeptime2'
+instance.save_vals = False
+
+get_optimizer.lr = 1e-3
+get_optimizer.lambda_lr = 1.
+get_optimizer.weight_decay = 0.
+
+get_scheduler.warmup_epochs = 5
+
+get_data.batch_size = 256
+
+train.loss_name = 'mse'
+train.epochs = 50
+train.clip = 10.
+
+Checkpoint.patience = 7
+
+deeptime2.layer_size = 256
+deeptime2.inr_layers = 5
+deeptime2.n_fourier_feats = 4096
+deeptime2.scales = [0.01, 0.1, 1, 5, 10, 20, 50, 100]
+
+ForecastDataset.data_path = 'stocks/OXY_2019.csv.gz'
+ForecastDataset.target = 'RSMKs_18_144_72'
+ForecastDataset.scale = True
+ForecastDataset.cross_learn = False
+ForecastDataset.time_features = []
+ForecastDataset.normalise_time_features = True
+ForecastDataset.features = 'S'
+ForecastDataset.horizon_len = 6
+ForecastDataset.lookback_mult = 8

experiments/configs/Stocks/96Sshort.gin

@@ -0,0 +1,37 @@
+build.experiment_name = 'Stocks/96Sshort'
+build.module = 'experiments.forecast'
+build.repeat = 1
+build.variables_dict = {
+}
+
+instance.model_type = 'deeptime'
+instance.save_vals = False
+
+get_optimizer.lr = 1e-3
+get_optimizer.lambda_lr = 1.
+get_optimizer.weight_decay = 0.
+
+get_scheduler.warmup_epochs = 5
+
+get_data.batch_size = 256
+
+train.loss_name = 'mse'
+train.epochs = 50
+train.clip = 10.
+
+Checkpoint.patience = 7
+
+deeptime.layer_size = 256
+deeptime.inr_layers = 5
+deeptime.n_fourier_feats = 4096
+deeptime.scales = [0.01, 0.1, 1, 5, 10, 20, 50, 100]
+
+ForecastDataset.data_path = 'stocks/OXY_2019.csv.gz'
+ForecastDataset.target = 'RSMKs_18_144_72'
+ForecastDataset.scale = True
+ForecastDataset.cross_learn = False
+ForecastDataset.time_features = []
+ForecastDataset.normalise_time_features = True
+ForecastDataset.features = 'S'
+ForecastDataset.horizon_len = 6
+ForecastDataset.lookback_mult = 8

mjc_notes.md

@@ -1,4 +1,5 @@
 # install environment
+
 ```sh
 # try with pip torch WORKS!
 export PROJ=deeptime
@@ -13,11 +14,12 @@ pip install tsai
 
 # note that I've also recorded the env in requirements
 
-python -m experiments.forecast --config_path=storage/experiments/Exchange/192S/repeat=0/config.gin run >> storage/experiments/Exchange/192S/repeat=0/instance.log 2>&1%
+python -m experiments.forecast --config_path=storage/experiments/Exchange/192S/repeat=0/config.gin run | tee -a storage/experiments/Exchange/192S/repeat=0/instance.log 2>&1%
 ```
+
 # run
 
-```
+```sh
 python -m experiments.forecast --config_path=storage/experiments/Exchange/96S/repeat=0/config.gin run
 
 python -m experiments.forecast --config_path=storage/experiments/Exchange/96Splus/repeat=0/config.gin run
@@ -31,3 +33,16 @@ python -m experiments.forecast --config_path=storage/experiments/Exchange/96Ssho
 # Lessons
 
 Single variate works much better. The output is not just a straight line. Likely because we have limited the output, not the input
+
+# stocks
+
+```sh
+python -m experiments.forecast --config_path=experiments/configs/Stocks/96S.gin build_experiment 
+python -m experiments.forecast --config_path=storage/experiments/Stocks/96S/repeat=0/config.gin run
+```
+
+
+```
+make build-all path=experiments/configs/Stocks
+./run.sh
+```

run.sh

@@ -1,4 +1,4 @@
-for dataset in ECL ETTm2 Exchange ILI Traffic Weather; do
+for dataset in ECL ETTm2 Exchange ILI Traffic Weather Stocks; do
   for instance in `/bin/ls -d storage/experiments/$dataset/*/*`; do
       echo $instance
       make run command=${instance}/command