feat(Pytorch): added custom model to pytorch-forecasting (#8)

* feat: Refractored and created new model. Pipeline not ready yet. * feat: Implemented and refactored a data pipeline. * ref: Refractored to make more sense. * feat: Training works now with models that you can change. * feat: Added predict function but without working instructions. * feat: gitignore.
2026-08-19 13:58:11 +00:00 · 2021-11-18 10:59:06 +01:00
parent 6e192ebc8a
commit d4676e099b
12 changed files with 353 additions and 18 deletions
@@ -0,0 +1,119 @@
+#%%
+import pytorch_lightning as pl
+from pytorch_lightning.callbacks import EarlyStopping, LearningRateMonitor
+
+from pytorch_forecasting import TimeSeriesDataSet, TemporalFusionTransformer
+from pytorch_forecasting.data import GroupNormalizer
+from pytorch_forecasting.metrics import QuantileLoss
+
+import sys
+sys.path.insert(0, '..')
+
+from load_data import load_files
+
+print("success")
+
+#%%
+# load data
+data = load_files('../data/', add_features=True, log_returns=False, narrow_format=True)
+# need to treat time as an independent column
+data = data.reset_index().rename({'index':'time'}, axis = 'columns')
+# we need a `time_idx` column for pytorch-forecasting, so we convert the time column to a time index by encoding the dates as consecutive days from the first date.
+data['time_idx'] = (data['time']-data['time'].min()).astype('timedelta64[D]').astype(int)+1
+# volume needs some love before we can use it
+data.drop(columns=['volume'], inplace=True)
+
+data['month'] = data['month'].astype(str)
+data['day_month'] = data['day_month'].astype(str)
+data['day_week'] = data['day_week'].astype(str)
+
+
+
+#%%
+# define dataset
+max_encoder_length = 36 # this is the look-back window, see https://github.com/jdb78/pytorch-forecasting/issues/448
+max_prediction_length = 6
+# training_cutoff = "YYYY-MM-DD"  # day for cutoff
+
+# training_cutoff = data["time_idx"].max() - max_prediction_length
+
+#%%
+data.head()
+data.describe()
+
+#%%
+training = TimeSeriesDataSet(
+    data, # data[lambda x: x.date < training_cutoff],
+    time_idx= 'time_idx',
+    target= 'returns',
+    # weight="weight",
+    group_ids=[ 'ticker' ],
+    min_encoder_length = max_encoder_length,##max_encoder_length//2,
+    max_encoder_length = max_encoder_length,
+    min_prediction_length = 1,
+    max_prediction_length = 1,#max_prediction_length,
+    static_categoricals=[  ],
+    static_reals=[  ],
+    time_varying_known_categoricals=[ 'day_month', 'day_week', 'month' ],
+    time_varying_known_reals=[ 'vol_10', 'vol_20', 'vol_30', 'vol_60', 'mom_10', 'mom_20', 'mom_30', 'mom_60', 'mom_90'],
+    time_varying_unknown_categoricals=[  ],
+    time_varying_unknown_reals=[ 'returns' ],
+    allow_missing_timesteps=True,
+    target_normalizer=GroupNormalizer(
+        groups=['ticker'], transformation="softplus")
+)
+
+#%%
+print(training.index.time)
+print(training.index.time.max())
+
+#%%
+# create validation and training dataset
+
+validation = TimeSeriesDataSet.from_dataset(training, data, predict=True, stop_randomization=True)
+batch_size = 128
+train_dataloader = training.to_dataloader(train=True, batch_size=batch_size, num_workers=0)
+val_dataloader = validation.to_dataloader(train=False, batch_size=batch_size, num_workers=0)
+
+# define trainer with early stopping
+early_stop_callback = EarlyStopping(monitor="val_loss", min_delta=1e-4, patience=1, verbose=False, mode="min")
+lr_logger = LearningRateMonitor()
+trainer = pl.Trainer(
+    max_epochs=100,
+    gpus=0,
+    gradient_clip_val=0.1,
+    limit_train_batches=30,
+    callbacks=[lr_logger, early_stop_callback],
+)
+
+# create the model
+tft = TemporalFusionTransformer.from_dataset(
+    training,
+    learning_rate=0.03,
+    hidden_size=32,
+    attention_head_size=1,
+    dropout=0.1,
+    hidden_continuous_size=16,
+    output_size=7,
+    loss=QuantileLoss(),
+    log_interval=2,
+    reduce_on_plateau_patience=4
+)
+print(f"Number of parameters in network: {tft.size()/1e3:.1f}k")
+
+# find optimal learning rate (set limit_train_batches to 1.0 and log_interval = -1)
+res = trainer.tuner.lr_find(
+    tft, train_dataloader=train_dataloader, val_dataloaders=val_dataloader, early_stop_threshold=1000.0, max_lr=0.3,
+)
+
+print(f"suggested learning rate: {res.suggestion()}")
+fig = res.plot(show=True, suggest=True)
+fig.show()
+
+# fit the model
+trainer.fit(
+    tft, train_dataloader=train_dataloader, val_dataloaders=val_dataloader,
+)
+
+
+# %%
@@ -0,0 +1,60 @@
+
+import torch
+from torch import nn
+from torch.nn import functional as F
+from torch.utils.data import DataLoader, random_split
+import pytorch_lightning as pl
+
+
+class LitManualAutoEncoder(pl.LightningModule):
+
+    def __init__(self):
+        super().__init__()
+        self.encoder = nn.Sequential(nn.Linear(28 * 28, 128), nn.ReLU(), nn.Linear(128, 3))
+        self.decoder = nn.Sequential(nn.Linear(3, 128), nn.ReLU(), nn.Linear(128, 28 * 28))
+        print("success")
+
+    def training_step(self, batch, batch_idx):
+        # --------------------------
+        # REPLACE WITH YOUR OWN
+        opt_a = self.optimizers()
+        
+        x, y = batch
+        x = x.view(x.size(0), -1)
+        z = self.encoder(x)
+        x_hat = self.decoder(z)
+        loss = F.mse_loss(x_hat, x)
+
+        # backward acts like normal backward
+        self.manual_backward(loss, opt_a, retain_graph=True)
+        self.manual_backward(loss, opt_a)
+        opt_a.step()
+        opt_a.zero_grad()
+
+        # --------------------------
+
+    def validation_step(self, batch, batch_idx):
+        # --------------------------
+        # REPLACE WITH YOUR OWN
+        x, y = batch
+        x = x.view(x.size(0), -1)
+        z = self.encoder(x)
+        x_hat = self.decoder(z)
+        loss = F.mse_loss(x_hat, x)
+        self.log('val_loss', loss)
+        # --------------------------
+
+    def test_step(self, batch, batch_idx):
+        # --------------------------
+        # REPLACE WITH YOUR OWN
+        x, y = batch
+        x = x.view(x.size(0), -1)
+        z = self.encoder(x)
+        x_hat = self.decoder(z)
+        loss = F.mse_loss(x_hat, x)
+        self.log('test_loss', loss)
+        # --------------------------
+
+    def configure_optimizers(self):
+        optimizer = torch.optim.Adam(self.parameters(), lr=1e-3)
+        return optimizer
@@ -0,0 +1,48 @@
+from load_data import load_files
+import pandas as pd
+# from tensorflow import keras
+from utils.normalize import normalize
+# import tensorflow as tf
+from utils.visualize import visualize_loss
+
+from torch.utils.data import DataLoader, random_split
+from model_lightning import LitManualAutoEncoder 
+import pytorch_lightning as pl
+
+#%%
+data = load_files('data/', False)
+data.reset_index(drop=True, inplace=True)
+data = data[[column for column in data.columns if not column.endswith('volume')]]
+
+data.head()
+
+#%%
+ticker_to_predict = 'ETH_returns'
+
+learning_rate = 0.002
+batch_size = 64
+epochs =  100
+
+split_fraction = 0.715
+train_split = int(split_fraction * int(data.shape[0]))
+
+past = 10
+future = 1
+
+start = past + future
+end = start + train_split
+
+
+# train = DataLoader(train, batch_size=32)
+# test = DataLoader(test, batch_size=32)
+# val = DataLoader(val, batch_size=32)
+
+
+# init model
+ae = LitManualAutoEncoder()
+
+# Initialize a trainer
+trainer = pl.Trainer(gpus=1, max_epochs=3, progress_bar_refresh_rate=20)
+
+# Train the model ⚡
+# trainer.fit(ae, train, val)