added parallel tuning of hyperpars

supsi-dacd-isaac · Apr 9, 2024 · 44bcf14 · 44bcf14
1 parent 6474478
commit 44bcf14
Show file tree

Hide file tree

Showing 2 changed files with 34 additions and 20 deletions.
diff --git a/pyforecaster/forecasting_models/holtwinters.py b/pyforecaster/forecasting_models/holtwinters.py
@@ -19,7 +19,24 @@ def hankel(x, n, generate_me=None):
         h.append(x[w + g].reshape(-1, 1))
     return np.hstack(h)
 
-def tune_hyperpars(x, model_class, hyperpars, n_trials=100, targets_names=None, verbose=True, return_model=True, **model_init_kwargs):
+def fit_sample(p, hyperpars,model_class, model_init_kwargs, x, targets_names=None):
+    pars_dict = {k: v for k, v in zip(hyperpars.keys(), p)}
+    model_init_kwargs.update(pars_dict)
+    model = model_class(**model_init_kwargs)
+    if targets_names is not None:
+        subscores = []
+        for c in targets_names:
+            model_init_kwargs.update({'target_name': c, 'targets_names': None})
+            model = model_class(**model_init_kwargs)
+            # features are all x columns except the targets and current target c
+            features_names = [n for n in x.columns if n not in set(targets_names) - {c}]
+            subscores.append(score_autoregressive(model, x[features_names], target_name=c))
+        score = np.mean(subscores)
+    else:
+        score = score_autoregressive(model, x)
+    return score
+
+def tune_hyperpars(x, model_class, hyperpars, n_trials=100, targets_names=None, verbose=True, return_model=True, parallel=True, **model_init_kwargs):
     """
     :param x: pd.DataFrame (n, n_cov)
     :param y: pd.Series (n)
@@ -37,23 +54,20 @@ def tune_hyperpars(x, model_class, hyperpars, n_trials=100, targets_names=None,
     pars_cartridge = np.random.rand(n_trials, n_pars) * (par_maxima-par_minima) + par_minima
     model_init_kwargs.update({'optimize_hyperpars':False})
     model_init_kwargs_0 = deepcopy(model_init_kwargs)
-    scores = []
-    generator = tqdm(pars_cartridge) if verbose else pars_cartridge
-    for i, p in enumerate(generator):
-        pars_dict = {k:v for k, v in zip(hyperpars.keys(), p)}
-        model_init_kwargs.update(pars_dict)
-        model = model_class(**model_init_kwargs)
-        if targets_names is not None:
-            subscores = []
-            for c in targets_names:
-                model_init_kwargs.update({'target_name':c, 'targets_names':None})
-                model = model_class(**model_init_kwargs)
-                # features are all x columns except the targets and current target c
-                features_names = [n for n in x.columns if n not in set(targets_names) - {c}]
-                subscores.append(score_autoregressive(model, x[features_names], target_name=c))
-            scores.append(np.mean(subscores))
-        else:
-            scores.append(score_autoregressive(model, x))
+
+
+    if parallel:
+        from concurrent.futures import ProcessPoolExecutor
+        with ProcessPoolExecutor() as executor:
+            scores = list(tqdm(executor.map(partial(fit_sample, hyperpars=hyperpars, model_class=model_class,
+                                               model_init_kwargs=model_init_kwargs, x=x, targets_names=targets_names),
+                                       pars_cartridge), total=n_trials, desc='Tuning hyperpars for {}'.format(model_class.__name__)))
+
+    else:
+        generator = tqdm(pars_cartridge) if verbose else pars_cartridge
+        scores = []
+        for i, p in enumerate(generator):
+            scores.append(fit_sample(p, hyperpars, model_class, model_init_kwargs, x, targets_names=targets_names))
     if verbose:
         plt.figure()
         t = np.linspace(0.01, 0.99, 30)

diff --git a/tests/test_nns.py b/tests/test_nns.py
@@ -501,8 +501,8 @@ def test_invertible_causal_nn(self):
         y_hat = m.predict(e_te.iloc[:, :144])
 
         m = CausalInvertibleNN(learning_rate=1e-2, batch_size=200, load_path=None, n_hidden_x=144,
-                               n_layers=2, normalize_target=False, n_epochs=5, stopping_rounds=20, rel_tol=-1,
-                               end_to_end='full', n_hidden_y=300, n_prediction_layers=3, n_out=144).fit(e_tr.iloc[:, :144], e_tr.iloc[:, -144:])
+                               n_layers=3, normalize_target=False, n_epochs=5, stopping_rounds=20, rel_tol=-1,
+                               end_to_end='quasi', n_hidden_y=300, n_prediction_layers=3, n_out=144).fit(e_tr.iloc[:, :144], e_tr.iloc[:, -144:])
 
         z_hat_ete = m.predict(e_te.iloc[:, :144])