Merge pull request #2315 from ds-wook/refactor/migrate-lightgcn-pytorch

refactor: migrate lightgcn pytorch

Author: miguelgfierro

examples/02_model_collaborative_filtering/lightgcn_deep_dive.ipynb

@@ -54,7 +54,6 @@
     pass  # skip this import if we are not in a Spark environment
 
 try:
-    from recommenders.models.deeprec.deeprec_utils import prepare_hparams
     from recommenders.models.deeprec.models.graphrec.lightgcn import LightGCN
     from recommenders.models.deeprec.DataModel.ImplicitCF import ImplicitCF
     from recommenders.models.ncf.ncf_singlenode import NCF
@@ -385,10 +384,32 @@ def prepare_training_lightgcn(train, test):
 
 
 def train_lightgcn(params, data):
-    hparams = prepare_hparams(**params)
-    model = LightGCN(hparams, data)
+    ctor_keys = {"embed_size", "n_layers", "seed"}
+    fit_keys = {
+        "epochs",
+        "learning_rate",
+        "batch_size",
+        "decay",
+        "eval_epoch",
+        "top_k",
+        "metrics",
+        "save_model",
+        "save_epoch",
+    }
+
+    ctor_kwargs = {k: params[k] for k in ctor_keys if k in params}
+    fit_kwargs = {k: params[k] for k in fit_keys if k in params}
+    if "MODEL_DIR" in params:
+        fit_kwargs["model_dir"] = params["MODEL_DIR"]
+
+    model = LightGCN(
+        n_users=data.n_users,
+        n_items=data.n_items,
+        norm_adj=data.get_norm_adj_mat(),
+        **ctor_kwargs,
+    )
     with Timer() as t:
-        model.fit()
+        model.fit(data, **fit_kwargs)
     return model, t
 
 

examples/06_benchmarks/benchmark_utils.py

@@ -1,7 +1,9 @@
 # Copyright (c) Recommenders contributors.
 # Licensed under the MIT License.
 
+import logging
 import random
+
 import numpy as np
 import pandas as pd
 import scipy.sparse as sp
@@ -13,6 +15,8 @@
     DEFAULT_PREDICTION_COL,
 )
 
+logger = logging.getLogger(__name__)
+
 
 class ImplicitCF(object):
     """Data processing class for GCN models which use implicit feedback.
@@ -156,7 +160,7 @@ def get_norm_adj_mat(self):
             if self.adj_dir is None:
                 raise FileNotFoundError
             norm_adj_mat = sp.load_npz(self.adj_dir + "/norm_adj_mat.npz")
-            print("Already load norm adj matrix.")
+            logger.info("Already load norm adj matrix.")
 
         except FileNotFoundError:
             norm_adj_mat = self.create_norm_adj_mat()
@@ -180,15 +184,15 @@ def create_norm_adj_mat(self):
         adj_mat[: self.n_users, self.n_users :] = R
         adj_mat[self.n_users :, : self.n_users] = R.T
         adj_mat = adj_mat.todok()
-        print("Already create adjacency matrix.")
+        logger.info("Already create adjacency matrix.")
 
         rowsum = np.array(adj_mat.sum(1))
         d_inv = np.power(rowsum + 1e-9, -0.5).flatten()
         d_inv[np.isinf(d_inv)] = 0.0
         d_mat_inv = sp.diags(d_inv)
         norm_adj_mat = d_mat_inv.dot(adj_mat)
         norm_adj_mat = norm_adj_mat.dot(d_mat_inv)
-        print("Already normalize adjacency matrix.")
+        logger.info("Already normalize adjacency matrix.")
 
         return norm_adj_mat.tocsr()