update best logq 0.2 mclsr, logq 1.0 sasrec, baselines

README.md

@@ -1,3 +1,5 @@
+
+# Steps on local machine
 <p align="center">
    <a>
       <img src="./assets/irec_logo.png" alt="cool irec logo" width="40%" height="40%">
@@ -40,26 +42,6 @@
    uv sync --frozen
    ```
 
-### Using pip
-
-1. Create and activate a virtual environment:
-   ```bash
-   python3 -m venv .venv
-   source ./.venv/bin/activate
-   ```
-
-2. Install dependencies:
-
-   **For development:**
-   ```bash
-   pip install -e ".[dev]"
-   ```
-
-   **For production:**
-   ```bash
-   pip install -e .
-   ```
-
 ## Preparing datasets
 All pre-processed datasets used in our experiments are available for download from our cloud storage. This is the fastest way to get started.
 
@@ -91,4 +73,4 @@ The script has 1 input argument: `params` which is the path to the json file wit
 
 -`callbacks` Different additional traning 
 
--`use_wandb` Enable Weights & Biases logging for experiment tracking
+-`use_wandb` Enable Weights & Biases logging for experiment tracking

configs/train/mclsr_baseline_clothing_sanity.json

@@ -0,0 +1,233 @@
+{
+  "experiment_name": "mclsr_Clothing",
+  "use_wandb": true,
+  "best_metric": "validation/ndcg@20",
+  "dataset": {
+    "type": "graph",
+    "use_user_graph": true,
+    "use_item_graph": true,
+    "neighborhood_size": 50,
+    "graph_dir_path": "./data/Clothing",
+    "dataset": {
+      "type": "mclsr",
+      "path_to_data_dir": "./data",
+      "name": "Clothing",
+      "max_sequence_length": 20,
+      "samplers": {
+        "num_negatives_val": 1280,
+        "num_negatives_train": 1280,
+        "type": "mclsr",
+        "negative_sampler_type": "random"
+      }
+    }
+  },
+  "dataloader": {
+    "train": {
+      "type": "torch",
+      "batch_size": 128,
+      "batch_processor": {
+        "type": "basic" 
+      },
+      "drop_last": true,
+      "shuffle": true,
+      "num_workers": 4,
+      "pin_memory" : true
+    },
+    "validation": {
+      "type": "torch",
+      "batch_size": 128,
+      "batch_processor": {
+        "type": "basic"
+      },
+      "drop_last": false,
+      "shuffle": false,
+      "num_workers": 4,
+      "pin_memory" : true
+    }
+  },
+  "model": {
+    "type": "mclsr",
+    "sequence_prefix": "item",
+    "user_prefix": "user",
+    "labels_prefix": "labels",
+    "candidate_prefix": "candidates",
+    "embedding_dim": 64,
+    "num_graph_layers": 2,
+    "dropout": 0.3,
+    "layer_norm_eps": 1e-9,
+    "graph_dropout": 0.3,
+    "initializer_range": 0.02,
+    "alpha": 0.5
+  },
+  "optimizer": {
+    "type": "basic",
+    "optimizer": {
+      "type": "adam",
+      "lr": 0.001
+    }
+  },
+  "loss": {
+    "type": "composite",
+    "losses": [
+      {
+        "type": "sampled_softmax",          
+        "queries_prefix": "combined_representation",
+        "positive_prefix": "label_representation",
+        "negative_prefix": "negative_representation", 
+        "output_prefix": "downstream_loss",
+        "weight": 1.0
+      },
+      {
+        "type": "fps",
+        "fst_embeddings_prefix": "sequential_representation",
+        "snd_embeddings_prefix": "graph_representation",
+        "output_prefix": "contrastive_interest_loss",
+        "weight": 1.0,
+        "temperature": 0.5
+      },
+      {
+        "type": "fps",
+        "fst_embeddings_prefix": "user_graph_user_embeddings",
+        "snd_embeddings_prefix": "common_graph_user_embeddings",
+        "output_prefix": "contrastive_user_feature_loss",
+        "weight": 0.05,
+        "temperature": 0.5
+      },
+      {
+        "type": "fps",
+        "fst_embeddings_prefix": "item_graph_item_embeddings",
+        "snd_embeddings_prefix": "common_graph_item_embeddings",
+        "output_prefix": "contrastive_item_feature_loss",
+        "weight": 0.05,
+        "temperature": 0.5
+      }
+    ],
+    "output_prefix": "loss"
+  },
+  "callback": {
+    "type": "composite",
+    "callbacks": [
+      {
+        "type": "metric",
+        "on_step": 1,
+        "loss_prefix": "loss"
+      },
+      {
+        "type": "metric",
+        "on_step": 1,
+        "loss_prefix": "downstream_loss"
+      },
+      {
+        "type": "metric",
+        "on_step": 1,
+        "loss_prefix": "contrastive_interest_loss"
+      },
+      {
+        "type": "metric",
+        "on_step": 1,
+        "loss_prefix": "contrastive_user_feature_loss"
+      },
+      {
+        "type": "metric",
+        "on_step": 1,
+        "loss_prefix": "contrastive_item_feature_loss"
+      },
+      {
+        "type": "validation",
+        "on_step": 64,
+        "pred_prefix": "predictions",
+        "labels_prefix": "labels",
+        "metrics": {
+          "ndcg@5": {
+            "type": "mclsr-ndcg",
+            "k": 5
+          },
+          "ndcg@10": {
+            "type": "mclsr-ndcg",
+            "k": 10
+          },
+          "ndcg@20": {
+            "type": "mclsr-ndcg",
+            "k": 20
+          },
+          "ndcg@50": {
+            "type": "mclsr-ndcg",
+            "k": 50
+          },
+          "recall@5": {
+            "type": "mclsr-recall",
+            "k": 5
+          },
+          "recall@10": {
+            "type": "mclsr-recall",
+            "k": 10
+          },
+          "recall@20": {
+            "type": "mclsr-recall",
+            "k": 20
+          },
+          "recall@50": {
+            "type": "mclsr-recall",
+            "k": 50
+          },
+          "hit@20": {
+            "type": "mclsr-hit",
+            "k": 20
+          },
+          "hit@50": {
+            "type": "mclsr-hit",
+            "k": 50
+          }
+        }
+      },
+      {
+        "type": "eval",
+        "on_step": 256,
+        "pred_prefix": "predictions",
+        "labels_prefix": "labels",
+        "metrics": {
+          "ndcg@5": {
+            "type": "mclsr-ndcg",
+            "k": 5
+          },
+          "ndcg@10": {
+            "type": "mclsr-ndcg",
+            "k": 10
+          },
+          "ndcg@20": {
+            "type": "mclsr-ndcg",
+            "k": 20
+          },
+          "ndcg@50": {
+            "type": "mclsr-ndcg",
+            "k": 50
+          },
+          "recall@5": {
+            "type": "mclsr-recall",
+            "k": 5
+          },
+          "recall@10": {
+            "type": "mclsr-recall",
+            "k": 10
+          },
+          "recall@20": {
+            "type": "mclsr-recall",
+            "k": 20
+          },
+          "recall@50": {
+            "type": "mclsr-recall",
+            "k": 50
+          },
+          "hit@20": {
+            "type": "mclsr-hit",
+            "k": 20
+          },
+          "hit@50": {
+            "type": "mclsr-hit",
+            "k": 50
+          }
+        }
+      }
+    ]
+  }
+}