model: graph_conv-relation_attention_h128_l3_edge_prediction | (graph_conv-relation_attention_h128_l3) | WandB: hh8kfhbv

config.json

@@ -2,7 +2,7 @@
   "name": null,
   "seed": 42,
   "deterministic": true,
-  "output_dir": "relation_prediction_relationattention_128e_v1",
+  "output_dir": "relation_prediction_relationattention_128e_v2",
   "model": {
     "encoder": "graph_conv",
     "hidden_channels": 128,
@@ -25,11 +25,11 @@
     "use_edge_encoder": true,
     "edge_encoder_dim": 32,
     "edge_encoder_dropout": 0.1,
-    "edge_encoder_init_bias": null,
-    "use_pretrained_model": false,
-    "pretrained_model_source": null,
-    "pretrained_model_path": null,
-    "pretrained_model_revision": null,
+    "edge_encoder_init_bias": 0.0,
+    "use_pretrained_model": true,
+    "pretrained_model_source": "huggingface",
+    "pretrained_model_path": "seanhacks/relation_prediction_relationattention_128e",
+    "pretrained_model_revision": "20251228",
     "pretrained_model_load_head": true,
     "pretrained_model_freeze_encoder_weights": false,
     "pretrained_model_freeze_head_weights": false
@@ -65,7 +65,7 @@
     "optimizer": "adamw",
     "scheduler": null,
     "gradient_clip_val": null,
-    "epochs": 1000,
+    "epochs": 1500,
     "batches_per_epoch": 20,
     "accelerator": "auto",
     "devices": 1,
@@ -93,7 +93,7 @@
       "hidden_128",
       "layers_3",
       "lr_0.0005",
-      "epochs_1000"
+      "epochs_1500"
     ],
     "log_model": false,
     "mode": "online",