best val_rmse 0.0102

config.json

@@ -1,5 +1,5 @@
 {
-  "best_val_loss": 0.0003782084968406707,
+  "best_val_loss": 0.010232063010334969,
   "comet_ai_key": null,
   "context_observations": {
     "add_rem": true,
@@ -34,7 +34,7 @@
     "same_route": true,
     "time": 0.0
   },
-  "experiment_dir": "/home/cesarali/Pharma/generative_pk/results/comet/node_pk_compartments/fb0874f7e0e7492393bb882525283b3a",
+  "experiment_dir": "/home/cesarali/Pharma/generative_pk/results/comet/node_pk_compartments/dacd9625a8524df685263795265fe04a",
   "experiment_indentifier": null,
   "experiment_name": "node_pk_compartments",
   "hf_model_card_path": [
@@ -171,11 +171,11 @@
       "simulated_ou_as_rates.tr"
     ],
     "test_protocol": "node",
-    "test_size": 5,
+    "test_size": 128,
     "tqdm_progress": false,
-    "train_size": 20,
+    "train_size": 1000,
     "val_protocol": "node",
-    "val_size": 5,
+    "val_size": 128,
     "z_score_normalization": false
   },
   "model_type": "node_pk",
@@ -188,15 +188,15 @@
     "combine_latent_mode": "mlp",
     "cov_proj_dim": 16,
     "decoder_attention_layers": 2,
-    "decoder_hidden_dim": 32,
+    "decoder_hidden_dim": 128,
     "decoder_name": "RNNDecoder",
-    "decoder_num_layers": 2,
-    "decoder_rnn_hidden_dim": 20,
+    "decoder_num_layers": 4,
+    "decoder_rnn_hidden_dim": 256,
     "drift_activation": "Tanh",
     "drift_num_layers": 2,
     "dropout": 0.1,
-    "encoder_rnn_hidden_dim": 20,
-    "exclusive_node_step": true,
+    "encoder_rnn_hidden_dim": 256,
+    "exclusive_node_step": false,
     "individual_encoder_name": "RNNContextEncoder",
     "individual_encoder_number_of_heads": 4,
     "init_hidden_num_layers": 2,
@@ -207,18 +207,18 @@
     "output_head_num_layers": 2,
     "prediction_latent_deterministic": false,
     "prediction_only": false,
-    "rnn_decoder_number_of_layers": 1,
-    "rnn_individual_encoder_number_of_layers": 1,
+    "rnn_decoder_number_of_layers": 4,
+    "rnn_individual_encoder_number_of_layers": 4,
     "study_latent_deterministic": false,
-    "time_obs_encoder_hidden_dim": 32,
-    "time_obs_encoder_output_dim": 32,
+    "time_obs_encoder_hidden_dim": 256,
+    "time_obs_encoder_output_dim": 256,
     "use_attention": true,
     "use_invariance_loss": true,
     "use_kl_i": true,
     "use_kl_i_np": true,
     "use_kl_init": true,
     "use_kl_s": true,
-    "zi_latent_dim": 20
+    "zi_latent_dim": 128
   },
   "run_index": 0,
   "tags": [
@@ -243,12 +243,12 @@
       0.9,
       0.999
     ],
-    "epochs": 50,
+    "epochs": 20,
     "eps": 1e-08,
     "eval_empirical_every_pct": 0.25,
     "gradient_clip_val": 1.0,
     "learning_rate": 0.00023,
-    "log_image_every_epoch": 2,
+    "log_image_every_epoch": 25,
     "log_interval": 1,
     "log_vcp": false,
     "num_batch_plot": 1,
@@ -262,7 +262,7 @@
       "last_epoch": -1
     },
     "shuffle_val": true,
-    "val_empirical_every_pct_epochs": 0.5,
+    "val_empirical_every_pct_epochs": 0.1,
     "weight_decay": 0.0001
   },
   "transformers_version": "4.52.4",

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:787f42a203b6312d5a4a9ac9f5ec93a44ffbadf319314eba9c258dfc476ff220
-size 87185
+oid sha256:a9633a4640c2332fcb691aa6339ce4f00fc6df11a1ccf4e03ef931b765abc12c
+size 20436185