config.json

{
  "model_type": "isrm-vae",
  "architecture": "distilbert-vae",
  "architecture_type": "dual-layer",
  "base_model": "distilbert-base-uncased",
  "latent_dim": 3,
  "hidden_dim": 768,
  "num_layers": 6,
  "num_unfrozen_layers": 2,
  "pad_dimensions": [
    "pleasure",
    "arousal",
    "dominance"
  ],
  "bdi_dimensions": [
    "belief",
    "goal",
    "intention",
    "ambiguity",
    "social"
  ],
  "pad_layer": 10,
  "bdi_layer": 19,
  "task": "affective-state-encoding",
  "training": {
    "dataset_size": 1500,
    "epochs": 15,
    "optimizer": "AdamW",
    "learning_rate": 2e-05,
    "loss_fn": "MSE + KL divergence",
    "kl_weight": 0.001,
    "final_mse": 0.018,
    "final_kld": 0.003
  }
}