fnbm-current-sweep-results

Grid-sweep trial results for fnbm-current.

Dataset Info

• Rows: 23
• Columns: 29

Columns

Column	Type	Description
trial_id	Value('string')	Unique id for this trial: {sweep_name}__{combo_name}
experiment_name	Value('string')	Experiment folder slug this trial belongs to, e.g. fnbm-current
sweep_name	Value('string')	Base run name for the whole sweep (config['run']['name'] before per-combo suffixing)
combo_name	Value('string')	Per-combo run name (base name + swept param values)
grid_overrides	Value('string')	JSON-encoded {dot.path: value} of just the swept parameters for this trial
job_id	Value('string')	Cluster job id, if known
cluster	Value('string')	Cluster name, if known
timestamp	Value('string')	ISO 8601 UTC time this record was built
config_yaml	Value('string')	Full resolved config for this trial (reproducibility source of truth)
best_val_loss	Value('float64')	Best validation loss achieved during training
best_val_pearson_r	Value('float64')	Validation Pearson r at the best-val-loss epoch
best_epoch	Value('int64')	Epoch index of the best validation loss
sim_r2	Value('float64')	Final-epoch R^2 against simulation ground truth
sim_pearson_r	Value('float64')	Final-epoch Pearson r against simulation ground truth
sim_mean_gt_z_abs_ratio	Value('float64')	Ground-truth mean fraction of
sim_mean_model_z_abs_ratio	Value('float64')	Model's mean fraction of
sim_z_abs_1_corr	Value('float64')	Correlation between model and ground-truth per-sample first-order
sim_z_abs_2_corr	Value('float64')	Correlation between model and ground-truth per-sample second-order
sim_z_abs_ratio_corr	Value('float64')	Correlation between model and ground-truth per-sample z_abs_ratio
sim_z_abs_ratio_mse	Value('float64')	MSE between model and ground-truth per-sample z_abs_ratio
absorption_gap	Value('float64')	sim_mean_model_z_abs_ratio - sim_mean_gt_z_abs_ratio (signed; >0 = over-attributes to second order)
promiscuity_mean_entropy	Value('float64')	Mean per-filter normalized positional entropy (diffuse activation proxy)
promiscuity_max_entropy	Value('float64')	Max per-filter normalized positional entropy
n_true_interaction_pairs	Value('float64')	Ground truth: number of true interacting motif pairs in the simulation (from sim_config.json)
interaction_effective_num_pairs	Value('float64')	Effective number of filter pairs carrying interaction mass (entropy-based)
interaction_concentration_gap	Value('float64')	interaction_effective_num_pairs - n_true_interaction_pairs
epoch_history	Value('string')	JSON-encoded list of per-epoch metrics.csv rows for this trial
wandb_run_url	Value('string')	Live wandb run URL for this trial, if wandb was enabled
report_pdf_path	Value('string')	Cluster-local path to this trial's latest simulation_eval report.pdf, if any

Generation Parameters

{
  "script_name": "sync_sweep_results.py",
  "model": "fnbm-current",
  "description": "Grid-sweep trial results for fnbm-current.",
  "experiment_name": "fnbm-current",
  "experiment_id": "fnbm-current",
  "visualizer_type": "table",
  "artifact_type": "eval_result",
  "hyperparameters": {},
  "input_datasets": []
}

Usage

from datasets import load_dataset

dataset = load_dataset("arushram/fnbm-current-sweep-results", split="train")
print(f"Loaded {len(dataset)} rows")

---