project:
  slug: ssb_rtx6000x2
  name: SSB RTX6000x2
  owner: null
experiment:
  name: ssb_npz_rtx6000x2_run
  output_dir: artifacts/ssb_rtx6000x2/runs
  tags:
    provider: local
    gpu: rtx6000x2
    dataset: ssb_npz_v1
  notes: null
data:
  dataset_name: npz_classification
  dataset_version: ssb_npz_v1
  source_uri: data/processed/ssb_npz_v1
  batch_size: 305
  num_workers: 8
  shuffle: true
  pin_memory: true
  split:
    train: 0.8
    validation: 0.1
    test: 0.1
  preprocessing:
    class_weighted_loss: true
  augmentations: []
  cache_dir: data/cache
model:
  name: mlp_classifier
  parameters:
    input_dim: 144
    hidden_dims:
    - 512
    - 256
    - 128
    num_classes: 3
    dropout: 0.226947161178787
  args: null
  pretrained: false
  checkpoint_path: null
training:
  epochs: 45
  max_steps: null
  log_every_n_steps: 25
  val_every_n_steps: null
  gradient_clip_norm: 1.0
  optimizer:
    name: adamw
    lr: 0.0002547762821605831
    weight_decay: 0.010098717791485455
    eps: 1.0e-08
    betas: null
    args: {}
  scheduler:
    name: cosine
    args:
      t_max: 43
      early_stop_patience: 0
  distributed:
    backend: nccl
    world_size: 2
    num_nodes: 1
    node_rank: 0
    master_addr: localhost
    master_port: 29500
    gradient_accumulation_steps: 4
    mixed_precision: true
    seed: 2026
  checkpoint:
    dir: artifacts/ssb_rtx6000x2/checkpoints
    monitor: val_loss
    mode: min
    save_top_k: 2
    save_every_n_epochs: 2
    save_last: true
    resume_from: null
  logging:
    level: INFO
    log_dir: logs/ssb_rtx6000x2
    tensorboard: true
    mlflow_tracking_uri: null
    mlflow_experiment: null
    flush_steps: 50
  timeout_seconds: null
evaluation:
  metrics:
  - accuracy
  - f1
  - precision
  - recall
  batch_size: 1024
  save_predictions: false
  output_dir: artifacts/ssb_rtx6000x2/eval
  generate_confusion_matrix: true
  generate_calibration_plot: false
  calibration_bins: 10