Linear-ODE-Gamma/train.yaml

model:
  si:
    class_path: omg.si.stochastic_interpolants.StochasticInterpolants
    init_args:
      stochastic_interpolants:
        # chemical species
        - class_path: omg.si.single_stochastic_interpolant_identity.SingleStochasticInterpolantIdentity
        # fractional coordinates
        - class_path: omg.si.single_stochastic_interpolant.SingleStochasticInterpolant
          init_args:
            interpolant: omg.si.interpolants.PeriodicLinearInterpolant
            gamma: 
              class_path: omg.si.gamma.LatentGammaSqrt
              init_args:
                a: 0.2575112227566439
            epsilon: null
            differential_equation_type: "ODE"
            integrator_kwargs:
              method: "euler"
            velocity_annealing_factor: 7.7611189744870925
            correct_center_of_mass_motion: true
        # lattice vectors
        - class_path: omg.si.single_stochastic_interpolant.SingleStochasticInterpolant
          init_args:
            interpolant: omg.si.interpolants.TrigonometricInterpolant
            gamma: 
              class_path: omg.si.gamma.LatentGammaSqrt
              init_args:
                a: 2.9759856920732597
            epsilon: null
            differential_equation_type: "ODE"
            integrator_kwargs:
              method: "euler"
            velocity_annealing_factor: 4.116061496782678
            correct_center_of_mass_motion: false
      data_fields:
        # if the order of the data_fields changes,
        # the order of the above StochasticInterpolant inputs must also change
        - "species"
        - "pos"
        - "cell"
      integration_time_steps: 690
  relative_si_costs:
    species_loss: 0.0
    pos_loss_b: 0.9976417941296929
    cell_loss_b: 0.002358205870307133
  sampler:
    class_path: omg.sampler.IndependentSampler
    init_args:
      pos_distribution:
        class_path: omg.sampler.position_distributions.UniformPositionDistribution
      cell_distribution:
        class_path: omg.sampler.cell_distributions.InformedLatticeDistribution
        init_args:
          dataset_name: alex_mp_20
      species_distribution:
        class_path: omg.sampler.species_distributions.MirrorSpecies
  model:
    class_path: omg.model.model.Model
    init_args:
      encoder:
        class_path: omg.model.encoders.cspnet_full.CSPNetFull
      head:
        class_path: omg.model.heads.pass_through.PassThrough
      time_embedder:
        class_path: omg.model.model_utils.SinusoidalTimeEmbeddings
        init_args:
          dim: 256
  use_min_perm_dist: False
  float_32_matmul_precision: "high"
  validation_mode: "match_rate"
  number_cpus: 7
  dataset_name: "alex_mp_20"
data:
  train_dataset:
    class_path: omg.datamodule.StructureDataset
    init_args:
      file_path: "data/alex_mp_20/train.lmdb"
      lazy_storage: True
      niggli_reduce: True
  val_dataset:
    class_path: omg.datamodule.StructureDataset
    init_args:
      file_path: "data/alex_mp_20/val.lmdb"
      lazy_storage: True
      niggli_reduce: True
  pred_dataset:
    class_path: omg.datamodule.StructureDataset
    init_args:
      file_path: "data/alex_mp_20/test.lmdb"
      lazy_storage: True
      niggli_reduce: True
  batch_size: 128
  num_workers: 4
  pin_memory: True
  persistent_workers: True
trainer:
  callbacks:
    - class_path: lightning.pytorch.callbacks.ModelCheckpoint
      init_args:
        filename: "best_val_loss_total"
        save_top_k: 1
        monitor: "val_loss_total"
        save_weights_only: true
    - class_path: lightning.pytorch.callbacks.ModelCheckpoint
      init_args:
        filename: "best_val_match_rate"
        save_top_k: 1
        monitor: "match_rate"
        save_weights_only: true
        mode: 'max'
    - class_path: lightning.pytorch.callbacks.ModelCheckpoint
      init_args:
        filename: "best_val_rmsd"
        save_top_k: 1
        monitor: "mean_rmsd"
        save_weights_only: true
    - class_path: lightning.pytorch.callbacks.ModelCheckpoint
      init_args:
        save_top_k: -1  # Store every checkpoint after 100 epochs.
        monitor: "val_loss_total"
        every_n_epochs: 100
        save_weights_only: false
  gradient_clip_val: 0.5
  num_sanity_val_steps: 0
  precision: "32-true"
  max_epochs: 2000
  enable_progress_bar: true
  limit_val_batches: 0.1
  check_val_every_n_epoch: 100
optimizer:
  class_path: torch.optim.Adam
  init_args:
    lr: 4.006249666984122e-05