rc-foundry 0.1.4__py3-none-any.whl → 0.1.6__py3-none-any.whl

rfd3/configs/datasets/conditions/dna_condition.yaml

@@ -0,0 +1,3 @@
+_target_: rfd3.transforms.training_conditions.SubtypeCondition
+frequency: 1.0
+subtype: ["is_dna", "is_rna"]

rfd3/configs/datasets/conditions/island.yaml

@@ -0,0 +1,28 @@
+
+_target_: rfd3.transforms.training_conditions.IslandCondition
+frequency: 1.0
+name: island
+
+# Island sampling (`is_motif_token` assignment)
+island_sampling_kwargs:
+  island_len_min: 1
+  island_len_max: 12 # Rec 25, kept lower because unindexed motifs get sampled too and create more tokens.
+  n_islands_min: 2
+  n_islands_max: 5
+
+# Subgraph / within-token sampling (`is_motif_atom` assignment)
+p_diffuse_motif_sidechains: 0.80  # 80% probability of diffusing sidechains
+p_diffuse_subgraph_atoms: 0.0      # 0% probability of sampling subgraph atoms (defaults to fully fixed)
+subgraph_sampling_kwargs:  # see tipatom
+  residue_p_seed_furthest_from_o: null
+  residue_n_bond_expectation: null
+  residue_p_fix_all: null
+  hetatom_n_bond_expectation: null
+  hetatom_p_fix_all: null
+
+# Sets `is_motif_atom_with_fixed_seq`
+p_fix_motif_sequence: 0.2       # probability that sequence is fixed for all motifs during training
+# Sets `is_motif_atom_with_fixed_coord`
+p_fix_motif_coordinates: 0.8    # Of the atoms that are sampled, should their coordinates be fixed?
+# Sets `is_motif_atom_with_unindexed`
+p_unindex_motif_tokens: 0.5     # probability of unindexing all motif atoms

rfd3/configs/datasets/conditions/ppi.yaml

@@ -0,0 +1,2 @@
+_target_: rfd3.transforms.training_conditions.PPICondition
+frequency: 1.0

rfd3/configs/datasets/conditions/sequence_design.yaml

@@ -0,0 +1,17 @@
+defaults:
+  - island
+  - _self_
+
+frequency: 1.0
+name: sequence_design
+
+island_sampling_kwargs:
+  island_len_min: 99999
+  island_len_max: 999999999
+
+p_diffuse_motif_sidechains: 1.0
+p_unindex_motif_tokens: 0.0
+
+# For ChemNet-style sampling  < 1.0
+p_fix_motif_coordinates: 0.8
+p_fix_motif_sequence: 0.1

rfd3/configs/datasets/conditions/tipatom.yaml

@@ -0,0 +1,28 @@
+
+defaults:
+  - island
+  - _self_
+
+frequency: 1.0
+name: tipatom
+
+# Island sampling (`is_motif_token` assignment)
+island_sampling_kwargs:
+  island_len_min: 1
+  island_len_max: 1
+  n_islands_min: 2
+  n_islands_max: 12
+
+# Subgraph / within-token sampling (`is_motif_atom` assignment)
+p_diffuse_motif_sidechains: 0.0  # 80% probability of diffusing sidechains
+p_diffuse_subgraph_atoms: 1.0
+subgraph_sampling_kwargs:
+  residue_p_seed_furthest_from_o: 0.8
+  residue_n_bond_expectation: 3.0
+  residue_p_fix_all: 0.05
+  hetatom_n_bond_expectation: 8
+  hetatom_p_fix_all: 0.5
+
+p_fix_motif_sequence: 0.7
+p_fix_motif_coordinates: 1.0
+p_unindex_motif_tokens: 0.5

rfd3/configs/datasets/conditions/unconditional.yaml

@@ -0,0 +1,21 @@
+# Unconditional that fixes non-protein targets
+
+defaults:
+  - island
+  - _self_
+
+frequency: 1.0
+name: unconditional
+
+island_sampling_kwargs:
+  island_len_min: 0
+  island_len_max: 0
+  n_islands_min: 0
+  n_islands_max: 0
+
+# Conditional assignments won't matter for protein regions since always diffused:
+p_diffuse_motif_sidechains: 0.0
+p_diffuse_subgraph_atoms: 0.0
+p_fix_motif_sequence: 0.0
+p_fix_motif_coordinates: 0.0
+p_unindex_motif_tokens: 0.0

rfd3/configs/datasets/design_base.yaml

@@ -0,0 +1,97 @@
+# base training dataset for training AF3 design models (atom14 variants):
+# protein subsampling only.
+
+defaults:
+  # Grab datasets
+  - train/pdb/rfd3_train_interface@train.pdb.sub_datasets.interface
+  - train/pdb/rfd3_train_pn_unit@train.pdb.sub_datasets.pn_unit
+  #- train/rfd3_monomer_distillation@train
+  
+  # Customized validation datasets
+  - val/unconditional@val.unconditional
+  - val/unconditional_deep@val.unconditional_deep
+  - val/indexed@val.indexed
+
+  # Customized train masks
+  - conditions/unconditional@global_transform_args.train_conditions.unconditional
+  - conditions/island@global_transform_args.train_conditions.island
+  - conditions/tipatom@global_transform_args.train_conditions.tipatom
+  - conditions/sequence_design@global_transform_args.train_conditions.sequence_design
+  - conditions/ppi@global_transform_args.train_conditions.ppi
+
+  - _self_
+
+# Create a dictionary used for transform arguments
+pipeline_target: rfd3.transforms.pipelines.build_atom14_base_pipeline
+
+# Base config overrides:
+diffusion_batch_size_train: 32
+diffusion_batch_size_inference: 8
+crop_size: 384
+n_recycles_train: 2
+n_recycles_validation: 1
+max_atoms_in_crop: 3840  # ~10x crop size.
+
+# Global transform arguments are necessary for arguments shared between training and inference
+global_transform_args:
+  n_atoms_per_token: 14
+  central_atom: CB
+  sigma_perturb: 2.0
+  sigma_perturb_com: 1.0
+  association_scheme: dense
+  center_option: diffuse  # options are ["all", "motif", "diffuse"]
+
+  # Reference conformer policy
+  generate_conformers: True
+  generate_conformers_for_non_protein_only: True
+  provide_reference_conformer_when_unmasked: True
+  ground_truth_conformer_policy: IGNORE  # Other options: REPLACE, ADD, FALLBACK. See atomworks.enums for details
+  provide_elements_for_unindexed_components: True
+  use_element_for_atom_names_of_atomized_tokens: True  # TODO: correct name, implies unindexed do too
+
+  # PPI Cropping
+  keep_full_binder_in_spatial_crop: False
+  max_binder_length: 170
+
+  # PPI Hotspots
+  max_ppi_hotspots_frac_to_provide: 0.2
+  ppi_hotspot_max_distance: 4.5
+
+  # Secondary structure features
+  max_ss_frac_to_provide: 0.4
+  min_ss_island_len: 1
+  max_ss_island_len: 10
+
+  train_conditions:
+    unconditional:
+      frequency: 5.0
+    sequence_design:
+      frequency: 2.0
+    island:
+      frequency: 1.0
+    tipatom:
+      frequency: 0.0
+    ppi:
+      frequency: 0.0
+
+  # Used to create simple boolean flags for downstream conditioning
+  meta_conditioning_probabilities:
+    calculate_hbonds: 0.2
+    calculate_rasa: 0.6
+
+    keep_protein_motif_rasa: 0.1  # Small to prevent noisy input to model
+    hbond_subsample: 0.5
+    
+    # fully indexed training
+    unindex_leak_global_index: 0.10
+    unindex_insert_random_break: 0.10
+    unindex_remove_random_break: 0.10
+
+    # Probability of adding 1d secondary structure conditioning
+    add_1d_ss_features: 0.1
+    featurize_plddt: 0.9  # Applied for monomer distillation only
+    add_global_is_non_loopy_feature: 0.99
+
+    # PPI
+    add_ppi_hotspots: 0.75
+    full_binder_crop: 0.75

rfd3/configs/datasets/train/pdb/af3_train_interface.yaml

@@ -0,0 +1,46 @@
+defaults:
+  - base
+
+dataset:
+  dataset_parser:
+    _target_: atomworks.ml.datasets.parsers.InterfacesDFParser
+    base_dir: ${paths.data.pdb_data_dir}
+  dataset:
+    name: interface
+    data: ${paths.data.pdb_parquet_dir}/interfaces_df_train.parquet
+    filters:
+      # filters common across all PDB datasets
+      - "deposition_date < '2021-09-30'"
+      - "resolution < 9.0"
+      - "num_polymer_pn_units <= 300" 
+      - "cluster.notnull()"
+      # interface specific filters
+      - "~(pn_unit_1_non_polymer_res_names.notnull() and pn_unit_1_non_polymer_res_names.str.contains('${resolve_import:atomworks.constants,AF3_EXCLUDED_LIGANDS_REGEX}', regex=True))"
+      - "~(pn_unit_2_non_polymer_res_names.notnull() and pn_unit_2_non_polymer_res_names.str.contains('${resolve_import:atomworks.constants,AF3_EXCLUDED_LIGANDS_REGEX}', regex=True))"
+      - "is_inter_molecule"
+    columns_to_load:
+      # columns common across all PDB datasets
+      - example_id
+      - pdb_id
+      - assembly_id
+      - deposition_date
+      - resolution
+      - num_polymer_pn_units
+      - method
+      - cluster
+      - n_prot
+      - n_nuc
+      - n_ligand
+      - n_peptide
+      # interface specific columns
+      - pn_unit_1_iid
+      - pn_unit_2_iid
+      - pn_unit_1_non_polymer_res_names
+      - pn_unit_2_non_polymer_res_names
+      - is_inter_molecule
+      - all_pn_unit_iids_after_processing
+      - involves_loi
+  transform:
+    # interface-specific Transform pipeline parameters
+    crop_contiguous_probability: 0.0
+    crop_spatial_probability: 1.0

rfd3/configs/datasets/train/pdb/af3_train_pn_unit.yaml

@@ -0,0 +1,42 @@
+defaults:
+  - base
+
+dataset: 
+  dataset_parser:
+    _target_: atomworks.ml.datasets.parsers.PNUnitsDFParser
+    base_dir: ${paths.data.pdb_data_dir}
+  dataset:
+    name: pn_unit
+    data: ${paths.data.pdb_parquet_dir}/pn_units_df_train.parquet
+    filters:
+        # filters common across all PDB datasets
+      - "deposition_date < '2021-09-30'"
+      - "resolution < 9.0"
+      - "num_polymer_pn_units <= 300" 
+      - "cluster.notnull()"
+      # pn_unit specific filters
+      - "~(q_pn_unit_non_polymer_res_names.notnull() and q_pn_unit_non_polymer_res_names.str.contains('${resolve_import:atomworks.constants,AF3_EXCLUDED_LIGANDS_REGEX}', regex=True))"
+    columns_to_load:
+      # columns common across all PDB datasets
+      - example_id
+      - pdb_id
+      - assembly_id
+      - deposition_date
+      - resolution
+      - num_polymer_pn_units
+      - method
+      - cluster
+      - n_prot
+      - n_nuc
+      - n_ligand
+      - n_peptide
+      - total_num_atoms_in_unprocessed_assembly
+      # pn_unit specific columns
+      - q_pn_unit_iid
+      - q_pn_unit_non_polymer_res_names
+      - all_pn_unit_iids_after_processing
+      - q_pn_unit_is_loi
+  transform:
+    # pn_unit-specific Transform pipeline parameters
+    crop_contiguous_probability: 0.3333333333333333
+    crop_spatial_probability: 0.6666666666666667

rfd3/configs/datasets/train/pdb/base.yaml

@@ -0,0 +1,14 @@
+# Adds weights to the sampler
+
+defaults:
+  - base_no_weights
+  - _self_
+
+weights:
+  _target_: atomworks.ml.samplers.calculate_weights_for_pdb_dataset_df
+  beta: 0.5
+  alphas:
+    a_prot: 3.0 # 3 for AF-3
+    a_nuc: 0.0 # 3 for AF-3
+    a_ligand: 1.0 # 1 for AF-3
+    a_loi: 5.0 # 5 for AF-3

rfd3/configs/datasets/train/pdb/base_no_weights.yaml

@@ -0,0 +1,19 @@
+defaults:
+  - base_transform_args
+  - _self_
+
+dataset:
+  _target_: atomworks.ml.datasets.StructuralDatasetWrapper
+  save_failed_examples_to_dir: ${paths.data.failed_examples_dir}
+  cif_parser_args:
+    cache_dir: null
+    load_from_cache: false
+    save_to_cache: false
+  dataset:
+    _target_: atomworks.ml.datasets.PandasDataset
+    # we will use the example_id as the unique column
+    id_column: example_id
+  transform:
+    # common Transform pipeline components for all PDB datasets
+    _target_: ${datasets.pipeline_target}
+    is_inference: False

rfd3/configs/datasets/train/pdb/base_transform_args.yaml

@@ -0,0 +1,59 @@
+# All required training args
+
+defaults:
+  - _self_
+
+dataset:
+  transform:
+    _target_: ${datasets.pipeline_target}
+    is_inference: False
+    return_atom_array: False
+
+    # Model
+    sigma_perturb: ${datasets.global_transform_args.sigma_perturb}
+    sigma_perturb_com: ${datasets.global_transform_args.sigma_perturb_com}
+    sigma_data: ${model.net.diffusion_module.sigma_data}
+    diffusion_batch_size: ${datasets.diffusion_batch_size_train}
+    central_atom: ${datasets.global_transform_args.central_atom}
+    n_atoms_per_token: ${datasets.global_transform_args.n_atoms_per_token}
+    association_scheme: ${datasets.global_transform_args.association_scheme}
+    center_option: ${datasets.global_transform_args.center_option}
+
+    # Conformers
+    generate_conformers: ${datasets.global_transform_args.generate_conformers}
+    generate_conformers_for_non_protein_only: ${datasets.global_transform_args.generate_conformers_for_non_protein_only}
+    provide_reference_conformer_when_unmasked: ${datasets.global_transform_args.provide_reference_conformer_when_unmasked}
+    ground_truth_conformer_policy: ${datasets.global_transform_args.ground_truth_conformer_policy}
+    provide_elements_for_unindexed_components: ${datasets.global_transform_args.provide_elements_for_unindexed_components}
+    use_element_for_atom_names_of_atomized_tokens: ${datasets.global_transform_args.use_element_for_atom_names_of_atomized_tokens}
+    residue_cache_dir: ${paths.data.residue_cache_dir}
+
+    # Conditions
+    train_conditions: ${datasets.global_transform_args.train_conditions}
+    meta_conditioning_probabilities: ${datasets.global_transform_args.meta_conditioning_probabilities}
+
+    # PPI Hypers
+    keep_full_binder_in_spatial_crop: ${datasets.global_transform_args.keep_full_binder_in_spatial_crop}
+    max_binder_length: ${datasets.global_transform_args.max_binder_length}
+    max_ppi_hotspots_frac_to_provide: ${datasets.global_transform_args.max_ppi_hotspots_frac_to_provide}
+    ppi_hotspot_max_distance: ${datasets.global_transform_args.ppi_hotspot_max_distance}
+
+    # 1D SS hypers
+    max_ss_frac_to_provide: ${datasets.global_transform_args.max_ss_frac_to_provide}
+    min_ss_island_len: ${datasets.global_transform_args.min_ss_island_len}
+    max_ss_island_len: ${datasets.global_transform_args.max_ss_island_len}
+
+    # Cropping
+    crop_size: ${datasets.crop_size}
+    max_atoms_in_crop: ${datasets.max_atoms_in_crop}
+    allowed_types: ALL
+    crop_spatial_probability: ???
+    crop_contiguous_probability: ???
+    dna_contact_crop_probability: 0.0
+    crop_center_cutoff_distance: 15.0
+    zero_occ_on_exposure_after_cropping: False
+    b_factor_min: null
+
+    # Other dataset-specific parameters
+    atom_1d_features: ${model.net.token_initializer.atom_1d_features}
+    token_1d_features: ${model.net.token_initializer.token_1d_features}

rfd3/configs/datasets/train/pdb/na_complex_distillation.yaml

@@ -0,0 +1,20 @@
+defaults:
+  - base_no_weights
+  - _self_
+
+dataset:
+  dataset_parser:
+    _target_: atomworks.ml.datasets.parsers.GenericDFParser
+    pn_unit_iid_colnames: null
+
+  dataset:
+    name: tf_distillation
+    data: /projects/ml/prot_dna/transcriptionFactor_distillation_rf3.newDL.csv
+    columns_to_load:
+      - example_id
+      - path
+    
+  transform:
+    crop_contiguous_probability: 0.4
+    crop_spatial_probability: 0.0
+    dna_contact_crop_probability: 0.6

rfd3/configs/datasets/train/pdb/pdb_base.yaml

@@ -0,0 +1,11 @@
+# Base config for all PDB datasets
+defaults:
+  - base
+  - _self_
+
+dataset:
+  # All PDB datasets load from this cache:
+  cif_parser_args:
+    cache_dir: ${paths.data.cif_cache_dir}
+    load_from_cache: True
+    save_to_cache: False

rfd3/configs/datasets/train/pdb/rfd3_train_interface.yaml

@@ -0,0 +1,22 @@
+# Inherit 
+defaults:
+  - af3_train_interface
+  - pdb_base
+  - _self_
+
+dataset:
+  transform:
+    crop_contiguous_probability: 0.0
+    crop_spatial_probability: 1.0
+    filters:
+        # filters common across all PDB datasets
+        - 'pdb_id not in ["7rte", "7m5w", "7n5u"]'
+        - 'pdb_id not in ["3di3", "5o45", "1z92", "2gy5", "4zxb"]'
+        - "deposition_date < '2024-12-16'"
+        - "resolution < 9.0"
+        - "num_polymer_pn_units <= 300" 
+        - "cluster.notnull()"
+        # interface specific filters
+        - "~(pn_unit_1_non_polymer_res_names.notnull() and pn_unit_1_non_polymer_res_names.str.contains('${resolve_import:atomworks.constants,AF3_EXCLUDED_LIGANDS_REGEX}', regex=True))"
+        - "~(pn_unit_2_non_polymer_res_names.notnull() and pn_unit_2_non_polymer_res_names.str.contains('${resolve_import:atomworks.constants,AF3_EXCLUDED_LIGANDS_REGEX}', regex=True))"
+        - "is_inter_molecule"

rfd3/configs/datasets/train/pdb/rfd3_train_pn_unit.yaml

@@ -0,0 +1,23 @@
+defaults:
+  - af3_train_pn_unit
+  - pdb_base
+  - _self_
+
+dataset: 
+  transform:
+    # pn_unit-specific Transform pipeline parameters
+    crop_contiguous_probability: 0.25
+    crop_spatial_probability: 0.75
+
+  # Modify: date & clustering parquet
+  dataset:
+    filters:
+        # filters common across all PDB datasets
+      - 'pdb_id not in ["7rte", "7m5w", "7n5u"]'
+      - 'pdb_id not in ["3di3", "5o45", "1z92", "2gy5", "4zxb"]'
+      - "deposition_date < '2024-12-16'"
+      - "resolution < 9.0"
+      - "num_polymer_pn_units <= 300" 
+      - "cluster.notnull()"
+      # pn_unit specific filters
+      - "~(q_pn_unit_non_polymer_res_names.notnull() and q_pn_unit_non_polymer_res_names.str.contains('${resolve_import:atomworks.constants,AF3_EXCLUDED_LIGANDS_REGEX}', regex=True))"

rfd3/configs/datasets/train/rfd3_monomer_distillation.yaml

@@ -0,0 +1,38 @@
+defaults:
+  - pdb/base_transform_args@monomer_distillation
+  - _self_
+
+monomer_distillation:
+  dataset:
+    _target_: atomworks.ml.datasets.StructuralDatasetWrapper
+    save_failed_examples_to_dir: ${paths.data.failed_examples_dir}
+    # Explicitly do not load from cache.
+    # Dataset too big, and structures are small
+    cif_parser_args:
+      cache_dir: null
+      load_from_cache: False
+      save_to_cache: False
+
+    # metadata dataset
+    dataset:
+      _target_: atomworks.ml.datasets.PandasDataset
+      name: af2fb_distillation
+      id_column: example_id
+      data: ${paths.data.monomer_distillation_parquet_dir}/af2_distillation_facebook.parquet
+      columns_to_load:
+        - example_id
+        - path
+      
+    # metadata parser
+    dataset_parser:
+      _target_: atomworks.ml.datasets.parsers.GenericDFParser
+      pn_unit_iid_colnames: null
+  
+    transform:
+      _target_: ${datasets.pipeline_target}
+      is_inference: False
+      # protein_msa_dirs: [{"dir": "${paths.data.monomer_distillation_data_dir}/msa", "extension": ".a3m", "directory_depth": 2}]
+      # rna_msa_dirs: []
+      crop_contiguous_probability: 0.25
+      crop_spatial_probability: 0.75
+      b_factor_min: 70

rfd3/configs/datasets/val/bcov_ppi_easy_medium.yaml

@@ -0,0 +1,9 @@
+
+defaults:
+  - design_validation_base
+  - val_examples/bcov_ppi_easy_medium_with_ori@dataset.data
+  - _self_
+
+dataset:
+  eval_every_n: 1
+  name: bcov-ppi-easy-medium

rfd3/configs/datasets/val/design_validation_base.yaml

@@ -0,0 +1,40 @@
+dataset:
+  _target_: rfd3.inference.datasets.ContigJsonDataset
+
+  # Required parameters for each inheriting dataset
+  data: ???  # Path to json file
+  name: ???  # Name for displaying and saving files
+  eval_every_n: ???  # Evaluate on this dataset every n epochs
+  subset_to_keys: null  # Specific keys in json to keep, ignores all others.
+
+  # NB: Used for parsing input files (not for atom_array reloading anymore)
+  cif_parser_args:
+    cache_dir: null
+    load_from_cache: False
+    save_to_cache: False
+    add_missing_atoms: False
+    
+  # Common Transform pipeline components for all PDB datasets
+  transform:
+    _target_: ${datasets.pipeline_target}
+    is_inference: True
+    return_atom_array: True
+    diffusion_batch_size: ${datasets.diffusion_batch_size_train}
+    sigma_data: ${model.net.diffusion_module.sigma_data}
+    central_atom: ${datasets.global_transform_args.central_atom}
+    n_atoms_per_token: ${datasets.global_transform_args.n_atoms_per_token}
+    association_scheme: ${datasets.global_transform_args.association_scheme}
+    center_option: ${datasets.global_transform_args.center_option}
+
+    # Conformers
+    generate_conformers: ${datasets.global_transform_args.generate_conformers}
+    generate_conformers_for_non_protein_only: ${datasets.global_transform_args.generate_conformers_for_non_protein_only}
+    provide_reference_conformer_when_unmasked: ${datasets.global_transform_args.provide_reference_conformer_when_unmasked}
+    ground_truth_conformer_policy: ${datasets.global_transform_args.ground_truth_conformer_policy}
+    provide_elements_for_unindexed_components: ${datasets.global_transform_args.provide_elements_for_unindexed_components}
+    use_element_for_atom_names_of_atomized_tokens: ${datasets.global_transform_args.use_element_for_atom_names_of_atomized_tokens}
+    residue_cache_dir: ${paths.data.residue_cache_dir}
+
+    # Other dataset-specific parameters
+    atom_1d_features: ${model.net.token_initializer.atom_1d_features}
+    token_1d_features: ${model.net.token_initializer.token_1d_features}

rfd3/configs/datasets/val/dna_binder_design5.yaml

@@ -0,0 +1,9 @@
+
+defaults:
+  - design_validation_base
+  - _self_
+
+dataset:
+  data: ${paths.data.design_benchmark_data_dir}/dna_binder.json
+  name: dna_binder_design
+  eval_every_n: 1

rfd3/configs/datasets/val/dna_binder_long.yaml

@@ -0,0 +1,13 @@
+
+defaults:
+  - design_validation_base
+  - _self_
+
+dataset:
+  data: ${paths.root_dir}/tests/dna.json
+  name: dna_binder_design
+  eval_every_n: 10
+  subset_to_keys:
+    - 7rte_sequence_only
+    - 7rte_with_structure
+

rfd3/configs/datasets/val/dna_binder_short.yaml

@@ -0,0 +1,13 @@
+
+defaults:
+  - design_validation_base
+  - _self_
+
+dataset:
+  data: ${paths.root_dir}/rfd3/tests/test_data/dna.json
+  name: dna_binder_design
+  eval_every_n: 1
+  subset_to_keys:
+    - 7rte_sequence_only
+    - 7rte_with_structure
+

rfd3/configs/datasets/val/indexed.yaml

@@ -0,0 +1,9 @@
+
+defaults:
+  - design_validation_base
+  - _self_
+
+dataset:
+  data: ${paths.data.design_benchmark_data_dir}/indexed.json
+  name: indexed-design
+  eval_every_n: 8

rfd3/configs/datasets/val/mcsa_41.yaml

@@ -0,0 +1,9 @@
+
+defaults:
+  - design_validation_base
+  - _self_
+
+dataset:
+  data: ${paths.data.design_benchmark_data_dir}/mcsa_41.json
+  name: woodys-benchmark
+  eval_every_n: 16

rfd3/configs/datasets/val/mcsa_41_short_rigid.yaml

@@ -0,0 +1,10 @@
+
+defaults:
+  - unindexed
+  - _self_
+
+dataset:
+  name: rigid-ligand-enzymes
+  eval_every_n: 1
+  data: ${paths.data.design_benchmark_data_dir}/mcsa_41_short_rigid_new.json
+

rfd3/configs/datasets/val/ppi_inference.yaml

@@ -0,0 +1,7 @@
+defaults:
+  - unconditional
+  - _self_
+
+dataset:
+  name: ppi_inference
+  data: ??? # This is a required override, specifying a path to the dataset json or yaml file.

rfd3/configs/datasets/val/sm_binder_hbonds.yaml

@@ -0,0 +1,13 @@
+defaults:
+  - design_validation_base
+  - _self_
+
+dataset:
+  data: ${paths.data.design_benchmark_data_dir}/sm_binder_hbonds.json
+  eval_every_n: 5
+  name: sm_binder_hbonds-design
+  subset_to_keys:
+    - FAD
+    - IAI
+    - OQO
+    - SAM

rfd3/configs/datasets/val/sm_binder_hbonds_short.yaml

@@ -0,0 +1,15 @@
+defaults:
+  - sm_binder_hbonds
+  - _self_
+
+dataset:
+  eval_every_n: 1
+  data: ${paths.data.design_benchmark_data_dir}/sm_binder_hbonds_sampled.json
+  name: sm_binder_hbonds-design-short
+  subset_to_keys:
+    - FAD_1
+    - FAD_2
+    - FAD_3
+    - IAI_1
+    - IAI_2
+    - IAI_3