Struct ProteinMPNNRunner 

pub struct ProteinMPNNRunner { /* private fields */ }

Implementations

impl ProteinMPNNRunner

pub fn from_pretrained( modeltype: ProteinMPNNModels, device: Device, ) -> Result<Self>

Load a ProteinMPNN model from HuggingFace hub.

pub fn load_model(modeltype: ProteinMPNNModels, device: Device) -> Result<Self>

👎Deprecated since 0.4.0:

renamed to from_pretrained

Renamed to from_pretrained (ferritin-100.8).

pub fn load_model_with( modeltype: ProteinMPNNModels, opts: &LoadOptions, ) -> Result<Self>

👎Deprecated since 0.4.0:

renamed to from_pretrained_with

Renamed to from_pretrained_with (ferritin-100.8).

pub fn from_pretrained_with( modeltype: ProteinMPNNModels, opts: &LoadOptions, ) -> Result<Self>

Load with an explicit device and dtype (ferritin-100.9).

pub fn from_path(path: impl AsRef<Path>, device: Device) -> Result<Self>

Load from a local .pt file (e.g. from ferritin-test-data or a cached download).

Assumes a ProteinMPNN checkpoint. For LigandMPNN use from_path_as — the two need different configs and a LigandMPNN checkpoint loaded as ProteinMPNN silently ignores its nine ligand tensors.

pub fn from_path_with( path: impl AsRef<Path>, opts: &LoadOptions, ) -> Result<Self>

Load from a local file with an explicit device and dtype.

pub fn from_path_as( path: impl AsRef<Path>, config: &ProteinMPNNConfig, opts: &LoadOptions, ) -> Result<Self>

Load from a local file against an explicit architecture config.

pub fn into_model(self) -> ProteinMPNN

Consume the runner and yield the loaded model.

Lets callers that need the bare ProteinMPNN — such as MPNNExecConfig::load_model — go through this tested loading path instead of duplicating the download (ferritin-100.10).

pub fn get_log_probs(&self, features: &ProteinFeatures) -> Result<Tensor>

Run ProteinMPNN and return a (L, 21) log-probability tensor for all positions.

Useful for numerical parity tests against a Python reference. Values are log-softmax of the raw logits from a single structure-conditioned forward pass (the same computation as simple_decode).

pub fn get_pseudo_probabilities( &self, features: &ProteinFeatures, ) -> Result<Vec<PseudoProbability>>

Run ProteinMPNN and return per-residue pseudo-probabilities for the 21 amino acids.