treeflow.model.phylo_model module

treeflow.model.phylo_model.parse_model(model: Dict[str, object] | str) → Tuple[str, Dict[str, object | Dict[str, object]] | None]

exception treeflow.model.phylo_model.PhyloModelParseError

Bases: ValueError

add_note()

Exception.add_note(note) – add a note to the exception

args

with_traceback()

Exception.with_traceback(tb) – set self.__traceback__ to tb and return self.

class treeflow.model.phylo_model.PhyloModel(model_dict: Dict[str, str | Dict[str, object]])

Bases: object

Class to represent the configuration of a basic phylogenetic model

classmethod check_model_dict(model_dict: Dict[str, str | Dict[str, object]])

all_params() → Dict[str, object]

free_params() → Dict[str, Dict[str, object]]

relaxed_clock() → bool

treeflow.model.phylo_model.is_scalar_or_vector_of_type(value: object, numeric_type: Type)

treeflow.model.phylo_model.constant(value: object) → Tensor | object

treeflow.model.phylo_model.get_prior(var_name: str, dist_name: str, params: Dict[str, Tensor | object]) → Distribution

treeflow.model.phylo_model.get_params(params: Dict[str, object | Dict[str, object]] | None) → Generator[Root, Tensor, Tuple[Dict[str, Tensor | object], bool]]

Get parameters for part of the model. Builds prior distributions or converts literals to Tensor constants.

Returns:

• dict – Dictionary mapping parameter names to values

• has_root – Whether one of the parameters is a root node

treeflow.model.phylo_model.wrap_in_root_if_needed(dist: Distribution, has_root_param: bool) → Distribution | Root

treeflow.model.phylo_model.get_tree_model(tree_model: str, tree_model_params: Dict[str, object], has_root_param: bool, initial_tree: TensorflowRootedTree, var_name: str = 'tree') → Generator[Distribution | Root, TensorflowRootedTree, TensorflowRootedTree]

treeflow.model.phylo_model.get_subst_model(subst_model: str) → EigendecompositionSubstitutionModel

treeflow.model.phylo_model.get_subst_model_params(subst_model: str, params: Dict[str, object | Dict[str, object]] | None, float_dtype: DType = tf.float64, include_jc_frequencies: bool = True) → Generator[Distribution, Tensor, Tuple[Dict[str, Tensor | object], bool]]

treeflow.model.phylo_model.get_strict_clock_rates(clock_rate: Tensor)

treeflow.model.phylo_model.get_relaxed_lognormal_clock_rate_distribution(branch_rate_loc: Tensor, branch_rate_scale: Tensor, has_root_param: bool, initial_tree: TensorflowRootedTree) → Distribution

treeflow.model.phylo_model.get_clock_model_rates(clock_model: str, clock_model_params: Dict[str, object], has_root_param: bool, initial_tree: TensorflowRootedTree) → Generator[Distribution, Tensor, Tensor]

treeflow.model.phylo_model.get_discrete_gamma_site_rate_distribution(category_count: Tensor, site_gamma_shape: Tensor) → DiscretizedDistribution

treeflow.model.phylo_model.get_discrete_weibull_site_rate_distribution(category_count: Tensor, site_weibull_concentration: Tensor, site_weibull_scale: Tensor | None = None) → DiscretizedDistribution

treeflow.model.phylo_model.get_site_rate_distribution(site_model, site_model_params) → FiniteDiscreteDistribution

treeflow.model.phylo_model.get_sequence_distribution(alignment: Alignment, tree: TensorflowRootedTree, subst_model: EigendecompositionSubstitutionModel, subst_model_params: Dict[str, Tensor], site_model: str, site_model_params: Dict[str, object], clock_model_rates: Tensor, pattern_counts: Tensor | None = None, use_native: str | bool = 'auto', unroll: str | bool = 'auto') → Distribution

treeflow.model.phylo_model.phylo_model_to_joint_distribution(model: PhyloModel, initial_tree: TensorflowRootedTree, initial_alignment: Alignment, pattern_counts: Tensor | None = None, use_native: str | bool = 'auto', unroll: str | bool = 'auto', include_likelihood: bool = True) → JointDistributionCoroutine

Build the joint distribution for a phylogenetic model.

Parameters:

• use_native – Forwarded to the sequence likelihood (LeafCTMC) to select the native C++ tree-likelihood op (True), the pure-TensorFlow implementation (False), or automatic selection ("auto").

• include_likelihood – When False the sequence (alignment) likelihood term is omitted, so the resulting distribution is the prior over all latent parameters. This is useful for profiling the prior and likelihood contributions to the target density separately.