omicverse.fm.ModelSpec

class omicverse.fm.ModelSpec(name, version, skill_ready=SkillReadyStatus.REFERENCE, tasks=<factory>, modalities=<factory>, species=<factory>, gene_id_scheme=GeneIDScheme.SYMBOL, requires_finetuning=False, zero_shot_embedding=True, zero_shot_annotation=False, output_keys=<factory>, embedding_dim=512, hardware=<factory>, differentiator='', prefer_when='', checkpoint_url='', documentation_url='', paper_url='', license_notes='')[source]

Complete specification for a foundation model.

Parameters:

  • name (str) – Lowercase model identifier (e.g. "scgpt").

  • version (str) – Model version string.

  • skill_ready (SkillReadyStatus) – Adapter readiness level.

  • tasks (list[TaskType]) – Supported tasks.

  • modalities (list[Modality]) – Supported data modalities.

  • species (list[str]) – Supported species (lowercase).

  • gene_id_scheme (GeneIDScheme) – Expected gene identifier format.

  • requires_finetuning (bool) – Whether annotation/some tasks need fine-tuning.

  • zero_shot_embedding (bool) – Whether zero-shot embedding is supported.

  • zero_shot_annotation (bool) – Whether zero-shot annotation is supported.

  • output_keys (OutputKeys) – Standard AnnData keys written by this model.

  • embedding_dim (int) – Dimension of cell embeddings.

  • hardware (HardwareRequirements) – GPU/CPU requirements.

  • differentiator (str) – Unique feature that distinguishes this model.

  • prefer_when (str) – When to specifically choose this model.

  • checkpoint_url (str) – URL for downloading model weights.

  • documentation_url (str) – URL for model documentation.

  • paper_url (str) – URL for the model paper.

  • license_notes (str) – License information.

__init__(name, version, skill_ready=SkillReadyStatus.REFERENCE, tasks=<factory>, modalities=<factory>, species=<factory>, gene_id_scheme=GeneIDScheme.SYMBOL, requires_finetuning=False, zero_shot_embedding=True, zero_shot_annotation=False, output_keys=<factory>, embedding_dim=512, hardware=<factory>, differentiator='', prefer_when='', checkpoint_url='', documentation_url='', paper_url='', license_notes='')

Methods

__init__(name, version[, skill_ready, ...])

supports_modality(modality)

Check if model supports a given modality.

supports_species(species)

Check if model supports a given species.

supports_task(task)

Check if model supports a given task.

to_dict()

Convert to dictionary for serialization.

Attributes

checkpoint_url

differentiator

documentation_url

embedding_dim

gene_id_scheme

license_notes

paper_url

prefer_when

requires_finetuning

skill_ready

zero_shot_annotation

zero_shot_embedding

name

version

tasks

modalities

species

output_keys

hardware