neurovlm.core.NeuroVLM

neurovlm.core.NeuroVLM#

class neurovlm.core.NeuroVLM(datasets=None, device='cpu')[source]#

Unified interface for text-to-brain and brain-to-text.

Parameters:

  • datasets (sequence of str, optional) – Text corpora to include for to_text.

  • device (str, optional) – Torch device for inference.

__init__(datasets=None, device='cpu')[source]#
Parameters:

  • datasets (Sequence[str] | None)

  • device (str)

Return type:

None

Methods

__init__([datasets, device])

brain(X)

Start a brain-driven chain.

decode_brain(latent)

Decode one latent brain vector into NIfTI.

decode_brains(latents)

Decode multiple latent brain vectors into NIfTI images.

generate_brain(X[, adapter])

Alias for text-to-brain generation via to_brain(head='mse').

generate_llm_response(backend, model_name[, ...])

Generate an LLM summary using the last retrieval result as context.

generate_text(X, *[, basis, prefix_text, ...])

Generate text for one or more brain maps with NeuroQFormer + NeuroQwen3.

get_niftis([index])

Return NIfTI image(s) from the latest generated (MSE) brain maps.

plot(image[, threshold, display_mode, ...])

Plot a provided NIfTI image.

retrieve_brain(X[, dataset])

Alias for contrastive brain retrieval via to_brain(head='infonce').

retrieve_text(X[, datasets, project])

Alias for contrastive text retrieval.

text(X)

Start a text-driven chain.

to_brain(X[, head, dataset, adapter])

Generate or retrieve brain results for one or many queries.

to_text(X[, head, datasets, project])

Generate or retrieve text results for one or many queries.

top_k([k, target, query_index, dataset])

Return top-k rows from the latest retrieval.

Attributes

results

Return the latest retrieval/generation result object.