Model Module (PsychiatryNLPKit.model)

Wrappers for HuggingFace language models and Vision Transformers with automatic device detection (CUDA / MPS / XPU / CPU) and lazy loading.

Attach model instances to a TextData object to enable analyses that require them: embedding models for similarity/density, generative models for perplexity, etc.

Language Model Wrappers

PsychiatryNLPKit.model.BaseLLM(name[, key])	Abstract base class for all model wrappers.
PsychiatryNLPKit.model.HuggingFaceLLM(name)	Concrete HuggingFace model wrapper with lazy loading.
PsychiatryNLPKit.model.HFEmbeddingLLM(name)	Marker for embedding models (sentence/word encoders).
PsychiatryNLPKit.model.HFGenerativeLLM(name)	Marker for autoregressive generative models.
PsychiatryNLPKit.model.HFMaskFillingModel(name)	Marker for masked language models.

Vision Model Wrappers

PsychiatryNLPKit.model.HuggingFaceViTModel(name)	Wrapper for SigLIP/CLIP-style vision-language models.
PsychiatryNLPKit.model.HFMultimodalEmbeddingModel(name)	Marker for multimodal embedding models (e.g. CLIP, SigLIP).