Method __init__

gemma_encoder.py:173–195

(self, device="cpu", dtype=dtype, model_options={})

def ltxv_gemma_clip(encoder_path, ltxv_path, processor=None, dtype=None):
    class _LTXVGemmaTextEncoderModel(LTXVGemmaTextEncoderModel):
        def __init__(self, device="cpu", dtype=dtype, model_options={}):
            dtype = torch.bfloat16  # TODO: make this configurable

            gemma_model = Gemma3ForConditionalGeneration.from_pretrained(
                encoder_path,
                local_files_only=True,
                torch_dtype=dtype,
            )

            feature_extractor, embeddings_processor = load_text_embeddings_pipeline(
                ltxv_path,
                dtype=dtype,
                fallback_proj_path=encoder_path / "proj_linear.safetensors",
            )

            super().__init__(
                model=gemma_model,
                feature_extractor=feature_extractor,
                embeddings_processor=embeddings_processor,
                processor=processor,
                dtype=dtype,
                device=device,
            )

    return _LTXVGemmaTextEncoderModel