CLIPEncodeMultipleAdvanced

Documentation

• Class name: CLIPEncodeMultipleAdvanced
• Category: Bmad/conditioning
• Output node: False

This node is designed for advanced text encoding using the CLIP model, capable of processing multiple inputs simultaneously. It extends the functionality of a standard CLIP text encoding by allowing for a customizable number of inputs, each potentially undergoing a unique encoding process based on token normalization and weight interpretation. The node is optimized for scenarios requiring batch processing of text inputs for conditioning generation models, making it a versatile tool in the generation of conditioned outputs.

Input types

Required

• clip
  • The CLIP model used for encoding the text inputs. It plays a crucial role in determining the quality and relevance of the encoded vectors.
  • Comfy dtype: CLIP
  • Python dtype: torch.nn.Module
• token_normalization
  • A flag indicating whether or not to normalize the tokens during the encoding process. This affects the consistency and comparability of the encoded vectors.
  • Comfy dtype: COMBO[STRING]
  • Python dtype: bool
• weight_interpretation
  • A parameter that influences how the weights are interpreted during the encoding process, affecting the emphasis on different aspects of the input text.
  • Comfy dtype: COMBO[STRING]
  • Python dtype: str
• inputs_len
  • Specifies the number of text inputs to be encoded. This allows for dynamic adjustment of batch sizes for encoding, accommodating a variable number of inputs.
  • Comfy dtype: INT
  • Python dtype: int

Output types

• conditioning
  • Comfy dtype: CONDITIONING
  • A list of conditioning vectors generated from the input texts, tailored for use in generation models requiring conditioned inputs.
  • Python dtype: List[torch.Tensor]

Usage tips

• Infra type: GPU
• Common nodes: unknown

Source code

class CLIPEncodeMultipleAdvanced(AdvancedCLIPTextEncode):
    @classmethod
    def INPUT_TYPES(s):
        types = super().INPUT_TYPES()  # TODO should refactor Grid class above to this too, so if original is changed, all the new options are added there too
        types["required"].pop("text")
        types["required"]["inputs_len"] = ("INT", {"default": 9, "min": 0, "max": 32})
        return types

    RETURN_TYPES = ("CONDITIONING",)
    FUNCTION = "gen2"
    CATEGORY = conditioning_category_path
    OUTPUT_IS_LIST = (True,)

    def gen2(self, clip, token_normalization, weight_interpretation, inputs_len, **kwargs):
        conds = []
        for i in range(inputs_len):
            arg_name = get_arg_name_from_multiple_inputs("string", i)
            conds.append(
                super().encode(clip, kwargs[arg_name], token_normalization, weight_interpretation, 'disable')[0])
        return (conds,)