IConditioningModule<T>
Interfaces · AiDotNet.Interfaces
Interface for conditioning modules that encode various inputs into embeddings for diffusion models.
Known implementations: CompositeConditioningBase<T>, DualTextConditioner<T>, TextConditioningBase<T>
For Beginners
A conditioning module is like a “translator” that converts your input (like a text prompt) into a format the diffusion model can understand.
Common types of conditioning:
- Text conditioning (CLIP, T5): “A cat sitting on a couch” → embedding vectors
- Image conditioning (IP-Adapter): An image → embedding vectors for style/content
- Control conditioning (ControlNet): Depth maps, edges, poses → spatial guidance
Why conditioning matters:
- Without conditioning: Model generates random images
- With text conditioning: Model generates images matching your description
- With image conditioning: Model preserves style or content from reference images
- With control conditioning: Model follows spatial structure (poses, edges, depth)
Different conditioning methods:
- Cross-attention: Text embeddings attend to image features (most common for text)
- Addition/Concatenation: Add or concat embeddings to time embedding
- Spatial: Add control signals directly to features at each resolution
How It Works
Conditioning modules convert various types of input (text, images, audio, etc.) into embedding tensors that guide the diffusion process. They are essential for controlled generation like text-to-image, image-to-image, or style transfer.
Properties
| Property | Summary |
|---|---|
ConditioningType ConditioningType { get; } |
Gets the type of conditioning this module provides. |
bool ProducesPooledOutput { get; } |
Gets whether this module produces pooled (global) or sequence embeddings. |
int EmbeddingDimension { get; } |
Gets the dimension of the output embeddings. |
int MaxSequenceLength { get; } |
Gets the maximum sequence length for text input. |
Methods
| Method | Summary |
|---|---|
Tensor<T> Encode(Tensor<T> input) |
Encodes the input into conditioning embeddings. |
Tensor<T> EncodeText(Tensor<T> tokenIds, Tensor<T> attentionMask) |
Encodes text input (convenience method for text conditioning). |
Tensor<T> GetPooledEmbedding(Tensor<T> sequenceEmbeddings) |
Gets the pooled (global) embedding from sequence embeddings. |
Tensor<T> GetUnconditionalEmbedding(int batchSize) |
Gets unconditioned (null) embeddings for classifier-free guidance. |
Tensor<T> Tokenize(string text) |
Tokenizes text input (for text conditioning modules). |
Tensor<T> TokenizeBatch(string[] texts) |
Tokenizes a batch of text inputs. |