Timestep
A timestep is a discrete position or interval in an ordered temporal or sequential process at which state, computation, or output may change.
Grokking AI Algorithms, Second Edition
A timestep is one discrete point in the sequence of stages used to progressively add or remove noise in a diffusion process.
- How it is used
- Controls how much noise a diffusion model adds or removes at each stage of image generation.
- How deep it goes
- Explains timestep progression, schedule values, reverse-loop behavior, and sinusoidal embeddings without pursuing deeper theory.
- Examples
- Uses a noisy-to-crisp tree sequence, schedule tables, timestep 22 embeddings, and denoising code that counts backward from 1,000.
- What you can do after
- Build it from scratch by generating timestep embeddings and tracing schedule-controlled denoising loops.
Read how Grokking AI Algorithms, Second Edition explains timestep →
Build a Text-to-Image Generator (from Scratch)
A time step is a discrete stage in the diffusion process at which a particular noise level is applied.
- How it is used
- Conditions noise addition, denoising direction, and U-Net predictions on a discrete position in the diffusion process.
- How deep it goes
- Explains timestep ordering, noise weighting, arbitrary-time sampling, and training use, but does not detail timestep-embedding internals.
- Examples
- Uses dog images, clothing-item batches, and flower images visualized at selected noise levels.
- What you can do after
- Apply it via a DDIMScheduler to add noise at selected timesteps and inspect the resulting diffusion stages.
Read how Build a Text-to-Image Generator (from Scratch) explains timestep →
Inside Deep Learning: Math, Algorithms, Models
A discrete position in a sequence or recurrent computation at which the model updates state.
- How it is used
- Organizes sequential inputs and recurrent state updates so models can carry information across an ordered sequence.
- How deep it goes
- Defines timestep notation, state propagation, tensor dimensions, and an explicit loop, but does not derive recurrent internals.
- Examples
- Uses ordered weather vectors, an unrolled three-timestep RNN, and (B, T, D) sequence tensors.
- What you can do after
- Apply via PyTorch by arranging inputs as (B, T, D) and iterating over T recurrent updates.
Read how Inside Deep Learning: Math, Algorithms, Models explains timestep →
Natural Language Processing in Action, Second Edition
A time step is one ordered position in a sequential computation, corresponding here to a token position in text.
- How it is used
- Places token processing and error propagation within recurrent neural networks.
- How deep it goes
- Explains shared weights and serial backpropagation across tokens, without detailing tensor shapes or implementation code.
- Examples
- Uses text tokens, an unrolled recurrent network, and token-by-token gradient propagation as concrete illustrations.
- What you can do after
- Understand how recurrent networks reuse weights and propagate error across ordered token positions.
Read how Natural Language Processing in Action, Second Edition explains timestep →
Learn Generative AI with PyTorch
A discrete index into the diffusion/denoising schedule that determines how much noise is added or removed at a given stage.
- How it is used
- Conditions denoising predictions and indexes noise addition or removal during diffusion training and generation.
- How deep it goes
- Explains forward and reverse schedules, timestep embeddings, tensor broadcasting, and inference loops, but does not derive the full equations.
- Examples
- Uses 1,000-step flower diffusion, 50-step generation, and visual checkpoints at t = 800, 600, 400, 200, and 0.
- What you can do after
- Apply it via a denoising U-Net and DDIM scheduler to train noise prediction and generate images through selected timesteps.
Read how Learn Generative AI with PyTorch explains timestep →
Related concepts
- simulationused by · A simulation is a modeled or controlled enactment of a system, process, or scenario used to examine or reproduce behavior under specified conditions.
- U-Netused by · U-Net is an encoder-decoder neural-network architecture with downsampling and upsampling paths connected by skip connections for combining features at multiple spatial scales.
- diffusion modelpart of · A diffusion model is a generative model that learns to produce data by reversing a process that progressively adds noise to data.
- timestep embeddingproduces · A timestep embedding is a numerical vector representation of a diffusion process timestep used to condition a denoising model.
- denoisingpart of · Denoising is the process of removing unwanted noise or corruption from data while preserving meaningful structure.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.