Prefetch
Prefetching is the proactive retrieval or preparation of anticipated data or resources before they are needed.
Latency
The proactive retrieval of anticipated data before it is explicitly requested.
- How it is used
- Hides read latency by retrieving anticipated data while other work proceeds.
- How deep it goes
- Explains prediction strategies, a sequential prefetch window, and failure costs, but provides no application implementation code.
- Examples
- Illustrates the idea with login-time user-data retrieval and sequential file processing.
- What you can do after
- Operate/configure it by choosing an access predictor and tuning a resident window while accounting for wasted reads.
Deep Learning with Python, Third Edition
A data-input optimization that overlaps data preparation with model execution by preparing upcoming batches in advance.
- How it is used
- Keeps training supplied with prepared batches by overlapping data preprocessing with model execution.
- How deep it goes
- Defines asynchronous preparation and shows buffered pipelines, but stops short of scheduling internals or buffer-size tuning beyond AUTOTUNE.
- Examples
- Demonstrates `.batch(16).prefetch(2)` for disk-loaded images and `.batch(128).prefetch(8)` for tokenized text.
- What you can do after
- Apply it via `tf.data.Dataset.prefetch()` before Keras `fit()`, using `tf.data.AUTOTUNE` when the buffer size is uncertain.
Read how Deep Learning with Python, Third Edition explains prefetch →
Deep Learning with R, Third Edition
Data prefetching prepares future input elements asynchronously while current elements are being processed.
- How it is used
- Improves training input flow by preparing future batches while current batches run, particularly for GPU and distributed workloads.
- How deep it goes
- Explains buffering, streaming, and buffer-size selection with R pipelines, but does not derive scheduling or runtime internals.
- Examples
- Demonstrates prefetching in image classification, COCO object detection, and mini-GPT text-tokenization pipelines.
- What you can do after
- Apply it via `dataset_prefetch()` to overlap input preparation with model computation and select a fixed or automatic buffer.
Read how Deep Learning with R, Third Edition explains prefetch →
Related concepts
- tf.data APIconfigures · The tf.data API is TensorFlow’s API for constructing input pipelines that read, transform, batch, and iterate over data efficiently.
- predictive techniquekind of · A predictive technique performs an operation before its need is confirmed by forecasting future access, demand, or execution behavior.
- lazy initializationuses · Lazy initialization is a technique that postpones creating, retrieving, computing, or loading a value or resource until it is needed.
- optimistic updatecontrasts with · An optimistic update presents an expected change in client-side state before the corresponding operation is confirmed, then reconciles or reverses the change if needed.
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.