Atlas record: https://www.aitimeline.dev/en/events/evt-0234/
Availability · January 2017
PyTorch
Facebook released a library in which the computation graph is built as it runs, so a model can be debugged with ordinary language tools.
Why it matters
Researchers got back the ability to write an experiment as an ordinary program, and the field moved onto this tool within three years.
TensorFlow required declaring the whole graph and then running it, so an error in the declaration surfaced far from where it was made. PyTorch executes operations immediately, so ordinary debuggers and printing intermediate values work. For research, where a model changes daily, that proved decisive. By 2020 most papers at the main conferences used PyTorch.
Event record
- Event date
- January 2017 · Approximate date
- Timeline date
- Event date
- Verification
- Sources gathered automatically · September 17, 2026
- Lines
- ID
- evt-0234
The public release was in early 2017; the paper describing it appeared in 2019. Sources do not give an exact release date.
Related events
Earlier
- Contrasts TensorFlow 2015
The computation graph is built during execution rather than declared in advance.
- Enables Chainer: the graph is built as it computes 2015links to this record
The PyTorch README names Chainer among its inspirations, and the 2019 PyTorch paper calls it a framework of the same approach.
Later