Published machine-learning models
Models published for others to use, held as records — so one can be found by what it is for rather than by what it is called.
- frameworks declared
- 6
- per record
- 2.7 KB
How it divides
1 dimension · press a band to read it- pyTorch3,135 · 31.1%
- Transformers2,938 · 29.1%
- tensorFlow21,065 · 10.6%
- Keras970 · 9.6%
- Other963 · 9.5%
- Remaining 3 values1,013 · 10.0%
3,135 models are filed under pyTorch — 31.1% of those that declare it.
10,084 of 14,075 declared · 71.6%. The rest carry no value for this field, so these shares are of the declared set and do not add up to the collection.
What you can do with it
Find one by what it does — Describe the task in plain words. Model names carry almost no information about their purpose, which is exactly the problem.
Read it beside the research — Repositories and papers are on neighbouring shelves, so a model and the work behind it turn up in one search.
Follow it back — Every record keeps the address it was read from, so a model can be checked where it was published.
Where it stops
It is not a benchmark table. Nothing here ranks models or tells you which one is better at your task.
It is not a download or a hosting service. The record says a model was published; running it is somewhere else entirely.