sktime/pytorch-forecasting
Time series forecasting with PyTorch
In plain words
Outside PRs: 6 in 10 merged. First reply usually within 3.6 days. 1 starter issue is available now.
Outside PRs merged
60%
6 merged, 4 closed, 42 still open
Higher than 14% of 565 other measured repositories
First human reply
3.6 days
median of 52 outside PRs, 20 unanswered
Outside PRs opened 30 to 120 days ago, so each has had time to be answered. A figure needs at least 5 pull requests, otherwise it reads "Not enough data". How this is calculated · A number looks wrong?
Your first steps here
If you have never contributed to this project, do these in order.
- 1
Read the contributing guide
It says how to set the project up and what the maintainers expect in a pull request.
Open the guide - 2
Take this free issue
[DOC] Add more tutorials for v2
Open issue #2354 - 3
Know where to ask
Questions go to the project's discord, or on the issue itself. Not in private messages.
Open it - 4
Open a small pull request
Link the issue in it. A first reply here usually takes 3.6 days; until then, silence is normal.
What happens to an outside pull request
Every pull request opened by someone outside the core team, 30 to 120 days ago.
- Opened by outsiders52100%
- Got a human reply3262%
- Merged612%
Time to first reply
After 2 days, 38% of outside PRs had a reply from a person. After a week, 54%. Bots do not count.
The last 52 weeks
Steady. Median first reply was 3.8 days over the last four weeks (21 PRs) and 3.1 days over the four before (26 PRs).
Starter issues
Issues labelled for beginners, with their real state. Available means open, unassigned, no linked pull request, nobody has claimed it in the last 14 days, and it was updated in the last 60 days.
- Available
- 1
- Claimed
- 0
- Has a pull request
- 15
- Stale
- 0
- Has a pull request[DOC] Add usage examples to all models in v1 and v2#2377
- Has a pull request[DOC] replace `conda` as recommended env manager with `uv` and/or `venv`#2364
- Has a pull request[BUG] `y` not being normalized when using `EncoderNormalizer` in `EncoderDecoderDataModule`
When replies arrive
All maintainers counted together. Shown only when at least 3 people replied in the last 90 days, so it never describes one person.
Replies are most likely between 19:00 and 22:00 (India time).
Show as table
| Hour (India time) | Replies, all weekdays |
|---|---|
| 00:00 | 18 |
| 01:00 | 11 |
| 02:00 | 2 |
| 03:00 | 2 |
| 04:00 | 2 |
| 05:00 | 0 |
| 06:00 | 0 |
| 07:00 | 4 |
| 08:00 | 4 |
| 09:00 | 1 |
| 10:00 | 3 |
| 11:00 | 8 |
| 12:00 | 14 |
| 13:00 | 13 |
| 14:00 | 9 |
| 15:00 | 17 |
| 16:00 | 24 |
| 17:00 | 14 |
| 18:00 | 12 |
| 19:00 | 22 |
| 20:00 | 24 |
| 21:00 | 15 |
| 22:00 | 7 |
| 23:00 | 11 |
How to start
- Contributing guidefound
- Code of conductfound
- Issue templatesfound
- Dev container for one-step setupnot found
- No CLA or sign-off requirement was found.
Where the project talks
Activity
- Commits, 90 days
- 35
- Releases, 12 months
- 4
- Active maintainers
- 5
- Last push
- 9 Oct 2026
Stack
Badge and feed
Two ways to use these figures outside this page.
For maintainers: a README badge
Shows newcomers how quickly this project answers and how often it merges. Paste the line into your README; the badge updates with the index.
[](https://contributable.vercel.app/repo/sktime/pytorch-forecasting)[](https://contributable.vercel.app/repo/sktime/pytorch-forecasting)For contributors: follow new issues
Add the feed to a feed reader (Feedly, Inoreader, Thunderbird) and you hear about new free starter issues here without checking the site.
https://contributable.vercel.app/feed/repo/sktime/pytorch-forecastingFor developers: every figure on this page as JSON at .
Similar repositories
Same main language (Python), most shared frameworks and topics first, then the faster reply.
A unified framework for machine learning with time series
- First reply
- 5.3 days
