Parallelise _only_ the convolution+loss
Nobody has claimed this yet.
Assessment
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Newbie friendliness
- 35/100
- Issue type
- Feature
- Clarity
- Mostly clear
- Activity status
- Stale
- Tech stack
- machine-learning, python
- Domain
- machine-learning
Research direction
The issue asks to refactor the training loop so that convolution and loss computation can run in parallel per model. Start by locating where models are stacked and fitted, then identify the fktable and loss computation that happens after sampling. The goal is to separate these components so each model trains independently without weight freezing, enabling per-model parallelization.
Written by the indexing model from the issue text.
Description
This is probably quite complicated to do without harming the performance of the "single-model" training but it could be quite interesting.
Right now we are stacking together all the models but keeping track of the separated component so that we can stop the fit for each of them independently.
This is very efficient until the fits don't stabilise. Then one by one the fit stop (by freezing the weights) until eventually the last one stops and the fit finishes.
However, the models we use are not very heavy. Most (almost all in the tests we did... 3 years ago at this point...) the advantage of GPU running happens after the model is sampled: fktable and loss. We can probably separate them completely such that each model trains completely by itself (meaning we have complete freedom on the model side, and are not restricted to the architecture).
- Dominant language
- Python
- Stars
- 44
- Forks
- 15
- Avg merge
- 8d 12h
- Merged PRs (30d)
- 7
Getting set up
This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from NNPDF/nnpdf
-
bug
Difficulty 2/5 1-3 hours Newbie friendliness 68/100
-
validphys
Difficulty 1/5 Under an hour Newbie friendliness 70/100
-
Difficulty 2/5 1-3 hours Newbie friendliness 65/100
-
Invalidate cached resourcesPossibly taken @felixhekhorn claimed this today. Open
-
Extend FK table hash computationPossibly taken @jekoorn claimed this today. Open
Similar issues
-
Difficulty 2/5 1-3 hours Newbie friendliness 77/100
mozilla-ai/cq#601 ·
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 68/100
Maintainers usually reply within 1 day
-
intcomma returns exponential notation for large floatsPossibly taken @Omarinho12345 claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
python-humanize/humanize#521 ·
Maintainers usually reply within 1 day
-
LUCJ pass manager raises NoEdgeBetweenNodes when an edge is bad for two two-qubit gatesPossibly taken @hat-im claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
Maintainers usually reply within 1 day
-
opensubtitlescom: provider rejects pt-PT even though its own converter maps itPossibly taken @paulo-0308 claimed this today. Open
Difficulty 1/5 Under an hour Newbie friendliness 88/100
Diaoul/subliminal#1405 ·