Companion slides for the pipeline reproducibility meetup at Adsmurai. See links for code.
https://github.com/Adsmurai/dvc-meetup
https://www.meetup.com/BCN-DL-School/events/268262173/
7. Index
● Explicit data and process dependencies
● Data and model caching
● Visualize metrics across model and data versions
● “One click” pipeline reproducibility
● 🍻 🍕
8. Explicit data and process dependencies
raw
Prepare
data
prepared
train
prepared
test
Extract
features
features
train
features
test
Select
model
Test
model
model metrics
19. Explicit data and process dependencies
$ dvc pipeline show --ascii select_model.dvc
$ dvc pipeline show --ascii --outs select_model.dvc
20. Data and model caching
raw
Prepare
data
prepared
train
prepared
test
Extract
features
features
train
features
test
Select
model
Test
model
model metrics
21. Data and model caching
raw
Prepare
data
prepared
train
prepared
test
Extract
features
features
train
features
test
Select
model
Test
model
model metrics
CHANGE HERE
22. Data and model caching
raw
Prepare
data
prepared
train
prepared
test
Extract
features
features
train
features
test
Select
model
Test
model
model metrics
CHANGE HERE
$ dvc repro test_model.dvc