I have the following setup:
- model: my_model
operations:
train:
main: scripts.train
flags:
a: 1
evaluate:
main: scripts.evaluate
requires:
- operation: train
train_evaluate:
flags:
a: 1
steps:
- run: train a=${a}
- run: evaluate
compare_evaluate:
main: scripts.compare
requires:
- train_evaluate_run_1
- train_evaluate_run_2
compare:
steps:
- run: train_evaluate a=1
- run: train_evaluate a=2
- run: compare_evaluate # HERE
resources:
train_evaluate_run_1:
- operation: train_evaluate
name: train_evaluate_run_1
train_evaluate_run_2:
- operation: train_evaluate
name: train_evaluate_run_2
How to tell guild to resolve two different resources in the compare_evaluate
run (with the # HERE tag)?
The goal is to have a script/notebook that takes as input two train_evaluate
runs and compares them using custom plotting etc. and have all that specified in a single pipeline.