I have a training script that has a parameter seed
and the output model’s name is dependent on the seed, e.g. when seed = 0
, the model name is model-0.ckpt
. What is the correct way to add a stage? I tried the following but the generated dvc.yaml
and .gitignore
file is not correct
dvc stage add -n train \
-d data/train.pkl -d data/val.pkl -d data/test.pkl \
-d train.py \
-p seed \
-o models/model-${seed}.ckpt \
python train.py -s ${seed}
train:
cmd: python train.py -s -d
deps:
- data/test.pkl
- data/train.pkl
- data/val.pkl
- train.py
params:
- seed
outs:
- models/model-.ckpt