From ef45018d242f0ce554fe390e4ab864cee9a9e863 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Fran=C3=A7ois=20Laurent?= <francois.laurent@posteo.net>
Date: Wed, 8 Feb 2023 00:30:34 +0100
Subject: [PATCH] implements #22
---
src/taggingbackends/main.py | 25 +++++++++++++++++++------
1 file changed, 19 insertions(+), 6 deletions(-)
diff --git a/src/taggingbackends/main.py b/src/taggingbackends/main.py
index a6c6d9c..31655f9 100644
--- a/src/taggingbackends/main.py
+++ b/src/taggingbackends/main.py
@@ -10,7 +10,8 @@ Usage: tagging-backend [train|predict] --model-instance <name>
tagging-backend train ... --sample-size <N> --balancing-strategy <strategy>
tagging-backend train ... --frame-interval <I> --window-length <T>
tagging-backend train ... --pretrained-model-instance <name>
- tagging-backend predict ... --skip-make-dataset --sandbox <token>
+ tagging-backend train ... --skip-make-dataset --skip-build-features
+ tagging-backend predict ... --make-dataset --build-features --sandbox <token>
`tagging-backend` typically is run using `poetry run`.
A name must be provided to identify the trained model and its location within
@@ -36,6 +37,10 @@ the `make_dataset` module is loaded and this may take quite some time due to
dependencies (e.g. Julia FFI). The `--skip-make-dataset` option makes `train`
truly skip this step; the corresponding module is not loaded.
+Since version 0.8, `predict` makes `--skip-make-dataset` and
+`--skip-build-features` the default behavior. As a counterpart, it admits
+arguments `--make-dataset` and `--build-features`.
+
`--sandbox <token>` makes `tagging-backend` use a token instead of <name> as
directory name in data/raw, data/interim and data/processed.
This is intended to prevent conflicts on running `predict` in parallel on
@@ -59,7 +64,7 @@ def main(fun=None):
input_files, labels = [], []
sample_size = window_length = frame_interval = None
trxmat_only = reuse_h5files = False
- skip_make_dataset = skip_build_features = False
+ make_dataset = build_features = None
pretrained_model_instance = None
sandbox = False
balancing_strategy = 'auto'
@@ -92,9 +97,13 @@ def main(fun=None):
elif sys.argv[k] == "--reuse-h5files":
reuse_h5files = True
elif sys.argv[k] == "--skip-make-dataset":
- skip_make_dataset = True
+ make_dataset = False
elif sys.argv[k] == "--skip-build-features":
- skip_build_features = True
+ build_features = False
+ elif sys.argv[k] == '--make-dataset':
+ make_dataset = True
+ elif sys.argv[k] == '--build-features':
+ build_features = True
elif sys.argv[k] == "--pretrained-model-instance":
k = k + 1
pretrained_model_instance = sys.argv[k]
@@ -116,7 +125,11 @@ def main(fun=None):
if input_files:
for file in input_files:
backend.move_to_raw(file)
- if not skip_make_dataset:
+ if make_dataset is None and train_or_predict == 'train':
+ make_dataset = True
+ if build_features is None and train_or_predict == 'train':
+ build_features = True
+ if make_dataset:
make_dataset_kwargs = dict(labels_expected=train_or_predict == "train",
balancing_strategy=balancing_strategy)
if labels:
@@ -134,7 +147,7 @@ def main(fun=None):
elif reuse_h5files:
logging.info("option --reuse-h5files is ignored in the absence of --trxmat-only")
backend._run_script(backend.make_dataset, **make_dataset_kwargs)
- if not skip_build_features:
+ if build_features:
backend._run_script(backend.build_features)
if train_or_predict == "predict":
backend._run_script(backend.predict_model, trailing=unknown_args)
--
GitLab