Skip to content

Commit

Permalink
Merge pull request #24 from Music-and-Culture-Technology-Lab/dev
Browse files Browse the repository at this point in the history
Fix bug while parsing the path of weight files according to issue #23
  • Loading branch information
BreezeWhite authored Jun 4, 2021
2 parents 19dd6a8 + 8ec2ac3 commit f4795eb
Show file tree
Hide file tree
Showing 13 changed files with 5,737 additions and 27 deletions.
24 changes: 24 additions & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
@@ -1,11 +1,30 @@
# Changelog

## 0.4.0 - 2021-05-
Various improvements on music module and some critical bug fixes.

## Enhancement
- Improve the peak finding and thresholding strategy for more stable and better performance.
- Modify the feeding strategy of feature slices with adjustable overlapping rate while making predictions.
- Apply learning rate scheduler for music module.
- Replace the usage of custom training loop of music module with the built-in TF `.fit()` function.

## Bugs
- Fix a critical bug of inference of music module that would lead to missing onsets.
- Fix generation of pertubation of vocal module while training.

## Documentation
- Merge the demo page into master from `build_doc` branch.

---

## 0.3.4 - 2021-05-10
Hotifx version according to issue #19.

## Bugs
- Fix bug of treating numpy array as list while appending elements.

---

## 0.3.3 - 2021-05-07
Hotfix version according to issue #19.
Expand All @@ -14,6 +33,7 @@ Hotfix version according to issue #19.
- Fix column inconsistency of `aggregate_f0_info` and `write_agg_f0_results`.
- Update version of dependencies according to the security alert.

---

## 0.3.2 - 2021-02-13

Expand All @@ -26,13 +46,17 @@ Hotfix version according to issue #19.
- Fix wrong access name of the dict in vocal midi inference function.
- Fix bug of generating beat module training labels.

---

## 0.3.1 - 2021-01-18

Hotfix release of spleeter error.

### Bugs
- Call Spleeter in CLI mode instead of using python class.

---

## 0.3.0 - 2021-01-17

Release the `beat` module for symbolic domain beat transcription.
Expand Down
2 changes: 1 addition & 1 deletion omnizart/__init__.py
Original file line number Diff line number Diff line change
Expand Up @@ -7,4 +7,4 @@
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
os.environ['VAMP_PATH'] = os.path.join(MODULE_PATH, "resource", "vamp")

__version__ = "0.3.4"
__version__ = "0.4.1-rc0"
7 changes: 6 additions & 1 deletion omnizart/base.py
Original file line number Diff line number Diff line change
Expand Up @@ -99,8 +99,13 @@ def _resolve_model_path(self, model_path=None):
logger.warning("There are multiple checkpoints in the directory. Default to use %s", cand_dirs[0])
model_path = os.path.join(model_path, cand_dirs[0])

# Parse weight files if there are multiple
weight_files = glob.glob(os.path.join(model_path, "*.h5"))
if len(weight_files) == 0:
weight_path = os.path.join(model_path, "weights.h5")
else:
weight_path = sorted(weight_files)[-1]
arch_path = os.path.join(model_path, "arch.yaml")
weight_path = os.path.join(model_path, "weights.h5")
conf_path = os.path.join(model_path, "configurations.yaml")

return arch_path, weight_path, conf_path
Expand Down
Loading

0 comments on commit f4795eb

Please sign in to comment.