Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

ENH: fix frame classification model to work with BioSoundSegBench #774

Merged
merged 24 commits into from
Sep 9, 2024
Merged

ENH: fix frame classification model to work with BioSoundSegBench #774

merged 24 commits into from
Sep 9, 2024

Commits on Jul 9, 2024

  1. Hack a learning rate scheduler into FrameClassificationModel

    @NickleDave
    NickleDave committed Jul 9, 2024
    Configuration menu
    Copy the full SHA
    3a1a827 View commit details
    Browse the repository at this point in the history

Commits on Jul 17, 2024

  1. Add boundary_labels parameter to transforms.frame_labels.transforms.P… 

    …ostProcess.__call__
    @NickleDave
    NickleDave committed Jul 17, 2024
    Configuration menu
    Copy the full SHA
    171fefe View commit details
    Browse the repository at this point in the history

  2. Add background_label to post_tfm_kwargs in src/vak/eval/frame_classif… 

    …ication.py
    @NickleDave
    NickleDave committed Jul 17, 2024
    Configuration menu
    Copy the full SHA
    e38f982 View commit details
    Browse the repository at this point in the history

  3. In validation_step of FrameClassificationModel, use boundary labels… 

    … when they are present to post-process multi-class frame labels'
    @NickleDave
    NickleDave committed Jul 17, 2024
    Configuration menu
    Copy the full SHA
    6f5b609 View commit details
    Browse the repository at this point in the history

Commits on Jul 18, 2024

  1. Unpack dataset_path from dataset_config in code block for built-in … 

    …datasets in eval/frame_classification.py, to make sure this variable exists when we build the DataFrame with eval results"
    @NickleDave
    NickleDave committed Jul 18, 2024
    Configuration menu
    Copy the full SHA
    d3f00dc View commit details
    Browse the repository at this point in the history

  2. Make variable frame_dur inside code block for built-in datasets ins… 

    …ide eval/frame_classification.py so this variable exists when we get the post-processing transform
    @NickleDave
    NickleDave committed Jul 18, 2024
    Configuration menu
    Copy the full SHA
    b5e6d49 View commit details
    Browse the repository at this point in the history

  3. Pass background_label into transforms.frame_labels.PostProcess insi… 

    …de eval/frame_classification.py, using `constants.DEFAULT_BACKGROUND_LABEL`
    @NickleDave
    NickleDave committed Jul 18, 2024
    Configuration menu
    Copy the full SHA
    5676d85 View commit details
    Browse the repository at this point in the history

  4. Fix how we call self.manual_backwward in FrameClassificationModel to … 

    …handle the case when the loss function returns a dict
    @NickleDave
    NickleDave committed Jul 18, 2024
    Configuration menu
    Copy the full SHA
    57ab70f View commit details
    Browse the repository at this point in the history

  5. In FrameClassificationModel.validation_step, convert boundary_preds t… 

    …o numpy when we pass them in to self.post_tfm
    @NickleDave
    NickleDave committed Jul 18, 2024
    Configuration menu
    Copy the full SHA
    a948b62 View commit details
    Browse the repository at this point in the history

  6. In FrameClassificationModel.validation_step, when logging accuracy, c… 

    …all it 'val_multi_acc' to distinguish from boundary_acc and for consistency with val_multi_acc_tfm
    @NickleDave
    NickleDave committed Jul 18, 2024
    Configuration menu
    Copy the full SHA
    bc40abc View commit details
    Browse the repository at this point in the history

  7. Change how we get and log frame_dur in train/frame_classification.py … 

    …so we have it as a separate variable; will use for post_tfm kwargs when we add those later
    @NickleDave
    NickleDave committed Jul 18, 2024
    Configuration menu
    Copy the full SHA
    36c67bf View commit details
    Browse the repository at this point in the history

  8. Change one-line summary of __call__ method for frame_labels.transform… 

    …s.PostProcess
    @NickleDave
    NickleDave committed Jul 18, 2024
    Configuration menu
    Copy the full SHA
    fd20e45 View commit details
    Browse the repository at this point in the history

Commits on Jul 19, 2024

  1. BUG: Ensure boundary_labels is 1d in post-process transform, fix #767

    @NickleDave
    NickleDave committed Jul 19, 2024
    Configuration menu
    Copy the full SHA
    2164f35 View commit details
    Browse the repository at this point in the history

Commits on Sep 4, 2024

  1. Fix what metric we use for learning rate scheduler: use val_multi_acc… 

    … for models with multiple accuracies
    @NickleDave
    NickleDave committed Sep 4, 2024
    Configuration menu
    Copy the full SHA
    f5d7046 View commit details
    Browse the repository at this point in the history

  2. Remove trainer module from common, code is used only for frame classi… 

    …fication model
    @NickleDave
    NickleDave committed Sep 4, 2024
    Configuration menu
    Copy the full SHA
    b055a07 View commit details
    Browse the repository at this point in the history

  3. Add get_trainer and get_callbacks to train/frame_classification.py, f… 

    …ix so that we monitor 'val_multi_acc' when a model has multiple targets, and just 'val_acc' otherwise
    @NickleDave
    NickleDave committed Sep 4, 2024
    Configuration menu
    Copy the full SHA
    1f83de8 View commit details
    Browse the repository at this point in the history

Commits on Sep 5, 2024

  1. Add missing self.manual_backward in training_step of FrameClassificat… 

    …ionModel
    @NickleDave
    NickleDave committed Sep 5, 2024
    Configuration menu
    Copy the full SHA
    2452922 View commit details
    Browse the repository at this point in the history

  2. Fix how we determine whether there are multiple targets and what to m… 

    …onitor in train/frame_classification.py
    @NickleDave
    NickleDave committed Sep 5, 2024
    Configuration menu
    Copy the full SHA
    f40a3d4 View commit details
    Browse the repository at this point in the history

Commits on Sep 6, 2024

  1. Fix how we validate boundary_labels in transforms.frame_labels.functi… 

    …onal.postprocess -- don't if boundary_labels is None
    @NickleDave
    NickleDave committed Sep 6, 2024
    Configuration menu
    Copy the full SHA
    8f98f1c View commit details
    Browse the repository at this point in the history

  2. Fix vak/predict/frame_classification.py to handle edge case where no … 

    …non-background segments are predicted for any sample in dataset
    @NickleDave
    NickleDave committed Sep 6, 2024
    Configuration menu
    Copy the full SHA
    6111356 View commit details
    Browse the repository at this point in the history

Commits on Sep 7, 2024

  1. Revise comment

    @NickleDave
    NickleDave committed Sep 7, 2024
    Configuration menu
    Copy the full SHA
    5f0da8d View commit details
    Browse the repository at this point in the history

  2. Catch edge case in transforms.frame_labels.functional.boundary_inds_f… 

    …rom_frame_boundary_labels
    @NickleDave
    NickleDave committed Sep 7, 2024
    Configuration menu
    Copy the full SHA
    02974d3 View commit details
    Browse the repository at this point in the history

Commits on Sep 9, 2024

  1. Add minimal unit tests for vak.transforms.frame_labels.functional.bou… 

    …ndary_inds_from_boundary_labels
    @NickleDave
    NickleDave committed Sep 9, 2024
    Configuration menu
    Copy the full SHA
    a07f17b View commit details
    Browse the repository at this point in the history

  2. Remove learning rate scheduler for now

    @NickleDave
    NickleDave committed Sep 9, 2024
    Configuration menu
    Copy the full SHA
    608c0fc View commit details
    Browse the repository at this point in the history