Actual main

configs/experiment/im2im/diffae.yaml

@@ -47,3 +47,5 @@ callbacks:
     num_pcs: 4
     n_steps: 3
     every_n_epoch: ${model.save_images_every_n_epochs}
+  model_checkpoint:
+    monitor: null

configs/logger/mlflow.yaml

@@ -1,5 +1,5 @@
 mlflow:
   _target_: cyto_dl.loggers.MLFlowLogger
-  tracking_uri: https://mlflow.a100.int.allencell.org/
+  tracking_uri:
   experiment_name: ${experiment_name}
   run_name: ${run_name}

configs/model/im2im/diffae.yaml

@@ -24,11 +24,10 @@ loss:
   reduction: none
 
 semantic_encoder:
-  _target_: cyto_dl.nn.resnet.ResNet
-  out_dim: 8
+  _target_: cyto_dl.nn.TorchVisionWrapper
   base_encoder:
-    _partial_: True
     _target_: torchvision.models.resnet18
+    num_classes: 8
 
   # 3D
   # _target_: monai.networks.nets.Regressor

cyto_dl/dataframe/readers.py

@@ -9,47 +9,9 @@
     import modin.pandas as pd
 except ModuleNotFoundError:
     import pandas as pd
-
-import anndata
 import pyarrow.parquet
 
 
-def read_h5ad(path, include_columns=None, backed=None):
-    """Read an annData object stored in a .h5ad file.
-
-    Parameters
-    ----------
-
-    path: Union[Path, str]
-        Path to the .h5ad file
-
-    include_columns: Optional[Sequence[str]] = None
-        List of column names and/or regex expressions, used to only include the
-        desired columns in the resulting dataframe.
-
-    backed: Optional[str] = None
-        Can be (either "r" or "r+").
-        See anndata's docs for details:
-        https://anndata.readthedocs.io/en/latest/generated/anndata.read_h5ad.html#anndata.read_h5ad
-
-    Returns
-    -------
-    annData
-    """
-
-    if backed:
-        assert backed in ("r", "r+")
-    dataframe = anndata.read_hda5(path, backed=backed)
-
-    if include_columns is not None:
-        columns = []
-        for filter_ in include_columns:
-            columns += filter_columns(dataframe.obs.columns.tolist(), regex=filter_)
-
-        dataframe.obs = dataframe.obs[columns]
-    return dataframe
-
-
 def read_parquet(path, include_columns=None):
     """Read a dataframe stored in a .parquet file, and optionally include only the columns given by
     `include_columns`
@@ -178,28 +140,12 @@ def read_dataframe(
                 columns += filter_columns(dataframe.columns, regex=filter_)
 
             dataframe = dataframe[columns]
-    elif isinstance(dataframe, anndata.AnnData):
-        if include_columns is not None:
-            columns = []
-            for filter_ in include_columns:
-                columns += filter_columns(dataframe.obs.columns, regex=filter_)
-
-            dataframe.obs = dataframe.obs[columns]
     else:
         raise TypeError(
             f"`dataframe` must be either a pd.DataFrame or a path to "
             f"a file to load one. You passed {type(dataframe)}"
         )
 
-    if isinstance(dataframe, anndata.AnnData):
-        # Make dataframe out of anndata object
-        X = dataframe.X.toarray()
-        X = pd.DataFrame(X, columns=[f"X_{i}" for i in range(X.shape[1])]).reset_index(drop=True)
-        index = pd.DataFrame(dataframe.obs_names)
-        dataframe = dataframe.obs.reset_index(drop=True)
-        dataframe = pd.concat([X, dataframe], axis=1)
-        dataframe = pd.concat([index, dataframe], axis=1)
-
     if required_columns is not None:
         missing_columns = set(required_columns) - set(dataframe.columns)
         if missing_columns:

cyto_dl/image/__init__.py

@@ -1,4 +1,4 @@
-from .io import MonaiBioReader, OmeZarrReader
+from .io import MonaiBioReader
 from .transforms import (
     BrightSampler,
     RandomMultiScaleCropd,

cyto_dl/image/io/__init__.py

@@ -1,6 +1,5 @@
 from .bioio_loader import BioIOImageLoaderd
 from .monai_bio_reader import MonaiBioReader
 from .numpy_reader import ReadNumpyFile
-from .ome_zarr_reader import OmeZarrReader
 from .polygon_loader import PolygonLoaderd
 from .skimage_reader import SkimageReader

cyto_dl/image/io/bioio_loader.py

@@ -84,6 +84,7 @@ def __call__(self, data):
         if self.resolution_key in data:
             img.set_resolution_level(data[self.resolution_key])
         kwargs = {k: self.split_args(data[k]) for k in self.kwargs_keys if k in data}
+
         if self.dask_load:
             img = img.get_image_dask_data(**kwargs).compute()
         else:

cyto_dl/image/transforms/__init__.py

@@ -5,7 +5,7 @@
 from .multiscale_cropper import RandomMultiScaleCropd
 from .pad import PadZd
 from .project import Projectd  # codespell:ignore
-from .save import Save, Saved
+# from .save import Save, Saved
 
 try:
     from .rotation_mask_transform import RotationMask, RotationMaskd

cyto_dl/image/transforms/save.py

@@ -1,68 +1,68 @@
-from pathlib import Path
-from typing import Sequence, Union
+# from pathlib import Path
+# from typing import Sequence, Union
 
-import numpy as np
-import torch
-from bioio.writers import OmeTiffWriter
-from monai.data.meta_tensor import MetaTensor
-from monai.transforms import Transform
-from omegaconf import ListConfig
+# import numpy as np
+# import torch
+# from bioio.writers import OmeTiffWriter
+# from monai.data.meta_tensor import MetaTensor
+# from monai.transforms import Transform
+# from omegaconf import ListConfig
 
 
-class Save(Transform):
-    def __init__(
-        self,
-        save_path: str = "./",
-    ):
-        """
-        Parameters
-        ----------
-        keys: Sequence[str]
-            keys to save
-        save_path: str
-            path to save images
-        """
-        super().__init__()
-        self.save_path = Path(save_path)
-        self.count = 0
+# class Save(Transform):
+#     def __init__(
+#         self,
+#         save_path: str = "./",
+#     ):
+#         """
+#         Parameters
+#         ----------
+#         keys: Sequence[str]
+#             keys to save
+#         save_path: str
+#             path to save images
+#         """
+#         super().__init__()
+#         self.save_path = Path(save_path)
+#         self.count = 0
 
-    def __call__(self, img, name="img"):
-        OmeTiffWriter.save(
-            uri=self.save_path / f"{name}_{self.count}.tif",
-            data=img if not isinstance(img, (torch.Tensor, MetaTensor)) else img.numpy(),
-        )
-        self.count += 1
-        return img
+#     def __call__(self, img, name="img"):
+#         OmeTiffWriter.save(
+#             uri=self.save_path / f"{name}_{self.count}.tif",
+#             data=img if not isinstance(img, (torch.Tensor, MetaTensor)) else img.numpy(),
+#         )
+#         self.count += 1
+#         return img
 
 
-class Saved(Transform):
-    """Save a batch of images to disk for debugging."""
+# class Saved(Transform):
+#     """Save a batch of images to disk for debugging."""
 
-    def __init__(
-        self,
-        keys: Sequence[str],
-        save_path: str = "./",
-        allow_missing_keys: bool = False,
-    ):
-        """
-        Parameters
-        ----------
-        keys: Sequence[str]
-            keys to save
-        save_path: str
-            path to save images
-        allow_missing_keys: bool
-            allow missing keys in batch
-        """
-        super().__init__()
-        self.keys = keys if isinstance(keys, (list, ListConfig)) else [keys]
-        self.allow_missing_keys = allow_missing_keys
-        self.saver = Save(save_path)
+#     def __init__(
+#         self,
+#         keys: Sequence[str],
+#         save_path: str = "./",
+#         allow_missing_keys: bool = False,
+#     ):
+#         """
+#         Parameters
+#         ----------
+#         keys: Sequence[str]
+#             keys to save
+#         save_path: str
+#             path to save images
+#         allow_missing_keys: bool
+#             allow missing keys in batch
+#         """
+#         super().__init__()
+#         self.keys = keys if isinstance(keys, (list, ListConfig)) else [keys]
+#         self.allow_missing_keys = allow_missing_keys
+#         self.saver = Save(save_path)
 
-    def __call__(self, img_dict):
-        for key in self.keys:
-            if key in img_dict:
-                self.saver(img_dict[key], key)
-            elif not self.allow_missing_keys:
-                raise ValueError(f"key {key} found in data. Available keys are {img_dict.keys()}")
-        return img_dict
+#     def __call__(self, img_dict):
+#         for key in self.keys:
+#             if key in img_dict:
+#                 self.saver(img_dict[key], key)
+#             elif not self.allow_missing_keys:
+#                 raise ValueError(f"key {key} found in data. Available keys are {img_dict.keys()}")
+#         return img_dict