slide2vec 1.3.0__tar.gz → 2.0.0__tar.gz

{slide2vec-1.3.0/slide2vec.egg-info → slide2vec-2.0.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: slide2vec
-Version: 1.3.0
+Version: 2.0.0
 Summary: Embedding of whole slide images with Foundation Models
 Home-page: https://github.com/clemsgrs/slide2vec
 Author: Clément Grisi
@@ -95,7 +95,7 @@ pip install slide2vec
 
    A good starting point is the default configuration file `slide2vec/configs/default.yaml` where parameters are documented.<br>
    We've also added default configuration files for each of the foundation models currently supported:
-   - tile-level: `uni`, `uni2`, `virchow`, `virchow2`, `prov-gigapath`, `h-optimus-0`, `h-optimus-1`
+   - tile-level: `uni`, `uni2`, `virchow`, `virchow2`, `prov-gigapath`, `h-optimus-0`, `h-optimus-1`, `h0-mini`, `conch`, `musk`, `phikonv2`, `hibou-b`, `hibou-L`, [`kaiko`](https://github.com/kaiko-ai/towards_large_pathology_fms)
    - slide-level: `prov-gigapath`, `titan`, `prism`
 
 

{slide2vec-1.3.0 → slide2vec-2.0.0}/README.md

@@ -41,7 +41,7 @@ pip install slide2vec
 
    A good starting point is the default configuration file `slide2vec/configs/default.yaml` where parameters are documented.<br>
    We've also added default configuration files for each of the foundation models currently supported:
-   - tile-level: `uni`, `uni2`, `virchow`, `virchow2`, `prov-gigapath`, `h-optimus-0`, `h-optimus-1`
+   - tile-level: `uni`, `uni2`, `virchow`, `virchow2`, `prov-gigapath`, `h-optimus-0`, `h-optimus-1`, `h0-mini`, `conch`, `musk`, `phikonv2`, `hibou-b`, `hibou-L`, [`kaiko`](https://github.com/kaiko-ai/towards_large_pathology_fms)
    - slide-level: `prov-gigapath`, `titan`, `prism`
 
 

{slide2vec-1.3.0 → slide2vec-2.0.0}/pyproject.toml

@@ -23,7 +23,7 @@ warn_unused_configs = true
 no_implicit_reexport = true
 
 [tool.bumpver]
-current_version = "1.3.0"
+current_version = "2.0.0"
 version_pattern = "MAJOR.MINOR.PATCH"
 commit = false       # We do version bumping in CI, not as a commit
 tag = false          # Git tag already exists — we don't auto-tag

{slide2vec-1.3.0 → slide2vec-2.0.0}/setup.cfg

@@ -1,6 +1,6 @@
 [metadata]
 name = slide2vec
-version = 1.3.0
+version = 2.0.0
 description = Embedding of whole slide images with Foundation Models
 author = Clément Grisi
 platforms = unix, linux, osx, cygwin, win32

slide2vec-2.0.0/slide2vec/__init__.py

@@ -0,0 +1 @@
+__version__ = "2.0.0"

{slide2vec-1.3.0 → slide2vec-2.0.0}/slide2vec/aggregate.py

@@ -28,14 +28,20 @@ def get_args_parser(add_help: bool = True):
         "--config-file", default="", metavar="FILE", help="path to config file"
     )
     parser.add_argument(
-        "--run-id",
+        "--output-dir",
         type=str,
-        default="",
-        help="Name of output subdirectory",
+        default=None,
+        help="output directory to save logs and checkpoints",
     )
     parser.add_argument(
         "--run-on-cpu", action="store_true", help="run inference on cpu"
     )
+    parser.add_argument(
+        "opts",
+        help="Modify config options at the end of the command using \"path.key=value\".",
+        default=None,
+        nargs=argparse.REMAINDER,
+    )
     return parser
 
 
@@ -54,7 +60,7 @@ def main(args):
     # setup configuration
     run_on_cpu = args.run_on_cpu
     cfg = get_cfg_from_file(args.config_file)
-    output_dir = Path(cfg.output_dir, args.run_id)
+    output_dir = Path(cfg.output_dir, args.output_dir)
     cfg.output_dir = str(output_dir)
 
     coordinates_dir = Path(cfg.output_dir, "coordinates")
@@ -71,6 +77,11 @@ def main(args):
         process_list.is_file()
     ), "Process list CSV not found. Ensure tiling has been run."
     process_df = pd.read_csv(process_list)
+    if "aggregation_status" not in process_df.columns:
+        process_df["aggregation_status"] = ["tbp"] * len(process_df)
+        cols = ["wsi_name", "wsi_path", "mask_path", "tiling_status", "feature_status", "aggregation_status", "error", "traceback"]
+        process_df = process_df[cols]
+
     skip_feature_aggregation = process_df["aggregation_status"].str.contains("success").all()
 
     if skip_feature_aggregation and distributed.is_main_process():
@@ -111,6 +122,9 @@ def main(args):
             coordinates = (np.array([coordinates_arr["x"], coordinates_arr["y"]]).T).astype(int)
 
             feature_path = features_dir / f"{name}.pt"
+            output_path = features_dir / f"{name}.pt"
+            if cfg.model.save_tile_embeddings:
+                feature_path = features_dir / f"{name}-tiles.pt"
 
             # run forward pass with slide encoder
             if cfg.model.name == "prov-gigapath":
@@ -132,13 +146,19 @@ def main(args):
                 with autocast_context:
                     features = torch.load(feature_path).to(model.device)
                     tile_size_lv0 = coordinates_arr["tile_size_lv0"][0]
-                    wsi_feature = model.forward_slide(
+                    output = model.forward_slide(
                         features,
                         tile_coordinates=coordinates,
                         tile_size_lv0=tile_size_lv0,
                     )
-
-            torch.save(wsi_feature, feature_path)
+                    wsi_feature = output["embedding"].cpu()
+                    if cfg.model.name == "prism" and cfg.model.save_latents:
+                        latent_path = features_dir / f"{name}-latents.pt"
+                        latents = output["latents"].cpu()
+                        torch.save(latents, latent_path)
+                        del latents
+
+            torch.save(wsi_feature, output_path)
             del wsi_feature
             if not run_on_cpu:
                 torch.cuda.empty_cache()

{slide2vec-1.3.0 → slide2vec-2.0.0}/slide2vec/data/dataset.py

@@ -2,6 +2,7 @@ import torch
 import numpy as np
 import wholeslidedata as wsd
 
+from transformers.image_processing_utils import BaseImageProcessor
 from PIL import Image
 from pathlib import Path
 
@@ -58,5 +59,8 @@ class TileDataset(torch.utils.data.Dataset):
         if self.tile_size[idx] != self.tile_size_resized[idx]:
             tile = tile.resize((self.tile_size[idx], self.tile_size[idx]))
         if self.transforms:
-            tile = self.transforms(tile)
+            if isinstance(self.transforms, BaseImageProcessor):  # Hugging Face (`transformer`) 
+                tile = self.transforms(tile, return_tensors="pt")["pixel_values"].squeeze(0)
+            else:  # general callable such as torchvision transforms
+                tile = self.transforms(tile)
         return idx, tile

{slide2vec-1.3.0 → slide2vec-2.0.0}/slide2vec/embed.py

@@ -28,14 +28,20 @@ def get_args_parser(add_help: bool = True):
         "--config-file", default="", metavar="FILE", help="path to config file"
     )
     parser.add_argument(
-        "--run-id",
+        "--output-dir",
         type=str,
-        default="",
-        help="Name of output subdirectory",
+        default=None,
+        help="output directory to save logs and checkpoints",
     )
     parser.add_argument(
         "--run-on-cpu", action="store_true", help="run inference on cpu"
     )
+    parser.add_argument(
+        "opts",
+        help="Modify config options at the end of the command using \"path.key=value\".",
+        default=None,
+        nargs=argparse.REMAINDER,
+    )
     return parser
 
 
@@ -80,7 +86,7 @@ def run_inference(dataloader, model, device, autocast_context, unit, batch_size,
             ):
                 idx, image = batch
                 image = image.to(device, non_blocking=True)
-                feature = model(image).cpu().numpy()
+                feature = model(image)["embedding"].cpu().numpy()
                 features.resize(features.shape[0] + feature.shape[0], axis=0)
                 features[-feature.shape[0]:] = feature
                 indices.resize(indices.shape[0] + idx.shape[0], axis=0)
@@ -123,7 +129,7 @@ def main(args):
     # setup configuration
     run_on_cpu = args.run_on_cpu
     cfg = get_cfg_from_file(args.config_file)
-    output_dir = Path(cfg.output_dir, args.run_id)
+    output_dir = Path(cfg.output_dir, args.output_dir)
     cfg.output_dir = str(output_dir)
 
     if not run_on_cpu:
@@ -148,6 +154,11 @@ def main(args):
         process_list.is_file()
     ), "Process list CSV not found. Ensure tiling has been run."
     process_df = pd.read_csv(process_list)
+    if "feature_status" not in process_df.columns:
+        process_df["feature_status"] = ["tbp"] * len(process_df)
+        cols = ["wsi_name", "wsi_path", "mask_path", "tiling_status", "feature_status", "error", "traceback"]
+        process_df = process_df[cols]
+
     skip_feature_extraction = process_df["feature_status"].str.contains("success").all()
 
     if skip_feature_extraction:
@@ -219,13 +230,15 @@ def main(args):
 
                 name = wsi_fp.stem.replace(" ", "_")
                 feature_path = features_dir / f"{name}.pt"
+                if cfg.model.save_tile_embeddings:
+                    feature_path = features_dir / f"{name}-tiles.pt"
                 tmp_feature_path = tmp_dir / f"{name}-rank_{distributed.get_global_rank()}.h5"
 
                 # get feature dimension and dtype using a dry run
                 with torch.inference_mode(), autocast_context:
                     sample_batch = next(iter(dataloader))
                     sample_image = sample_batch[1].to(model.device)
-                    sample_feature = model(sample_image).cpu().numpy()
+                    sample_feature = model(sample_image)["embedding"].cpu().numpy()
                     feature_dim = sample_feature.shape[1:]
                     dtype = sample_feature.dtype
 

{slide2vec-1.3.0 → slide2vec-2.0.0}/slide2vec/main.py

@@ -24,6 +24,17 @@ def get_args_parser(add_help: bool = True):
     parser.add_argument(
         "--run-on-cpu", action="store_true", help="run inference on cpu"
     )
+    parser.add_argument(
+        "--output-dir",
+        type=str,
+        help="output directory to save logs and checkpoints",
+    )
+    parser.add_argument(
+        "opts",
+        help="Modify config options at the end of the command using \"path.key=value\".",
+        default=None,
+        nargs=argparse.REMAINDER,
+    )
     return parser
 
 
@@ -37,35 +48,26 @@ def log_progress(features_dir: Path, stop_event: threading.Event, log_interval:
         time.sleep(log_interval)
 
 
-def run_tiling(config_file, run_id):
-    print("Running tiling.py...")
+def run_tiling(root_dir, config_file, output_dir):
+    print(f"Running tiling.py from {root_dir}...")
     cmd = [
         sys.executable,
-        "slide2vec/tiling.py",
-        "--run-id",
-        run_id,
+        "hs2p/tiling.py",
         "--config-file",
-        config_file,
+        os.path.abspath(config_file),
+        "--output-dir",
+        os.path.abspath(output_dir),
+        "--skip-datetime",
+        "--skip-logging",
+        "wandb.enable=false", # disable wandb to avoid dupliacte logging
     ]
-    proc = subprocess.Popen(
-        cmd,
-        stdout=subprocess.PIPE,
-        stderr=subprocess.STDOUT,
-        text=True,
-        bufsize=1,
-        universal_newlines=True
-    )
-    # forward output in real-time
-    for line in proc.stdout:
-        print(line.rstrip())
-        sys.stdout.flush()
-    proc.wait()
+    proc = subprocess.run(cmd, cwd=root_dir)
     if proc.returncode != 0:
         print("Slide tiling failed. Exiting.")
         sys.exit(proc.returncode)
 
 
-def run_feature_extraction(config_file, run_id, run_on_cpu: False):
+def run_feature_extraction(config_file, output_dir, run_on_cpu: False):
     print("Running embed.py...")
     # find a free port
     with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
@@ -78,36 +80,24 @@ def run_feature_extraction(config_file, run_id, run_on_cpu: False):
         f"--master_port={free_port}",
         "--nproc_per_node=gpu",
         "slide2vec/embed.py",
-        "--run-id",
-        run_id,
         "--config-file",
-        config_file,
+        os.path.abspath(config_file),
+        "--output-dir",
+        os.path.abspath(output_dir),
     ]
     if run_on_cpu:
         cmd = [
             sys.executable,
             "slide2vec/embed.py",
-            "--run-id",
-            run_id,
             "--config-file",
-            config_file,
+            os.path.abspath(config_file),
+            "--output-dir",
+            os.path.abspath(output_dir),
             "--run-on-cpu",
         ]
     # launch in its own process group.
-    proc = subprocess.Popen(
-        cmd,
-        preexec_fn=os.setsid,
-        stdout=subprocess.PIPE,
-        stderr=subprocess.STDOUT,
-        text=True,
-        bufsize=1,
-        universal_newlines=True
-    )
+    proc = subprocess.Popen(cmd)
     try:
-        # forward output in real-time
-        for line in proc.stdout:
-            print(line.rstrip())
-            sys.stdout.flush()
         proc.wait()
     except KeyboardInterrupt:
         print("Received CTRL+C, terminating embed.py process group...")
@@ -119,34 +109,22 @@ def run_feature_extraction(config_file, run_id, run_on_cpu: False):
         sys.exit(proc.returncode)
 
 
-def run_feature_aggregation(config_file, run_id, run_on_cpu: False):
+def run_feature_aggregation(config_file, output_dir, run_on_cpu: False):
     print("Running aggregate.py...")
     # find a free port
     cmd = [
         sys.executable,
         "slide2vec/aggregate.py",
-        "--run-id",
-        run_id,
         "--config-file",
-        config_file,
+        os.path.abspath(config_file),
+        "--output-dir",
+        os.path.abspath(output_dir),
     ]
     if run_on_cpu:
         cmd.append("--run-on-cpu")
     # launch in its own process group.
-    proc = subprocess.Popen(
-        cmd,
-        preexec_fn=os.setsid,
-        stdout=subprocess.PIPE,
-        stderr=subprocess.STDOUT,
-        text=True,
-        bufsize=1,
-        universal_newlines=True
-    )
+    proc = subprocess.Popen(cmd)
     try:
-        # forward output in real-time
-        for line in proc.stdout:
-            print(line.rstrip())
-            sys.stdout.flush()
         proc.wait()
     except KeyboardInterrupt:
         print("Received CTRL+C, terminating aggregate.py process group...")
@@ -159,19 +137,19 @@ def run_feature_aggregation(config_file, run_id, run_on_cpu: False):
 
 
 def main(args):
-    config_file = args.config_file
-    skip_datetime = args.skip_datetime
     run_on_cpu = args.run_on_cpu
 
-    cfg, run_id = setup(config_file, skip_datetime=skip_datetime)
+    cfg, cfg_path = setup(args)
+    output_dir = Path(cfg.output_dir)
+
     hf_login()
 
-    run_tiling(config_file, run_id)
+    root_dir = "slide2vec/hs2p"
+    run_tiling(root_dir, cfg_path, output_dir)
 
     print("Tiling completed.")
     print("=+=" * 10)
 
-    output_dir = Path(cfg.output_dir)
     features_dir = output_dir / "features"
     if cfg.wandb.enable:
         stop_event = threading.Event()
@@ -180,10 +158,10 @@ def main(args):
         )
         log_thread.start()
 
-    run_feature_extraction(config_file, run_id, run_on_cpu)
+    run_feature_extraction(cfg_path, output_dir, run_on_cpu)
 
     if cfg.model.level == "slide":
-        run_feature_aggregation(config_file, run_id, run_on_cpu)
+        run_feature_aggregation(cfg_path, output_dir, run_on_cpu)
         print("Feature extraction completed.")
         print("=+=" * 10)
     else:
@@ -203,9 +181,9 @@ if __name__ == "__main__":
 
     import warnings
     import torchvision
-    
+
     torchvision.disable_beta_transforms_warning()
-    
+
     warnings.filterwarnings("ignore", message=".*Could not set the permissions.*")
     warnings.filterwarnings("ignore", message=".*antialias.*", category=UserWarning)
     warnings.filterwarnings("ignore", message=".*TypedStorage.*", category=UserWarning)