how about this

app.py

@@ -1,39 +0,0 @@
-import streamlit as st
-import huggingface_hub as hf_hub
-import monai
-import os
-import zipfile
-import torch
-
-hf_hub.login(token=st.secrets["HF_TOKEN"])
-
-with st.spinner("Downloading Dataset"):
-    data_path = hf_hub.hf_hub_download(repo_id="osbm/prostate158", filename="data.zip", repo_type="dataset")
-
-st.write(data_path)
-with st.spinner("Unzipping..."):
-    with zipfile.ZipFile(data_path, 'r') as zip_ref:
-        zip_ref.extractall(".")
-
-
-# st.write(os.listdir(os.getcwd()))
-# st.write(os.getcwd())
-
-
-model = monai.networks.nets.UNet(
-    in_channels=1,
-    out_channels=3,
-    spatial_dims=3,
-    channels=[16, 32, 64, 128, 256, 512],
-    strides=[2, 2, 2, 2, 2],
-    num_res_units=4,
-    act="PRELU",
-    norm="BATCH",
-    dropout=0.15,
-)
-
-# load this model using anatomy.pt 
-model.load_state_dict(torch.load('anatomy.pt', map_location=torch.device('cpu')))
-
-print(model)
-

inference.py

@@ -43,7 +43,32 @@ from monai.transforms import (
 
 # model.load_state_dict(torch.load("anatomy.pt", map_location=device))
 
-keys = ("t2", "t2_anatomy_reader1")
+# keys = ("t2", "t2_anatomy_reader1")
+# transforms = Compose(
+#     [
+#         LoadImaged(keys=keys, image_only=False),
+#         EnsureChannelFirstd(keys=keys),
+#         Spacingd(keys=keys, pixdim=[0.5, 0.5, 0.5], mode=("bilinear", "nearest")),
+#         Orientationd(keys=keys, axcodes="RAS"),
+#         ScaleIntensityd(keys=keys, minv=0, maxv=1),
+#         NormalizeIntensityd(keys=keys),
+#         EnsureTyped(keys=keys),
+#         ConcatItemsd(keys=("t2"), name=CommonKeys.IMAGE, dim=0),
+#         ConcatItemsd(keys=("t2_anatomy_reader1"), name=CommonKeys.LABEL, dim=0),
+#     ],
+# )
+
+# postprocessing = Compose(
+#     [
+#         EnsureTyped(keys=[CommonKeys.PRED, CommonKeys.LABEL]),
+#         KeepLargestConnectedComponentd(
+#             keys=CommonKeys.PRED, 
+#             applied_labels=list(range(1, 3))
+#         ),
+#     ],
+# )
+
+keys = ("t2")
 transforms = Compose(
     [
         LoadImaged(keys=keys, image_only=False),
@@ -54,24 +79,21 @@ transforms = Compose(
         NormalizeIntensityd(keys=keys),
         EnsureTyped(keys=keys),
         ConcatItemsd(keys=("t2"), name=CommonKeys.IMAGE, dim=0),
-        ConcatItemsd(keys=("t2_anatomy_reader1"), name=CommonKeys.LABEL, dim=0),
     ],
-    allow_missing_keys=True,
 )
 
-
-
-
 postprocessing = Compose(
     [
-        EnsureTyped(keys=[CommonKeys.PRED, CommonKeys.LABEL]),
+        EnsureTyped(keys=[CommonKeys.PRED]),
         KeepLargestConnectedComponentd(
             keys=CommonKeys.PRED, 
             applied_labels=list(range(1, 3))
         ),
     ],
-    allow_missing_keys=True,
 )
+
+
+
 inferer = monai.inferers.SlidingWindowInferer(
     roi_size=(96, 96, 96),
     sw_batch_size=4,