spacr 0.3.46__tar.gz → 0.3.50__tar.gz

{spacr-0.3.46/spacr.egg-info → spacr-0.3.50}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: spacr
-Version: 0.3.46
+Version: 0.3.50
 Summary: Spatial phenotype analysis of crisp screens (SpaCr)
 Home-page: https://github.com/EinarOlafsson/spacr
 Author: Einar Birnir Olafsson
@@ -66,6 +66,7 @@ Requires-Dist: gdown
 Requires-Dist: IPython<9.0,>=8.18.1
 Requires-Dist: ipykernel
 Requires-Dist: ipywidgets<9.0,>=8.1.2
+Requires-Dist: brokenaxes<1.0,>=0.6.2
 Requires-Dist: huggingface-hub<0.25,>=0.24.0
 Provides-Extra: dev
 Requires-Dist: pytest<3.11,>=3.9; extra == "dev"

{spacr-0.3.46 → spacr-0.3.50}/setup.py

@@ -62,12 +62,13 @@ dependencies = [
     'IPython>=8.18.1,<9.0',
     'ipykernel',
     'ipywidgets>=8.1.2,<9.0',
+    'brokenaxes>=0.6.2,<1.0',
     'huggingface-hub>=0.24.0,<0.25'
 ]
 
 setup(
     name="spacr",
-    version="0.3.46",
+    version="0.3.50",
     author="Einar Birnir Olafsson",
     author_email="olafsson@med.umich.com",
     description="Spatial phenotype analysis of crisp screens (SpaCr)",

spacr-0.3.50/spacr/chat_bot.py

@@ -0,0 +1,31 @@
+import os
+import torch
+from multiprocessing import Queue
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+# Disable parallelism warnings
+os.environ["TOKENIZERS_PARALLELISM"] = "false"
+
+def load_model(model_queue):
+    """Load the Falcon-7B model and send the status via a queue."""
+    try:
+        print("Loading Falcon-7B model...")
+        tokenizer = AutoTokenizer.from_pretrained("tiiuae/falcon-7b")
+
+        # Set `pad_token` to `eos_token` if not already set
+        if tokenizer.pad_token is None:
+            tokenizer.pad_token = tokenizer.eos_token
+
+        model = AutoModelForCausalLM.from_pretrained("tiiuae/falcon-7b")
+        model.config.pad_token_id = tokenizer.pad_token_id
+
+        # Send the loaded model and tokenizer back via the queue
+        model_queue.put((model, tokenizer, "Model loaded successfully!"))
+    except Exception as e:
+        model_queue.put((None, None, f"Error loading model: {e}"))
+
+def language_chat(tokenizer, model, user_input):
+    inputs = tokenizer(user_input, return_tensors="pt", padding=True, truncation=True)
+    outputs = model.generate(inputs.input_ids, max_length=256, pad_token_id=tokenizer.eos_token_id)
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
+

{spacr-0.3.46 → spacr-0.3.50}/spacr/gui_elements.py

@@ -16,7 +16,8 @@ from collections import deque
 from skimage.draw import polygon, line
 from skimage.transform import resize
 from scipy.ndimage import binary_fill_holes, label
-from tkinter import ttk, scrolledtext 
+from tkinter import ttk, scrolledtext
+from skimage.color import rgb2gray 
 
 fig = None
 
@@ -1534,20 +1535,45 @@ class ModifyMaskApp:
         tk.Label(self.zoom_toolbar, text="Upper Percentile:", bg='black', fg='white').pack(side='left')
         self.upper_entry = tk.Entry(self.zoom_toolbar, textvariable=self.upper_quantile, bg='black', fg='white')
         self.upper_entry.pack(side='left')
-
-        
+    
     def load_image_and_mask(self, index):
+        # Load the image
         image_path = os.path.join(self.folder_path, self.image_filenames[index])
-        image = imageio.imread(image_path)        
+        image = imageio.imread(image_path)
+        print(f"Original Image shape: {image.shape}, dtype: {image.dtype}")
+
+        # Handle multi-channel or transparency issues
+        if image.ndim == 3:
+            if image.shape[2] == 4:  # If the image has an alpha channel (RGBA)
+                image = image[..., :3]  # Remove the alpha channel
+
+            # Convert RGB to grayscale using weighted average
+            image = np.dot(image[..., :3], [0.2989, 0.5870, 0.1140]).astype(np.uint8)
+            print(f"Converted to grayscale: {image.shape}")
+
+        # Ensure the shape is (height, width) without extra channel
+        if image.ndim == 3 and image.shape[2] == 1:
+            image = np.squeeze(image, axis=-1)
+
+        if image.dtype != np.uint16:
+            # Scale the image to fit the 16-bit range (0–65535)
+            image = (image / image.max() * 65535).astype(np.uint16)
+            # eventually remove this images should not have to be 16 bit look into downstream function (non 16bit images are jsut black)
+
+        # Load the corresponding mask
         mask_path = os.path.join(self.masks_folder, self.image_filenames[index])
         if os.path.exists(mask_path):
-            print(f'loading mask:{mask_path} for image: {image_path}')
+            print(f'Loading mask: {mask_path} for image: {image_path}')
             mask = imageio.imread(mask_path)
+
+            # Ensure mask is uint8
             if mask.dtype != np.uint8:
-                mask = (mask / np.max(mask) * 255).astype(np.uint8)
+                mask = (mask / mask.max() * 255).astype(np.uint8)
         else:
+            # Create a new mask with the same size as the image
             mask = np.zeros(image.shape[:2], dtype=np.uint8)
-            print(f'loaded new mask for image: {image_path}')
+            print(f'Loaded new mask for image: {image_path}')
+
         return image, mask
     
     ####################################################################################################