Merge pull request #26 from full-stack-deep-learning/lab9

Lab 9

Author: sergeyk

.gitignore

@@ -45,3 +45,4 @@ logs
 .mypy_cache
 notebooks/lightning_logs
 lightning_logs/
+lab9/requirements.txt

lab1/text_recognizer/util.py

@@ -1,14 +1,15 @@
 """Utility functions for text_recognizer module."""
+from io import BytesIO
 from pathlib import Path
 from typing import Union
 from urllib.request import urlretrieve
-
-# import base64
+import base64
 import hashlib
 
 from PIL import Image
 from tqdm import tqdm
 import numpy as np
+import smart_open
 
 
 def to_categorical(y, num_classes):
@@ -17,7 +18,12 @@ def to_categorical(y, num_classes):
 
 
 def read_image_pil(image_uri: Union[Path, str], grayscale=False) -> Image:
-    with Image.open(image_uri) as image:
+    with smart_open.open(image_uri, "rb") as image_file:
+        return read_image_pil_file(image_file, grayscale)
+
+
+def read_image_pil_file(image_file, grayscale=False) -> Image:
+    with Image.open(image_file) as image:
         if grayscale:
             image = image.convert(mode="L")
         else:

lab2/text_recognizer/util.py

@@ -1,14 +1,15 @@
 """Utility functions for text_recognizer module."""
+from io import BytesIO
 from pathlib import Path
 from typing import Union
 from urllib.request import urlretrieve
-
-# import base64
+import base64
 import hashlib
 
 from PIL import Image
 from tqdm import tqdm
 import numpy as np
+import smart_open
 
 
 def to_categorical(y, num_classes):
@@ -17,7 +18,12 @@ def to_categorical(y, num_classes):
 
 
 def read_image_pil(image_uri: Union[Path, str], grayscale=False) -> Image:
-    with Image.open(image_uri) as image:
+    with smart_open.open(image_uri, "rb") as image_file:
+        return read_image_pil_file(image_file, grayscale)
+
+
+def read_image_pil_file(image_file, grayscale=False) -> Image:
+    with Image.open(image_file) as image:
         if grayscale:
             image = image.convert(mode="L")
         else:

lab3/text_recognizer/util.py

@@ -1,14 +1,15 @@
 """Utility functions for text_recognizer module."""
+from io import BytesIO
 from pathlib import Path
 from typing import Union
 from urllib.request import urlretrieve
-
-# import base64
+import base64
 import hashlib
 
 from PIL import Image
 from tqdm import tqdm
 import numpy as np
+import smart_open
 
 
 def to_categorical(y, num_classes):
@@ -17,7 +18,12 @@ def to_categorical(y, num_classes):
 
 
 def read_image_pil(image_uri: Union[Path, str], grayscale=False) -> Image:
-    with Image.open(image_uri) as image:
+    with smart_open.open(image_uri, "rb") as image_file:
+        return read_image_pil_file(image_file, grayscale)
+
+
+def read_image_pil_file(image_file, grayscale=False) -> Image:
+    with Image.open(image_file) as image:
         if grayscale:
             image = image.convert(mode="L")
         else:

lab4/text_recognizer/lit_models/transformer.py

@@ -1,5 +1,9 @@
 import torch.nn as nn
-import wandb
+try:
+    import wandb
+except ModuleNotFoundError:
+    pass
+
 
 from .metrics import CharacterErrorRate
 from .base import BaseLitModel

lab4/text_recognizer/util.py

@@ -1,14 +1,15 @@
 """Utility functions for text_recognizer module."""
+from io import BytesIO
 from pathlib import Path
 from typing import Union
 from urllib.request import urlretrieve
-
-# import base64
+import base64
 import hashlib
 
 from PIL import Image
 from tqdm import tqdm
 import numpy as np
+import smart_open
 
 
 def to_categorical(y, num_classes):
@@ -17,7 +18,12 @@ def to_categorical(y, num_classes):
 
 
 def read_image_pil(image_uri: Union[Path, str], grayscale=False) -> Image:
-    with Image.open(image_uri) as image:
+    with smart_open.open(image_uri, "rb") as image_file:
+        return read_image_pil_file(image_file, grayscale)
+
+
+def read_image_pil_file(image_file, grayscale=False) -> Image:
+    with Image.open(image_file) as image:
         if grayscale:
             image = image.convert(mode="L")
         else:

lab5/text_recognizer/lit_models/transformer.py

@@ -1,5 +1,9 @@
 import torch.nn as nn
-import wandb
+try:
+    import wandb
+except ModuleNotFoundError:
+    pass
+
 
 from .metrics import CharacterErrorRate
 from .base import BaseLitModel

lab5/text_recognizer/util.py

@@ -1,14 +1,15 @@
 """Utility functions for text_recognizer module."""
+from io import BytesIO
 from pathlib import Path
 from typing import Union
 from urllib.request import urlretrieve
-
-# import base64
+import base64
 import hashlib
 
 from PIL import Image
 from tqdm import tqdm
 import numpy as np
+import smart_open
 
 
 def to_categorical(y, num_classes):
@@ -17,7 +18,12 @@ def to_categorical(y, num_classes):
 
 
 def read_image_pil(image_uri: Union[Path, str], grayscale=False) -> Image:
-    with Image.open(image_uri) as image:
+    with smart_open.open(image_uri, "rb") as image_file:
+        return read_image_pil_file(image_file, grayscale)
+
+
+def read_image_pil_file(image_file, grayscale=False) -> Image:
+    with Image.open(image_file) as image:
         if grayscale:
             image = image.convert(mode="L")
         else:

lab7/text_recognizer/lit_models/transformer.py

@@ -1,5 +1,9 @@
 import torch.nn as nn
-import wandb
+try:
+    import wandb
+except ModuleNotFoundError:
+    pass
+
 
 from .metrics import CharacterErrorRate
 from .base import BaseLitModel

lab7/text_recognizer/models/resnet_transformer.py

@@ -173,7 +173,7 @@ def predict(self, x: torch.Tensor) -> torch.Tensor:
             y = output_tokens[:, :Sy]  # (B, Sy)
             output = self.decode(x, y)  # (Sy, B, C)
             output = torch.argmax(output, dim=-1)  # (Sy, B)
-            output_tokens[:, Sy] = output[-1:]  # Set the last output token
+            output_tokens[:, Sy : Sy + 1] = output[-1:]  # Set the last output token
 
             # Early stopping of prediction loop to speed up prediction
             if ((output_tokens[:, Sy] == self.end_token) | (output_tokens[:, Sy] == self.padding_token)).all():