add jpeg quality option

docs/source/preparing_datasets/basic_dataset_conversion.md

@@ -43,7 +43,7 @@ out = ('/local/data', 'oci://bucket/data')
 | Numerical String   | 'str_float'   | `StrFloat`   | stores in UTF-8          |
 | Numerical String   | 'str_decimal' | `StrDecimal` | stores in UTF-8          |
 | Image              | 'pil'         | `PIL`        | raw PIL image class ([link]((https://pillow.readthedocs.io/en/stable/reference/Image.html)))            |
-| Image              | 'jpeg'        | `JPEG`       | PIL image as JPEG        |
+| Image              | 'jpeg:quality'        | `JPEG`       | PIL image as JPEG, quality between 0 and 100        |
 | Image              | 'png'         | `PNG`        | PIL image as PNG         |
 | Pickle             | 'pkl'         | `Pickle`     | arbitrary Python objects |
 | JSON               | 'json'        | `JSON`       | arbitrary data as JSON   |
@@ -52,7 +52,7 @@ Here's an example where the field `x` is an image, and `y` is a class label, as
 <!--pytest.mark.skip-->
 ```python
 column = {
-    'x': 'jpeg',
+    'x': 'jpeg:90',
     'y': 'int8',
 }
 ```

streaming/base/format/mds/encodings.py

@@ -466,6 +466,28 @@ def decode(self, data: bytes) -> Image.Image:
 class JPEG(Encoding):
     """Store PIL image as JPEG."""
 
+    def __init__(self, quality: int = 75):
+        assert 0 <= quality <= 100
+        self.quality = quality
+
+    @classmethod
+    def from_str(cls, text: str) -> Self:
+        """Parse this encoding from string.
+
+        Args:
+            text (str): The string to parse.
+
+        Returns:
+            Self: The initialized Encoding.
+        """
+        args = text.split(':') if text else []
+        assert len(args) in {0, 1}
+        if len(args) == 1:
+            quality = int(args[0])
+        else:
+            quality = 75
+        return cls(quality)
+
     def encode(self, obj: Image.Image) -> bytes:
         self._validate(obj, Image.Image)
         if isinstance(obj, JpegImageFile) and hasattr(obj, 'filename'):
@@ -474,7 +496,7 @@ def encode(self, obj: Image.Image) -> bytes:
                 return f.read()
         else:
             out = BytesIO()
-            obj.save(out, format='JPEG')
+            obj.save(out, format='JPEG', quality=self.quality)
             return out.getvalue()
 
     def decode(self, data: bytes) -> Image.Image: