Improve test coverage and refactor BaseModel (#26)

* Update `available_feature_keys` and `available_weights` * Fix serialization * Update version * Update `README`
james77777778 · Jan 21, 2024 · 4506330 · 4506330
1 parent c608f1c
commit 4506330
Show file tree

Hide file tree

Showing 20 changed files with 1,083 additions and 667 deletions.
diff --git a/README.md b/README.md
@@ -29,9 +29,9 @@ pip install keras kimm
 [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/14WxYgVjlwCIO9MwqPYW-dskbTL2UHsVN?usp=sharing)
 
 ```python
-import cv2
 import keras
 from keras import ops
+from keras import utils
 from keras.applications.imagenet_utils import decode_predictions
 
 import kimm
@@ -43,15 +43,15 @@ print(kimm.list_models())
 print(kimm.list_models("efficientnet", weights="imagenet"))  # fuzzy search
 
 # Initialize the model with pretrained weights
-model = kimm.models.EfficientNetV2B0()
-image_size = model._default_size
+model = kimm.models.VisionTransformerTiny16()
+image_size = (model._default_size, model._default_size)
 
 # Load an image as the model input
 image_path = keras.utils.get_file(
     "african_elephant.jpg", "https://i.imgur.com/Bvro0YD.png"
 )
-image = cv2.imread(image_path)
-image = cv2.resize(image, (image_size, image_size))
+image = utils.load_img(image_path, target_size=image_size)
+image = utils.img_to_array(image)
 x = ops.convert_to_tensor(image)
 x = ops.expand_dims(x, axis=0)
 
@@ -62,9 +62,9 @@ print("Predicted:", decode_predictions(preds, top=3)[0])
 
 ```bash
 ['ConvMixer1024D20', 'ConvMixer1536D20', 'ConvMixer736D32', 'ConvNeXtAtto', ...]
-['EfficientNetB0', 'EfficientNetB1', 'EfficientNetB2', 'EfficientNetB3', ...]
-1/1 ━━━━━━━━━━━━━━━━━━━━ 11s 11s/step
-Predicted: [('n02504458', 'African_elephant', 0.90578836), ('n01871265', 'tusker', 0.024864597), ('n02504013', 'Indian_elephant', 0.01161992)]
+['VisionTransformerBase16', 'VisionTransformerBase32', 'VisionTransformerSmall16', ...]
+1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step
+Predicted: [('n02504458', 'African_elephant', 0.6895825), ('n01871265', 'tusker', 0.17934209), ('n02504013', 'Indian_elephant', 0.12927249)]
 ```
 
 ### An end-to-end example: fine-tuning an image classification model on a cats vs. dogs dataset

diff --git a/kimm/__init__.py b/kimm/__init__.py
@@ -1,4 +1,4 @@
 from kimm import models  # force to add models to the registry
 from kimm.utils.model_registry import list_models
 
-__version__ = "0.1.2"
+__version__ = "0.1.3"
diff --git a/kimm/models/base_model.py b/kimm/models/base_model.py
@@ -1,4 +1,3 @@
-import abc
 import pathlib
 import typing
 import urllib.parse
@@ -12,6 +11,12 @@
 
 
 class BaseModel(models.Model):
+    default_origin = (
+        "https://github.com/james77777778/kimm/releases/download/0.1.0/"
+    )
+    available_feature_keys = []
+    available_weights = []
+
     def __init__(
         self,
         inputs,
@@ -183,12 +188,6 @@ def load_pretrained_weights(self, weights_url: typing.Optional[str] = None):
             )
             self.load_weights(weights_path)
 
-    @staticmethod
-    @abc.abstractmethod
-    def available_feature_keys():
-        # TODO: add docstring
-        raise NotImplementedError
-
     def get_config(self):
         # Don't chain to super here. The default `get_config()` for functional
         # models is nested and cannot be passed to BaseModel.
@@ -215,6 +214,19 @@ def get_config(self):
     def fix_config(self, config: typing.Dict):
         return config
 
-    @property
-    def default_origin(self):
-        return "https://github.com/james77777778/kimm/releases/download/0.1.0/"
+    def get_weights_url(self, weights):
+        if weights is None:
+            return None
+
+        for _weights, _origin, _file_name in self.available_weights:
+            if weights == _weights:
+                return f"{_origin}/{_file_name}"
+
+        # Failed to find the weights
+        _available_weights_name = [
+            _weights for _weights, _ in self.available_weights
+        ]
+        raise ValueError(
+            f"Available weights are {_available_weights_name}. "
+            f"Received weights={weights}"
+        )
diff --git a/kimm/models/convmixer.py b/kimm/models/convmixer.py
@@ -52,6 +52,8 @@ def __init__(
         activation: str = "relu",
         **kwargs,
     ):
+        kwargs["weights_url"] = self.get_weights_url(kwargs["weights"])
+
         input_tensor = kwargs.pop("input_tensor", None)
         self.set_properties(kwargs)
         inputs = self.determine_input_tensor(
@@ -100,10 +102,6 @@ def __init__(
         self.kernel_size = kernel_size
         self.activation = activation
 
-    @staticmethod
-    def available_feature_keys():
-        raise NotImplementedError
-
     def get_config(self):
         config = super().get_config()
         config.update(
@@ -136,6 +134,15 @@ def fix_config(self, config):
 
 
 class ConvMixer736D32(ConvMixer):
+    available_feature_keys = ["STEM", *[f"BLOCK{i}" for i in range(32)]]
+    available_weights = [
+        (
+            "imagenet",
+            ConvMixer.default_origin,
+            "convmixer736d32_convmixer_768_32.in1k.keras",
+        )
+    ]
+
     def __init__(
         self,
         input_tensor: keras.KerasTensor = None,
@@ -151,9 +158,6 @@ def __init__(
         **kwargs,
     ):
         kwargs = self.fix_config(kwargs)
-        if weights == "imagenet":
-            file_name = "convmixer736d32_convmixer_768_32.in1k.keras"
-            kwargs["weights_url"] = f"{self.default_origin}/{file_name}"
         super().__init__(
             32,
             768,
@@ -173,14 +177,17 @@ def __init__(
             **kwargs,
         )
 
-    @staticmethod
-    def available_feature_keys():
-        feature_keys = ["STEM"]
-        feature_keys.extend([f"BLOCK{i}" for i in range(32)])
-        return feature_keys
-
 
 class ConvMixer1024D20(ConvMixer):
+    available_feature_keys = ["STEM", *[f"BLOCK{i}" for i in range(20)]]
+    available_weights = [
+        (
+            "imagenet",
+            ConvMixer.default_origin,
+            "convmixer1024d20_convmixer_1024_20_ks9_p14.in1k.keras",
+        )
+    ]
+
     def __init__(
         self,
         input_tensor: keras.KerasTensor = None,
@@ -196,9 +203,6 @@ def __init__(
         **kwargs,
     ):
         kwargs = self.fix_config(kwargs)
-        if weights == "imagenet":
-            file_name = "convmixer1024d20_convmixer_1024_20_ks9_p14.in1k.keras"
-            kwargs["weights_url"] = f"{self.default_origin}/{file_name}"
         super().__init__(
             20,
             1024,
@@ -218,14 +222,17 @@ def __init__(
             **kwargs,
         )
 
-    @staticmethod
-    def available_feature_keys():
-        feature_keys = ["STEM"]
-        feature_keys.extend([f"BLOCK{i}" for i in range(20)])
-        return feature_keys
-
 
 class ConvMixer1536D20(ConvMixer):
+    available_feature_keys = ["STEM", *[f"BLOCK{i}" for i in range(20)]]
+    available_weights = [
+        (
+            "imagenet",
+            ConvMixer.default_origin,
+            "convmixer1536d20_convmixer_1536_20.in1k.keras",
+        )
+    ]
+
     def __init__(
         self,
         input_tensor: keras.KerasTensor = None,
@@ -241,9 +248,6 @@ def __init__(
         **kwargs,
     ):
         kwargs = self.fix_config(kwargs)
-        if weights == "imagenet":
-            file_name = "convmixer1536d20_convmixer_1536_20.in1k.keras"
-            kwargs["weights_url"] = f"{self.default_origin}/{file_name}"
         super().__init__(
             20,
             1536,
@@ -263,12 +267,6 @@ def __init__(
             **kwargs,
         )
 
-    @staticmethod
-    def available_feature_keys():
-        feature_keys = ["STEM"]
-        feature_keys.extend([f"BLOCK{i}" for i in range(20)])
-        return feature_keys
-
 
 add_model_to_registry(ConvMixer736D32, "imagenet")
 add_model_to_registry(ConvMixer1024D20, "imagenet")