Fix pip package #55

Author: monatis

clip.cpp

@@ -13,7 +13,7 @@
 #define STB_IMAGE_IMPLEMENTATION
 #include "stb_image.h"
 
-#define CLIP_DEBUG
+// #define CLIP_DEBUG
 
 // utility function for a workaround until https://github.com/ggerganov/ggml/issues/260 is resolved
 // after that, remove this and use the mechanism implemented in GGML directly

examples/python_bindings/README.md

@@ -67,25 +67,27 @@ def tokenize(self, text: str) -> List[int]:
 
 ```python
 def encode_text(
-    self, tokens: List[int], n_threads: int = os.cpu_count()
+    self, tokens: List[int], n_threads: int = os.cpu_count(), normalize: bool = True
 ) -> List[float]:
 ```
 
 - **Description**: Encodes a list of token IDs into a text embedding.
 - `tokens` (List[int]): A list of token IDs obtained through tokenization.
 - `n_threads` (int, optional): The number of CPU threads to use for encoding (default is the number of CPU cores).
+- `normalize` (bool, optional): Whether or not to normalize the output vector (default is `True`).
 
 #### 5. `load_preprocess_encode_image`
 
 ```python
 def load_preprocess_encode_image(
-    self, image_path: str, n_threads: int = os.cpu_count()
+    self, image_path: str, n_threads: int = os.cpu_count(), normalize: bool = True
 ) -> List[float]:
 ```
 
 - **Description**: Loads an image, preprocesses it, and encodes it into an image embedding.
 - `image_path` (str): The path to the image file to be encoded.
 - `n_threads` (int, optional): The number of CPU threads to use for encoding (default is the number of CPU cores).
+- `normalize` (bool, optional): Whether or not to normalize the output vector (default is `True`).
 
 #### 6. `calculate_similarity`
 

examples/python_bindings/clip_cpp/clip.py

@@ -4,12 +4,15 @@
 
 # Note: Pass -DBUILD_SHARED_LIBS=ON to cmake to create the shared library file
 
-# Load the shared library
-path_to_dll = os.environ.get(
-    "CLIP_DLL", os.path.join(os.path.abspath(os.path.dirname(__file__)), "libclip.so")
-)
+cur_dir = os.getcwd()
+this_dir = os.path.abspath(os.path.dirname(__file__))
 
-clip_lib = ctypes.CDLL(path_to_dll)
+# Load the shared library
+path_to_dll = os.environ.get("CLIP_DLL", this_dir)
+os.chdir(path_to_dll)
+ggml_lib = ctypes.CDLL("./libggml.so")
+clip_lib = ctypes.CDLL("./libclip.so")
+os.chdir(cur_dir)
 
 
 # Define the ctypes structures
@@ -278,7 +281,7 @@ def encode_text(
         return [txt_vec[i] for i in range(self.vec_dim)]
 
     def load_preprocess_encode_image(
-        self, image_path: str, n_threads: int = os.cpu_count(), normalize: bool = true
+        self, image_path: str, n_threads: int = os.cpu_count(), normalize: bool = True
     ) -> List[float]:
         image_ptr = make_clip_image_u8()
         if not clip_image_load_from_file(image_path.encode("utf8"), image_ptr):

examples/python_bindings/pyproject.toml

@@ -1,13 +1,13 @@
 [tool.poetry]
 name = "clip_cpp"
-version = "0.2.0"
+version = "0.3.5"
 description = "CLIP inference with no big dependencies as PyTorch, TensorFlow, Numpy"
 authors = ["Yusuf Sarıgöz <[email protected]>"]
 packages = [
     {include = "clip_cpp"},
 ]
 include = [
-{path = "clip_cpp/libclip.so"}
+{path = "clip_cpp/*.so"}
 ]
 
 readme = "README.md"