sparrow-parse 0.4.3__py3-none-any.whl → 0.4.5__py3-none-any.whl

sparrow_parse/__init__.py

@@ -1 +1 @@
-__version__ = '0.4.3'
+__version__ = '0.4.5'

sparrow_parse/extractors/vllm_extractor.py

@@ -7,7 +7,6 @@ from rich import print
 import os
 import tempfile
 import shutil
-from typing import Any, Dict, List, Union
 
 
 class VLLMExtractor(object):
@@ -153,34 +152,34 @@ if __name__ == "__main__":
 
     extractor = VLLMExtractor()
 
-    # # export HF_TOKEN="hf_"
-    # config = {
-    #     "method": "mlx",  # Could be 'huggingface', 'mlx' or 'local_gpu'
-    #     "model_name": "mlx-community/Qwen2-VL-72B-Instruct-4bit",
-    #     # "hf_space": "katanaml/sparrow-qwen2-vl-7b",
-    #     # "hf_token": os.getenv('HF_TOKEN'),
-    #     # Additional fields for local GPU inference
-    #     # "device": "cuda", "model_path": "model.pth"
-    # }
-    #
-    # # Use the factory to get the correct instance
-    # factory = InferenceFactory(config)
-    # model_inference_instance = factory.get_inference_instance()
-    #
-    # input_data = [
-    #     {
-    #         "file_path": "/Users/andrejb/Work/katana-git/sparrow/sparrow-ml/llm/data/invoice_1.jpg",
-    #         "text_input": "retrieve document data. return response in JSON format"
-    #     }
-    # ]
-    #
-    # # Now you can run inference without knowing which implementation is used
-    # results_array, num_pages = extractor.run_inference(model_inference_instance, input_data, tables_only=True,
-    #                                                    generic_query=False,
-    #                                                    debug_dir="/Users/andrejb/Work/katana-git/sparrow/sparrow-ml/llm/data/",
-    #                                                    debug=True,
-    #                                                    mode=None)
-    #
-    # for i, result in enumerate(results_array):
-    #     print(f"Result for page {i + 1}:", result)
-    # print(f"Number of pages: {num_pages}")
+    # export HF_TOKEN="hf_"
+    config = {
+        "method": "mlx",  # Could be 'huggingface', 'mlx' or 'local_gpu'
+        "model_name": "mlx-community/Qwen2-VL-7B-Instruct-8bit",
+        # "hf_space": "katanaml/sparrow-qwen2-vl-7b",
+        # "hf_token": os.getenv('HF_TOKEN'),
+        # Additional fields for local GPU inference
+        # "device": "cuda", "model_path": "model.pth"
+    }
+
+    # Use the factory to get the correct instance
+    factory = InferenceFactory(config)
+    model_inference_instance = factory.get_inference_instance()
+
+    input_data = [
+        {
+            "file_path": "/Users/andrejb/Work/katana-git/sparrow/sparrow-ml/llm/data/invoice_1.jpg",
+            "text_input": "retrieve document data. return response in JSON format"
+        }
+    ]
+
+    # Now you can run inference without knowing which implementation is used
+    results_array, num_pages = extractor.run_inference(model_inference_instance, input_data, tables_only=True,
+                                                       generic_query=False,
+                                                       debug_dir="/Users/andrejb/Work/katana-git/sparrow/sparrow-ml/llm/data/",
+                                                       debug=True,
+                                                       mode=None)
+
+    for i, result in enumerate(results_array):
+        print(f"Result for page {i + 1}:", result)
+    print(f"Number of pages: {num_pages}")

sparrow_parse/vllm/mlx_inference.py

@@ -4,6 +4,7 @@ from mlx_vlm.utils import load_image
 from sparrow_parse.vllm.inference_base import ModelInference
 import os
 import json
+import gc
 
 
 class MLXInference(ModelInference):
@@ -19,18 +20,47 @@ class MLXInference(ModelInference):
         :param model_name: Name of the model to load.
         """
         self.model_name = model_name
+        self.model = None
+        self.processor = None
         print(f"MLXInference initialized with model: {model_name}")
 
 
-    @staticmethod
-    def _load_model_and_processor(model_name):
+    def __del__(self):
         """
-        Load the model and processor for inference.
+        Destructor to clean up resources when the object is deleted.
+        """
+        self.unload_model()
+
+    def unload_model(self):
+        """
+        Unload the model and release resources.
+        """
+        if self.model:
+            print(f"Unloading model: {self.model_name}")
+            del self.model
+            self.model = None
+        if self.processor:
+            print(f"Unloading processor for model: {self.model_name}")
+            del self.processor
+            self.processor = None
+
+        # Force garbage collection to release memory
+        gc.collect()
+        print(f"Model {self.model_name} and its resources have been unloaded, memory cleared.")
 
+
+    def _load_model_and_processor(self, model_name):
+        """
+        Load the model and processor for inference.
         :param model_name: Name of the model to load.
         :return: Tuple containing the loaded model and processor.
         """
-        return load(model_name)
+        print(f"Loading model and processor for: {model_name}...")
+        model, processor = load(model_name)
+        self.model = model  # Store model instance
+        self.processor = processor  # Store processor instance
+        print(f"Model and processor for '{model_name}' loaded successfully.")
+        return model, processor
 
 
     def process_response(self, output_text):
@@ -79,48 +109,54 @@ class MLXInference(ModelInference):
     def inference(self, input_data, mode=None):
         """
         Perform inference on input data using the specified model.
-
         :param input_data: A list of dictionaries containing image file paths and text inputs.
         :param mode: Optional mode for inference ("static" for simple JSON output).
         :return: List of processed model responses.
         """
-        if mode == "static":
-            return [self.get_simple_json()]
-
-        # Load the model and processor
-        model, processor = self._load_model_and_processor(self.model_name)
-        config = model.config
-
-        # Prepare absolute file paths
-        file_paths = self._extract_file_paths(input_data)
-
-        results = []
-        for file_path in file_paths:
-            image, width, height = self.load_image_data(file_path)
-
-            # Prepare messages for the chat model
-            messages = [
-                {"role": "system", "content": "You are an expert at extracting structured text from image documents."},
-                {"role": "user", "content": input_data[0]["text_input"]},
-            ]
-
-            # Generate and process response
-            prompt = apply_chat_template(processor, config, messages)  # Assuming defined
-            response = generate(
-                model,
-                processor,
-                image,
-                prompt,
-                resize_shape=(width, height),
-                max_tokens=4000,
-                temperature=0.0,
-                verbose=False
-            )
-            results.append(self.process_response(response))
-
-            print("Inference completed successfully for: ", file_path)
+        try:
+            if mode == "static":
+                return [self.get_simple_json()]
+
+            # Load the model and processor
+            model, processor = self._load_model_and_processor(self.model_name)
+            config = model.config
+
+            # Prepare absolute file paths
+            file_paths = self._extract_file_paths(input_data)
+
+            results = []
+            for file_path in file_paths:
+                image, width, height = self.load_image_data(file_path)
+
+                # Prepare messages for the chat model
+                messages = [
+                    {"role": "system",
+                     "content": "You are an expert at extracting structured text from image documents."},
+                    {"role": "user", "content": input_data[0]["text_input"]},
+                ]
+
+                # Generate and process response
+                prompt = apply_chat_template(processor, config, messages)
+                response = generate(
+                    model,
+                    processor,
+                    image,
+                    prompt,
+                    resize_shape=(width, height),
+                    max_tokens=4000,
+                    temperature=0.0,
+                    verbose=False
+                )
+                results.append(self.process_response(response))
+
+                print("Inference completed successfully for: ", file_path)
+
+            return results
+
+        finally:
+            # Always unload the model after inference
+            self.unload_model()
 
-        return results
 
     @staticmethod
     def _extract_file_paths(input_data):

{sparrow_parse-0.4.3.dist-info → sparrow_parse-0.4.5.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: sparrow-parse
-Version: 0.4.3
+Version: 0.4.5
 Summary: Sparrow Parse is a Python package (part of Sparrow) for parsing and extracting information from documents.
 Home-page: https://github.com/katanaml/sparrow/tree/main/sparrow-data/parse
 Author: Andrej Baranovskij

{sparrow_parse-0.4.3.dist-info → sparrow_parse-0.4.5.dist-info}/RECORD

@@ -1,7 +1,7 @@
-sparrow_parse/__init__.py,sha256=udnlByVnFcZDwWir50pEbTU0bIwgBrpNtAiVExFEzu0,21
+sparrow_parse/__init__.py,sha256=aYJpblaKx7C4NugypZ2lZHBzsGsZoG5sUbcsPwgoi9s,21
 sparrow_parse/__main__.py,sha256=Xs1bpJV0n08KWOoQE34FBYn6EBXZA9HIYJKrE4ZdG78,153
 sparrow_parse/extractors/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-sparrow_parse/extractors/vllm_extractor.py,sha256=ybWpRpDH0YHoYpHkjIJtm7DQoHJBKNsirK2YIAlMvGo,7863
+sparrow_parse/extractors/vllm_extractor.py,sha256=G4kQh0GoZ4V4TdyeDwZWFkOG15MQCQMvSf2UbFQDWeI,7746
 sparrow_parse/helpers/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 sparrow_parse/helpers/pdf_optimizer.py,sha256=GIqQYWtixFeZGCRFXL0lQfQByapCDuQzzRHAkzcPwLE,3302
 sparrow_parse/processors/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -11,9 +11,9 @@ sparrow_parse/vllm/huggingface_inference.py,sha256=EJnG6PesGKMc_0qGPN8ufE6pSnhAg
 sparrow_parse/vllm/inference_base.py,sha256=4mwGoAY63MB4cHZpV0czTkJWEzimmiTzqqzKmLNzgjw,820
 sparrow_parse/vllm/inference_factory.py,sha256=FTM65O-dW2WZchHOrNN7_Q3-FlVoAc65iSptuuUuClM,1166
 sparrow_parse/vllm/local_gpu_inference.py,sha256=aHoJTejb5xrXjWDIGu5RBQWEyRCOBCB04sMvO2Wyvg8,628
-sparrow_parse/vllm/mlx_inference.py,sha256=cx-PLXf1t8ro50YALddj70FiR7s0gk_Ddp-I9XlPQQU,4788
-sparrow_parse-0.4.3.dist-info/METADATA,sha256=W7zeOHa09rgn-58aIdTkNOSqBLgpziDF7sZ_059jaoo,6432
-sparrow_parse-0.4.3.dist-info/WHEEL,sha256=tZoeGjtWxWRfdplE7E3d45VPlLNQnvbKiYnx7gwAy8A,92
-sparrow_parse-0.4.3.dist-info/entry_points.txt,sha256=8CrvTVTTcz1YuZ8aRCYNOH15ZOAaYLlcbYX3t28HwJY,54
-sparrow_parse-0.4.3.dist-info/top_level.txt,sha256=n6b-WtT91zKLyCPZTP7wvne8v_yvIahcsz-4sX8I0rY,14
-sparrow_parse-0.4.3.dist-info/RECORD,,
+sparrow_parse/vllm/mlx_inference.py,sha256=0LP6tDeaID7Rtarcs_ffmK6I77e2xv5tjs4xyt0qxLo,6100
+sparrow_parse-0.4.5.dist-info/METADATA,sha256=HbGmqb5Kq7BKVNUi5wE_w69LZMAYQXI34cpRN6pdWAs,6432
+sparrow_parse-0.4.5.dist-info/WHEEL,sha256=tZoeGjtWxWRfdplE7E3d45VPlLNQnvbKiYnx7gwAy8A,92
+sparrow_parse-0.4.5.dist-info/entry_points.txt,sha256=8CrvTVTTcz1YuZ8aRCYNOH15ZOAaYLlcbYX3t28HwJY,54
+sparrow_parse-0.4.5.dist-info/top_level.txt,sha256=n6b-WtT91zKLyCPZTP7wvne8v_yvIahcsz-4sX8I0rY,14
+sparrow_parse-0.4.5.dist-info/RECORD,,