sparrow-parse 0.4.4__py3-none-any.whl → 0.4.6__py3-none-any.whl

sparrow_parse/__init__.py

@@ -1 +1 @@
-__version__ = '0.4.4'
+__version__ = '0.4.6'

sparrow_parse/extractors/vllm_extractor.py

@@ -152,34 +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-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}")
+    # 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/inference_factory.py

@@ -23,3 +23,9 @@ class InferenceFactory:
         # model = torch.load('model.pth')
         # return model
         raise NotImplementedError("Model loading logic not implemented")
+
+
+    def unload_inference_instance(self, instance):
+        if instance and hasattr(instance, "unload_model"):
+            instance.unload_model()
+            print(f"Inference instance of type {type(instance).__name__} has been unloaded.")

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):
@@ -24,24 +25,22 @@ class MLXInference(ModelInference):
         print(f"MLXInference initialized with model: {model_name}")
 
 
-    def __del__(self):
-        """
-        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
-        print(f"Model {self.model_name} and its resources have been unloaded.")
+
+        # 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):

{sparrow_parse-0.4.4.dist-info → sparrow_parse-0.4.6.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: sparrow-parse
-Version: 0.4.4
+Version: 0.4.6
 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-0.4.4.dist-info → sparrow_parse-0.4.6.dist-info}/RECORD

@@ -1,7 +1,7 @@
-sparrow_parse/__init__.py,sha256=jP9l7AhBCN2A-6tezbTIihxoMTDna4SLTYhvVxwbdNM,21
+sparrow_parse/__init__.py,sha256=IIRkY9z46sq_aFXyT8WySWX9GdnZHB1GyedV8VfB_lA,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=PDLgLlKiq3Bv-UOQTzX3AgxNOLcEU2EniGAXLjMC30U,7820
+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
@@ -9,11 +9,11 @@ sparrow_parse/processors/table_structure_processor.py,sha256=PQHHFdQUuTin3Mm2USu
 sparrow_parse/vllm/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 sparrow_parse/vllm/huggingface_inference.py,sha256=EJnG6PesGKMc_0qGPN8ufE6pSnhAgFu0XjCbaLCNVyM,1980
 sparrow_parse/vllm/inference_base.py,sha256=4mwGoAY63MB4cHZpV0czTkJWEzimmiTzqqzKmLNzgjw,820
-sparrow_parse/vllm/inference_factory.py,sha256=FTM65O-dW2WZchHOrNN7_Q3-FlVoAc65iSptuuUuClM,1166
+sparrow_parse/vllm/inference_factory.py,sha256=J5_sPIL6YlkttaGeTc6_6t4004klQxESeeLLhXHvyZg,1407
 sparrow_parse/vllm/local_gpu_inference.py,sha256=aHoJTejb5xrXjWDIGu5RBQWEyRCOBCB04sMvO2Wyvg8,628
-sparrow_parse/vllm/mlx_inference.py,sha256=c6-s493jLXE3DfYnwsybiqgk3GU9GEaWt3CrfqLSWKQ,5872
-sparrow_parse-0.4.4.dist-info/METADATA,sha256=x_jaR76FUv5-kR9R5YR9So3OZbWj_rG8hjFZ07lZuto,6432
-sparrow_parse-0.4.4.dist-info/WHEEL,sha256=tZoeGjtWxWRfdplE7E3d45VPlLNQnvbKiYnx7gwAy8A,92
-sparrow_parse-0.4.4.dist-info/entry_points.txt,sha256=8CrvTVTTcz1YuZ8aRCYNOH15ZOAaYLlcbYX3t28HwJY,54
-sparrow_parse-0.4.4.dist-info/top_level.txt,sha256=n6b-WtT91zKLyCPZTP7wvne8v_yvIahcsz-4sX8I0rY,14
-sparrow_parse-0.4.4.dist-info/RECORD,,
+sparrow_parse/vllm/mlx_inference.py,sha256=NNBJ_eQZTzJ6X05wUfJhGLJjm1ANMB5BxpW7GMR4S3g,5955
+sparrow_parse-0.4.6.dist-info/METADATA,sha256=YiIuculsNFZVH2YKtSGTRpOMPpc6B2DhH0sbW99onBM,6432
+sparrow_parse-0.4.6.dist-info/WHEEL,sha256=tZoeGjtWxWRfdplE7E3d45VPlLNQnvbKiYnx7gwAy8A,92
+sparrow_parse-0.4.6.dist-info/entry_points.txt,sha256=8CrvTVTTcz1YuZ8aRCYNOH15ZOAaYLlcbYX3t28HwJY,54
+sparrow_parse-0.4.6.dist-info/top_level.txt,sha256=n6b-WtT91zKLyCPZTP7wvne8v_yvIahcsz-4sX8I0rY,14
+sparrow_parse-0.4.6.dist-info/RECORD,,