matrice-inference 0.1.2__py3-none-any.whl → 0.1.22__py3-none-any.whl

matrice_inference/__init__.py

@@ -2,6 +2,7 @@
 
 import os
 import sys
+import platform
 from matrice_common.utils import dependencies_check
 
 base = [
@@ -24,36 +25,52 @@ base = [
     "transformers"
 ]
 
-# Install base dependencies first
-dependencies_check(base)
-
 # Helper to attempt installation and verify importability
 def _install_and_verify(pkg: str, import_name: str):
     """Install a package expression and return True if the import succeeds."""
-    if dependencies_check([pkg]):
-        try:
-            __import__(import_name)
-            return True
-        except ImportError:
-            return False
-    return False
+    try:
+        if pkg=='onnxruntime-gpu':
+            pkg = 'onnxruntime'
+        __import__(pkg)
+        return True
+    except:
+        if dependencies_check([pkg]):
+            try:
+                __import__(import_name)
+                return True
+            except ImportError:
+                return False
+        return False
+
+# Runtime gating for optional OCR bootstrap (default OFF), and never on Jetson
+_ENABLE_OCR_BOOTSTRAP = os.getenv("MATRICE_ENABLE_OCR_BOOTSTRAP", "0")
+_IS_JETSON = (platform.machine().lower() in ("aarch64", "arm64"))
 
-if not dependencies_check(["opencv-python"]):
-    dependencies_check(["opencv-python-headless"])
+print("*******************************Deployment ENV Info**********************************")
+print(f"ENABLE_JETSON_PIP_SETTINGS: {_ENABLE_OCR_BOOTSTRAP}") #0 if OFF, 1 if ON, this will be set to 1 in jetson byom codebase.
+print(f"IS_JETSON_ARCH?: {_IS_JETSON}") #True if Jetson, False otherwise
+print("*************************************************************************************")
 
-# Attempt GPU-specific dependencies first
-_gpu_ok = _install_and_verify("onnxruntime-gpu", "onnxruntime") and _install_and_verify(
-    "fast-plate-ocr[onnx-gpu]", "fast_plate_ocr"
-)
+if not int(_ENABLE_OCR_BOOTSTRAP) and not _IS_JETSON:
+    # Install base dependencies first
+    dependencies_check(base)
 
-if not _gpu_ok:
-    # Fallback to CPU variants
-    _cpu_ok = _install_and_verify("onnxruntime", "onnxruntime") and _install_and_verify(
-        "fast-plate-ocr[onnx]", "fast_plate_ocr"
+    if not dependencies_check(["opencv-python"]):
+        dependencies_check(["opencv-python-headless"])
+
+    # Attempt GPU-specific dependencies first
+    _gpu_ok = _install_and_verify("onnxruntime-gpu", "onnxruntime") and _install_and_verify(
+        "fast-plate-ocr[onnx-gpu]", "fast_plate_ocr"
     )
-    if not _cpu_ok:
-        # Last-chance fallback without extras tag (PyPI sometimes lacks them)
-        _install_and_verify("fast-plate-ocr", "fast_plate_ocr")
+
+    if not _gpu_ok:
+        # Fallback to CPU variants
+        _cpu_ok = _install_and_verify("onnxruntime", "onnxruntime") and _install_and_verify(
+            "fast-plate-ocr[onnx]", "fast_plate_ocr"
+        )
+        if not _cpu_ok:
+            # Last-chance fallback without extras tag (PyPI sometimes lacks them)
+            _install_and_verify("fast-plate-ocr", "fast_plate_ocr")
 
 # matrice_deps = ["matrice_common", "matrice_analytics", "matrice"]
 

matrice_inference/server/__init__.py

@@ -1,23 +1,29 @@
 import os
 import logging
 
-# Root logger
-logging.basicConfig(level=logging.DEBUG)
+# Define paths
+log_path = os.path.join(os.getcwd(), "deploy_server.log")
 
-# Console handler (INFO+)
+# Create handlers explicitly
 console_handler = logging.StreamHandler()
-console_handler.setLevel(logging.INFO)
-
-# File handler (DEBUG+)
-log_path = os.path.join(os.getcwd(), "deploy_server.log")
 file_handler = logging.FileHandler(log_path)
+
+# Set levels
+console_handler.setLevel(logging.INFO)
 file_handler.setLevel(logging.DEBUG)
 
-# Formatter
+# Define a formatter
 formatter = logging.Formatter("%(asctime)s - %(levelname)s - %(message)s")
 console_handler.setFormatter(formatter)
 file_handler.setFormatter(formatter)
 
-# Add handlers to root logger
-logging.getLogger().addHandler(console_handler)
-logging.getLogger().addHandler(file_handler)
+# Get the root logger
+logger = logging.getLogger()
+logger.setLevel(logging.DEBUG)  # Root level must be the lowest (DEBUG)
+
+# Optional: remove any default handlers if basicConfig was called earlier
+if logger.hasHandlers():
+    logger.handlers.clear()
+
+logger.addHandler(console_handler)
+logger.addHandler(file_handler)

matrice_inference/server/model/triton_server.py

@@ -17,8 +17,6 @@ from matrice_common.utils import dependencies_check
 TRITON_DOCKER_IMAGE = "nvcr.io/nvidia/tritonserver:23.08-py3"
 BASE_PATH = "./model_repository"
 
-logging.basicConfig(level=logging.INFO, format="%(asctime)s [%(levelname)s] %(message)s")
-
 class TritonServer:
     def __init__(
         self,
@@ -1161,7 +1159,7 @@ class TritonInference:
         model_alphabet: str = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ_",
         return_confidence: bool = True,
         confidence_threshold: float = 0.0,  # Disabled threshold to match ONNX
-        ) -> Tuple[list[str], np.ndarray] | list[str]:
+        ) -> Union[Tuple[List[str], np.ndarray], List[str]]:
         """Postprocess OCR model outputs into license plate strings.
 
         Args:

matrice_inference/server/server.py

@@ -320,7 +320,10 @@ class MatriceDeployServer:
             post_processing_config = {}
         if isinstance(post_processing_config, dict):
             post_processing_config["facial_recognition_server_id"] = self.job_params.get("facial_recognition_server_id", None)
+            post_processing_config["lpr_server_id"] = self.job_params.get("lpr_server_id", None)
             post_processing_config["session"] = self.session  # Pass the session to post-processing
+            # Pass deployment_id for facial recognition deployment update
+            post_processing_config["deployment_id"] = self.deployment_id 
         
         # Get index_to_category from action_tracker if available
         index_to_category = None
@@ -374,10 +377,6 @@ class MatriceDeployServer:
             self.streaming_pipeline = StreamingPipeline(
                 inference_interface=self.inference_interface,
                 post_processor=self.post_processor,
-                consumer_threads=self.job_params.get("consumer_threads", 4),
-                producer_threads=self.job_params.get("producer_threads", 2),
-                inference_threads=self.job_params.get("inference_threads", 4),
-                postprocessing_threads=self.job_params.get("postprocessing_threads", 2),
                 inference_queue_maxsize=self.job_params.get("inference_queue_maxsize", 5000),
                 postproc_queue_maxsize=self.job_params.get("postproc_queue_maxsize", 5000),
                 output_queue_maxsize=self.job_params.get("output_queue_maxsize", 5000),