added bugfix for detr hpc. removed inference method from it.

detr_object_detection/hpc/detr_hpc_model/Dockerfile

@@ -10,6 +10,7 @@ RUN pip install --no-cache-dir -r requirements.docker.txt
 
 ENV PRODUCTION true
 ENV CUSTOM_MODEL_DIR custom_model
+ENV DEFAULT_MODEL_DIR default_model
 ENV DATASET_DIR dataset
 ENV TENSOR_BOARD_LOG_DIR tensorboard_logs/
 ENV ANNOTATION_FILE_NAME _annotations.coco.json
@@ -18,6 +19,7 @@ ENV MAX_NU_METRIC_IMAGES 10
 ENV DETECTION_THRESHOLD 0.8
 
 WORKDIR /home/objectdetection
+RUN ls
 
 #RUN pip install --no-cache-dir jupyter
 

detr_object_detection/hpc/detr_hpc_model/config.py

@@ -8,6 +8,7 @@ class Config:
         # get path to the models saving location
         if PROD_FLAG:
             
+            self.default_model_path = os.path.join(SHARED_FOLDER, os.environ['DEFAULT_MODEL_DIR'])
             self.custom_model_path = os.path.join(SHARED_FOLDER, os.environ['CUSTOM_MODEL_DIR'])
             self.dataset_path = os.path.join(SHARED_FOLDER, os.environ['DATASET_DIR'])
             self.lightning_logs_dir =  os.path.join(SHARED_FOLDER, os.environ['TENSOR_BOARD_LOG_DIR'])
@@ -34,7 +35,7 @@ class Config:
 
         self.TENSORBOARD_FOLDER = os.getenv(self.lightning_logs_dir, "./detr/dev/tensorboard_logs")
 
-        self.create_not_existing_dirs([SHARED_FOLDER,self.custom_model_path,self.dataset_path,self.lightning_logs_dir,self.input_path,self.output_path,self.status_folder])
+        self.create_not_existing_dirs([SHARED_FOLDER,self.custom_model_path,self.default_model_path,self.dataset_path,self.lightning_logs_dir,self.input_path,self.output_path,self.status_folder])
 
 
     def create_not_existing_dirs(self, list_of_paths):

detr_object_detection/hpc/detr_hpc_model/detr_client_test.py

@@ -26,10 +26,6 @@ def run_inference(input_path, output_path):
         ui_request.output = output_path
         response = stub.detect_objects(ui_request)
 
-        # training
-        # ui_request = model_pb2.TrainingConfig(config="test")
-        # response = stub.startTraining(ui_request)
-
         logger.debug("Inference finished")
     
     print(response)

detr_object_detection/hpc/detr_hpc_model/model.proto

@@ -14,17 +14,7 @@ message TrainStatus {
   int32 status = 1 ;
 }
 
-message ObjectDetectionInputFile {
-    string input = 1;
-    string output = 2;
-}
-
-message ObjectDetectionOutputFile {
-    string path = 1;
-}
-
 service DetrModelServicer {
-    rpc detect_objects(ObjectDetectionInputFile) returns (ObjectDetectionOutputFile);
     rpc startTraining(TrainingConfig) returns (TrainStatus);
 }
 

detr_object_detection/hpc/detr_hpc_model/model_pb2.py

@@ -14,15 +14,13 @@ _sym_db = _symbol_database.Default()
 
 
 
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\x0bmodel.proto\"\x07\n\x05\x45mpty\" \n\x0eTrainingConfig\x12\x0e\n\x06\x63onfig\x18\x01 \x01(\t\"\x1d\n\x0bTrainStatus\x12\x0e\n\x06status\x18\x01 \x01(\x05\"9\n\x18ObjectDetectionInputFile\x12\r\n\x05input\x18\x01 \x01(\t\x12\x0e\n\x06output\x18\x02 \x01(\t\")\n\x19ObjectDetectionOutputFile\x12\x0c\n\x04path\x18\x01 \x01(\t2\x8c\x01\n\x11\x44\x65trModelServicer\x12G\n\x0e\x64\x65tect_objects\x12\x19.ObjectDetectionInputFile\x1a\x1a.ObjectDetectionOutputFile\x12.\n\rstartTraining\x12\x0f.TrainingConfig\x1a\x0c.TrainStatusb\x06proto3')
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\x0bmodel.proto\"\x07\n\x05\x45mpty\" \n\x0eTrainingConfig\x12\x0e\n\x06\x63onfig\x18\x01 \x01(\t\"\x1d\n\x0bTrainStatus\x12\x0e\n\x06status\x18\x01 \x01(\x05\x32\x43\n\x11\x44\x65trModelServicer\x12.\n\rstartTraining\x12\x0f.TrainingConfig\x1a\x0c.TrainStatusb\x06proto3')
 
 
 
 _EMPTY = DESCRIPTOR.message_types_by_name['Empty']
 _TRAININGCONFIG = DESCRIPTOR.message_types_by_name['TrainingConfig']
 _TRAINSTATUS = DESCRIPTOR.message_types_by_name['TrainStatus']
-_OBJECTDETECTIONINPUTFILE = DESCRIPTOR.message_types_by_name['ObjectDetectionInputFile']
-_OBJECTDETECTIONOUTPUTFILE = DESCRIPTOR.message_types_by_name['ObjectDetectionOutputFile']
 Empty = _reflection.GeneratedProtocolMessageType('Empty', (_message.Message,), {
   'DESCRIPTOR' : _EMPTY,
   '__module__' : 'model_pb2'
@@ -44,20 +42,6 @@ TrainStatus = _reflection.GeneratedProtocolMessageType('TrainStatus', (_message.
   })
 _sym_db.RegisterMessage(TrainStatus)
 
-ObjectDetectionInputFile = _reflection.GeneratedProtocolMessageType('ObjectDetectionInputFile', (_message.Message,), {
-  'DESCRIPTOR' : _OBJECTDETECTIONINPUTFILE,
-  '__module__' : 'model_pb2'
-  # @@protoc_insertion_point(class_scope:ObjectDetectionInputFile)
-  })
-_sym_db.RegisterMessage(ObjectDetectionInputFile)
-
-ObjectDetectionOutputFile = _reflection.GeneratedProtocolMessageType('ObjectDetectionOutputFile', (_message.Message,), {
-  'DESCRIPTOR' : _OBJECTDETECTIONOUTPUTFILE,
-  '__module__' : 'model_pb2'
-  # @@protoc_insertion_point(class_scope:ObjectDetectionOutputFile)
-  })
-_sym_db.RegisterMessage(ObjectDetectionOutputFile)
-
 _DETRMODELSERVICER = DESCRIPTOR.services_by_name['DetrModelServicer']
 if _descriptor._USE_C_DESCRIPTORS == False:
 
@@ -68,10 +52,6 @@ if _descriptor._USE_C_DESCRIPTORS == False:
   _TRAININGCONFIG._serialized_end=56
   _TRAINSTATUS._serialized_start=58
   _TRAINSTATUS._serialized_end=87
-  _OBJECTDETECTIONINPUTFILE._serialized_start=89
-  _OBJECTDETECTIONINPUTFILE._serialized_end=146
-  _OBJECTDETECTIONOUTPUTFILE._serialized_start=148
-  _OBJECTDETECTIONOUTPUTFILE._serialized_end=189
-  _DETRMODELSERVICER._serialized_start=192
-  _DETRMODELSERVICER._serialized_end=332
+  _DETRMODELSERVICER._serialized_start=89
+  _DETRMODELSERVICER._serialized_end=156
 # @@protoc_insertion_point(module_scope)

detr_object_detection/hpc/detr_hpc_model/model_pb2_grpc.py

@@ -14,11 +14,6 @@ class DetrModelServicerStub(object):
         Args:
             channel: A grpc.Channel.
         """
-        self.detect_objects = channel.unary_unary(
-                '/DetrModelServicer/detect_objects',
-                request_serializer=model__pb2.ObjectDetectionInputFile.SerializeToString,
-                response_deserializer=model__pb2.ObjectDetectionOutputFile.FromString,
-                )
         self.startTraining = channel.unary_unary(
                 '/DetrModelServicer/startTraining',
                 request_serializer=model__pb2.TrainingConfig.SerializeToString,
@@ -29,12 +24,6 @@ class DetrModelServicerStub(object):
 class DetrModelServicerServicer(object):
     """Missing associated documentation comment in .proto file."""
 
-    def detect_objects(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
     def startTraining(self, request, context):
         """Missing associated documentation comment in .proto file."""
         context.set_code(grpc.StatusCode.UNIMPLEMENTED)
@@ -44,11 +33,6 @@ class DetrModelServicerServicer(object):
 
 def add_DetrModelServicerServicer_to_server(servicer, server):
     rpc_method_handlers = {
-            'detect_objects': grpc.unary_unary_rpc_method_handler(
-                    servicer.detect_objects,
-                    request_deserializer=model__pb2.ObjectDetectionInputFile.FromString,
-                    response_serializer=model__pb2.ObjectDetectionOutputFile.SerializeToString,
-            ),
             'startTraining': grpc.unary_unary_rpc_method_handler(
                     servicer.startTraining,
                     request_deserializer=model__pb2.TrainingConfig.FromString,
@@ -64,23 +48,6 @@ def add_DetrModelServicerServicer_to_server(servicer, server):
 class DetrModelServicer(object):
     """Missing associated documentation comment in .proto file."""
 
-    @staticmethod
-    def detect_objects(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/DetrModelServicer/detect_objects',
-            model__pb2.ObjectDetectionInputFile.SerializeToString,
-            model__pb2.ObjectDetectionOutputFile.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
     @staticmethod
     def startTraining(request,
             target,

detr_object_detection/hpc/detr_hpc_model/od/detrOD.py

@@ -23,14 +23,15 @@ class DetrOD():
         self.config = config
         # get path to folder with scripts
         self.current_dir_path = os.path.dirname(os.path.abspath(__file__))
-        self.default_model_path = os.path.join(self.current_dir_path, "default_model")
+        #self.default_model_path = os.path.join(self.current_dir_path, "default_model")
+        self.default_model_path = config.default_model_path
             
         model_path = self.default_model_path
         extractor_path = self.default_model_path
         try:
             # loading default model & extractor on the first stop always
-            self.model = DetrForObjectDetection.from_pretrained(model_path, local_files_only=config.PROD_FLAG)
-            self.extractor = DetrImageProcessor.from_pretrained(extractor_path, local_files_only=config.PROD_FLAG)
+            self.model = DetrForObjectDetection.from_pretrained(model_path)
+            self.extractor = DetrImageProcessor.from_pretrained(extractor_path)
         except Exception:
             logger.info("Could not find the default model. Downloading it now")
             self.download_default_model()
@@ -38,10 +39,11 @@ class DetrOD():
 
     def download_default_model(self):
         checkpoint = "ciasimbaya/ObjectDetection"
-        model = DetrForObjectDetection.from_pretrained(checkpoint)
-        extractor = DetrImageProcessor.from_pretrained(checkpoint)
-        extractor.save_pretrained(self.default_model_path)
-        model.model.save_pretrained(self.default_model_path)
+        self.model = DetrForObjectDetection.from_pretrained(checkpoint)
+        self.extractor = DetrImageProcessor.from_pretrained(checkpoint)
+        # In HPC files cannot be saved under apptainer folders.
+        self.extractor.save_pretrained(self.default_model_path)
+        self.model.model.save_pretrained(self.default_model_path)
 
 
     def load_custom_model(self):

detr_object_detection/hpc/detr_hpc_s3_databroker/Dockerfile

@@ -15,6 +15,8 @@ ENV ANNOTATION_FILE_NAME _annotations.coco.json
 
 WORKDIR /home/databroker
 
+RUN ls
+
 COPY main.py app.py config.py logger.py s3_func.py model.proto model_pb2.py model_pb2_grpc.py detr_databroker_server.py ./
 COPY static ./static
 COPY templates ./templates

detr_object_detection/hpc/detr_hpc_s3_databroker/detr_databroker_server.py

@@ -4,7 +4,6 @@ import model_pb2_grpc
 from concurrent import futures 
 import threading
 import os
-import app as webui
 
 from logger import Logger
 logger = Logger(__name__)
@@ -19,48 +18,11 @@ class DetrDataBrokerServicer(model_pb2_grpc.DetrDatabroker):
     def send_training_signal(self, request, context):
         logger.info("Preparing training signal to be sent to the model.")
         out_message = model_pb2.TrainingConfig()
-
-        # checking if training message is already being sent
-        if not webui.is_training_startable():
-            context.set_code(grpc.StatusCode.NOT_FOUND)
-            context.set_details("No new data for training has been detected.")
-            logger.info("Training not startable. Not sending training signal to the model.")
-            return out_message
         
         training_status = "starting"
         out_message.config=training_status
         logger.info("Sending training signal to the model.")
-        webui.change_new_train_flag(False)
         return out_message
-        
-
-    def send_detect_signal(self, request, context):
-        logger.info("Preparing inference signal to be sent to the model.")
-        image_path = webui.get_image_path()
-        out_message = model_pb2.ObjectDetectionInputFile()
-        
-        # check if the request is already sent
-        if not webui.is_inference_startable():
-            context.set_code(grpc.StatusCode.NOT_FOUND)
-            context.set_details("No new data for inference has been detected.")
-            logger.info("Inference not startable. Not sending inference signal to the model.")
-            return out_message
-
-        if not image_path:
-            logger.error("No image to be found under the path")
-            context.set_code(grpc.StatusCode.NOT_FOUND)
-            context.set_details("No image to be found under the path")
-            return out_message
-        
-        if os.path.exists(image_path):
-            out_message.path = image_path
-            logger.info("Sending inference signal to the model.")
-            webui.change_new_inference_flag(False)
-            return out_message
-        else:
-            logger.error("Inference image is not saved and cannot be found. No request was sent")
-            context.set_code(grpc.StatusCode.NOT_FOUND)
-            context.set_details("Inference image is not saved.")
 
 
 def serve_grpc(config,port):

detr_object_detection/hpc/detr_hpc_s3_databroker/model.proto

@@ -3,22 +3,10 @@ syntax = "proto3";
 message Empty {
 }
 
-message DatasetFeatues{
-  string datasetname = 1;
-  string description = 2;
-  string size = 3;
-  string DOI_ID = 4;
-}
-
-message ObjectDetectionInputFile {
-    string path = 1;
-}
-
 message TrainingConfig {
   string config = 1;
 }
 
 service DetrDatabroker {
-  rpc send_detect_signal(Empty) returns (ObjectDetectionInputFile);
   rpc send_training_signal(Empty) returns (TrainingConfig);
 }
\ No newline at end of file

detr_object_detection/hpc/detr_hpc_s3_databroker/model_pb2.py

@@ -14,13 +14,11 @@ _sym_db = _symbol_database.Default()
 
 
 
-DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\x0bmodel.proto\"\x07\n\x05\x45mpty\"X\n\x0e\x44\x61tasetFeatues\x12\x13\n\x0b\x64\x61tasetname\x18\x01 \x01(\t\x12\x13\n\x0b\x64\x65scription\x18\x02 \x01(\t\x12\x0c\n\x04size\x18\x03 \x01(\t\x12\x0e\n\x06\x44OI_ID\x18\x04 \x01(\t\"(\n\x18ObjectDetectionInputFile\x12\x0c\n\x04path\x18\x01 \x01(\t\" \n\x0eTrainingConfig\x12\x0e\n\x06\x63onfig\x18\x01 \x01(\t2\xab\x01\n\x0e\x44\x65trDatabroker\x12\x37\n\x12send_detect_signal\x12\x06.Empty\x1a\x19.ObjectDetectionInputFile\x12/\n\x14get_dataset_metadata\x12\x06.Empty\x1a\x0f.DatasetFeatues\x12/\n\x14send_training_signal\x12\x06.Empty\x1a\x0f.TrainingConfigb\x06proto3')
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(b'\n\x0bmodel.proto\"\x07\n\x05\x45mpty\" \n\x0eTrainingConfig\x12\x0e\n\x06\x63onfig\x18\x01 \x01(\t2A\n\x0e\x44\x65trDatabroker\x12/\n\x14send_training_signal\x12\x06.Empty\x1a\x0f.TrainingConfigb\x06proto3')
 
 
 
 _EMPTY = DESCRIPTOR.message_types_by_name['Empty']
-_DATASETFEATUES = DESCRIPTOR.message_types_by_name['DatasetFeatues']
-_OBJECTDETECTIONINPUTFILE = DESCRIPTOR.message_types_by_name['ObjectDetectionInputFile']
 _TRAININGCONFIG = DESCRIPTOR.message_types_by_name['TrainingConfig']
 Empty = _reflection.GeneratedProtocolMessageType('Empty', (_message.Message,), {
   'DESCRIPTOR' : _EMPTY,
@@ -29,20 +27,6 @@ Empty = _reflection.GeneratedProtocolMessageType('Empty', (_message.Message,), {
   })
 _sym_db.RegisterMessage(Empty)
 
-DatasetFeatues = _reflection.GeneratedProtocolMessageType('DatasetFeatues', (_message.Message,), {
-  'DESCRIPTOR' : _DATASETFEATUES,
-  '__module__' : 'model_pb2'
-  # @@protoc_insertion_point(class_scope:DatasetFeatues)
-  })
-_sym_db.RegisterMessage(DatasetFeatues)
-
-ObjectDetectionInputFile = _reflection.GeneratedProtocolMessageType('ObjectDetectionInputFile', (_message.Message,), {
-  'DESCRIPTOR' : _OBJECTDETECTIONINPUTFILE,
-  '__module__' : 'model_pb2'
-  # @@protoc_insertion_point(class_scope:ObjectDetectionInputFile)
-  })
-_sym_db.RegisterMessage(ObjectDetectionInputFile)
-
 TrainingConfig = _reflection.GeneratedProtocolMessageType('TrainingConfig', (_message.Message,), {
   'DESCRIPTOR' : _TRAININGCONFIG,
   '__module__' : 'model_pb2'
@@ -56,12 +40,8 @@ if _descriptor._USE_C_DESCRIPTORS == False:
   DESCRIPTOR._options = None
   _EMPTY._serialized_start=15
   _EMPTY._serialized_end=22
-  _DATASETFEATUES._serialized_start=24
-  _DATASETFEATUES._serialized_end=112
-  _OBJECTDETECTIONINPUTFILE._serialized_start=114
-  _OBJECTDETECTIONINPUTFILE._serialized_end=154
-  _TRAININGCONFIG._serialized_start=156
-  _TRAININGCONFIG._serialized_end=188
-  _DETRDATABROKER._serialized_start=191
-  _DETRDATABROKER._serialized_end=362
+  _TRAININGCONFIG._serialized_start=24
+  _TRAININGCONFIG._serialized_end=56
+  _DETRDATABROKER._serialized_start=58
+  _DETRDATABROKER._serialized_end=123
 # @@protoc_insertion_point(module_scope)

detr_object_detection/hpc/detr_hpc_s3_databroker/model_pb2_grpc.py

@@ -14,16 +14,6 @@ class DetrDatabrokerStub(object):
         Args:
             channel: A grpc.Channel.
         """
-        self.send_detect_signal = channel.unary_unary(
-                '/DetrDatabroker/send_detect_signal',
-                request_serializer=model__pb2.Empty.SerializeToString,
-                response_deserializer=model__pb2.ObjectDetectionInputFile.FromString,
-                )
-        self.get_dataset_metadata = channel.unary_unary(
-                '/DetrDatabroker/get_dataset_metadata',
-                request_serializer=model__pb2.Empty.SerializeToString,
-                response_deserializer=model__pb2.DatasetFeatues.FromString,
-                )
         self.send_training_signal = channel.unary_unary(
                 '/DetrDatabroker/send_training_signal',
                 request_serializer=model__pb2.Empty.SerializeToString,
@@ -34,18 +24,6 @@ class DetrDatabrokerStub(object):
 class DetrDatabrokerServicer(object):
     """Missing associated documentation comment in .proto file."""
 
-    def send_detect_signal(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
-    def get_dataset_metadata(self, request, context):
-        """Missing associated documentation comment in .proto file."""
-        context.set_code(grpc.StatusCode.UNIMPLEMENTED)
-        context.set_details('Method not implemented!')
-        raise NotImplementedError('Method not implemented!')
-
     def send_training_signal(self, request, context):
         """Missing associated documentation comment in .proto file."""
         context.set_code(grpc.StatusCode.UNIMPLEMENTED)
@@ -55,16 +33,6 @@ class DetrDatabrokerServicer(object):
 
 def add_DetrDatabrokerServicer_to_server(servicer, server):
     rpc_method_handlers = {
-            'send_detect_signal': grpc.unary_unary_rpc_method_handler(
-                    servicer.send_detect_signal,
-                    request_deserializer=model__pb2.Empty.FromString,
-                    response_serializer=model__pb2.ObjectDetectionInputFile.SerializeToString,
-            ),
-            'get_dataset_metadata': grpc.unary_unary_rpc_method_handler(
-                    servicer.get_dataset_metadata,
-                    request_deserializer=model__pb2.Empty.FromString,
-                    response_serializer=model__pb2.DatasetFeatues.SerializeToString,
-            ),
             'send_training_signal': grpc.unary_unary_rpc_method_handler(
                     servicer.send_training_signal,
                     request_deserializer=model__pb2.Empty.FromString,
@@ -80,40 +48,6 @@ def add_DetrDatabrokerServicer_to_server(servicer, server):
 class DetrDatabroker(object):
     """Missing associated documentation comment in .proto file."""
 
-    @staticmethod
-    def send_detect_signal(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/DetrDatabroker/send_detect_signal',
-            model__pb2.Empty.SerializeToString,
-            model__pb2.ObjectDetectionInputFile.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
-    @staticmethod
-    def get_dataset_metadata(request,
-            target,
-            options=(),
-            channel_credentials=None,
-            call_credentials=None,
-            insecure=False,
-            compression=None,
-            wait_for_ready=None,
-            timeout=None,
-            metadata=None):
-        return grpc.experimental.unary_unary(request, target, '/DetrDatabroker/get_dataset_metadata',
-            model__pb2.Empty.SerializeToString,
-            model__pb2.DatasetFeatues.FromString,
-            options, channel_credentials,
-            insecure, call_credentials, compression, wait_for_ready, timeout, metadata)
-
     @staticmethod
     def send_training_signal(request,
             target,

detr_object_detection/normal/detr_model/od/detrOD.py

@@ -38,10 +38,10 @@ class DetrOD():
 
     def download_default_model(self):
         checkpoint = "ciasimbaya/ObjectDetection"
-        model = DetrForObjectDetection.from_pretrained(checkpoint)
-        extractor = DetrImageProcessor.from_pretrained(checkpoint)
-        extractor.save_pretrained(self.default_model_path)
-        model.model.save_pretrained(self.default_model_path)
+        self.model = DetrForObjectDetection.from_pretrained(checkpoint)
+        self.extractor = DetrImageProcessor.from_pretrained(checkpoint)
+        self.extractor.save_pretrained(self.default_model_path)
+        self.model.model.save_pretrained(self.default_model_path)
 
 
     def load_custom_model(self):