Add model metric logging

examples/ensemble/preprocess.py

@@ -9,11 +9,11 @@ class Preprocess(object):
         # set internal state, this will be called only once. (i.e. not per request)
         pass
 
-    def preprocess(self, body: dict) -> Any:
+    def preprocess(self, body: dict, collect_custom_statistics_fn=None) -> Any:
         # we expect to get two valid on the dict x0, and x1
         return [[body.get("x0", None), body.get("x1", None)], ]
 
-    def postprocess(self, data: Any) -> dict:
+    def postprocess(self, data: Any, collect_custom_statistics_fn=None) -> dict:
         # post process the data returned from the model inference engine
         # data is the return value from model.predict we will put is inside a return value as Y
         return dict(y=data.tolist() if isinstance(data, np.ndarray) else data)

examples/ensemble/readme.md

@@ -29,3 +29,4 @@ Or add Canary endpoint
 
 > **_Notice:_**  You can also change the serving service while it is already running!
 This includes adding/removing endpoints, adding canary model routing etc.
+by default new endpoints/models will be automatically updated after 1 minute

examples/keras/preprocess.py

@@ -13,7 +13,7 @@ class Preprocess(object):
         # set internal state, this will be called only once. (i.e. not per request)
         pass
 
-    def preprocess(self, body: dict) -> Any:
+    def preprocess(self, body: dict, collect_custom_statistics_fn=None) -> Any:
         # we expect to get two valid on the dict x0, and x1
         url = body.get("url")
         if not url:
@@ -25,7 +25,7 @@ class Preprocess(object):
 
         return np.array(image).flatten()
 
-    def postprocess(self, data: Any) -> dict:
+    def postprocess(self, data: Any, collect_custom_statistics_fn=None) -> dict:
         # post process the data returned from the model inference engine
         # data is the return value from model.predict we will put is inside a return value as Y
         if not isinstance(data, np.ndarray):

examples/keras/readme.md

@@ -12,6 +12,8 @@ The output will be a model created on the project "serving examples", by the nam
 
 ## setting up the serving service
 
+Prerequisites, Keras/Tensorflow models require Triton engine support, please use `docker-compose-triton.yml` / `docker-compose-triton-gpu.yml` or if running on Kubernetes, the matching helm chart.
+
 1. Create serving Service: `clearml-serving create --name "serving example"` (write down the service ID)
 2. Create model endpoint: 
 
@@ -36,3 +38,4 @@ Or add Canary endpoint
 
 > **_Notice:_**  You can also change the serving service while it is already running!
 This includes adding/removing endpoints, adding canary model routing etc.
+by default new endpoints/models will be automatically updated after 1 minute

examples/lightgbm/preprocess.py

@@ -9,14 +9,14 @@ class Preprocess(object):
         # set internal state, this will be called only once. (i.e. not per request)
         pass
 
-    def preprocess(self, body: dict) -> Any:
+    def preprocess(self, body: dict, collect_custom_statistics_fn=None) -> Any:
         # we expect to get four valid numbers on the dict: x0, x1, x2, x3
         return np.array(
             [[body.get("x0", None), body.get("x1", None), body.get("x2", None), body.get("x3", None)], ],
             dtype=np.float32
         )
 
-    def postprocess(self, data: Any) -> dict:
+    def postprocess(self, data: Any, collect_custom_statistics_fn=None) -> dict:
         # post process the data returned from the model inference engine
         # data is the return value from model.predict we will put is inside a return value as Y
         # we pick the most probably class and return the class index (argmax)

examples/pipeline/preprocess.py

@@ -7,14 +7,14 @@ class Preprocess(object):
         # set internal state, this will be called only once. (i.e. not per request)
         pass
 
-    def postprocess(self, data: List[dict]) -> dict:
+    def postprocess(self, data: List[dict], collect_custom_statistics_fn=None) -> dict:
         # we will here average the results and return the new value
         # assume data is a list of dicts greater than 1
 
         # average result
         return dict(y=0.5 * data[0]['y'][0] + 0.5 * data[1]['y'][0])
 
-    def process(self, data: Any) -> Any:
+    def process(self, data: Any, collect_custom_statistics_fn=None) -> Any:
         """
         do something with the actual data, return any type of object.
         The returned object will be passed as is to the postprocess function engine

examples/pipeline/readme.md

@@ -24,3 +24,4 @@ Training a scikit-learn model (see example/sklearn)
 
 > **_Notice:_**  You can also change the serving service while it is already running!
 This includes adding/removing endpoints, adding canary model routing etc.
+by default new endpoints/models will be automatically updated after 1 minute

examples/pytorch/preprocess.py

@@ -13,7 +13,7 @@ class Preprocess(object):
         # set internal state, this will be called only once. (i.e. not per request)
         pass
 
-    def preprocess(self, body: dict) -> Any:
+    def preprocess(self, body: dict, collect_custom_statistics_fn=None) -> Any:
         # we expect to get two valid on the dict x0, and x1
         url = body.get("url")
         if not url:
@@ -24,7 +24,7 @@ class Preprocess(object):
         image = ImageOps.grayscale(image).resize((28, 28))
         return np.array(image).flatten()
 
-    def postprocess(self, data: Any) -> dict:
+    def postprocess(self, data: Any, collect_custom_statistics_fn=None) -> dict:
         # post process the data returned from the model inference engine
         # data is the return value from model.predict we will put is inside a return value as Y
         if not isinstance(data, np.ndarray):

examples/pytorch/readme.md

@@ -13,6 +13,9 @@ The output will be a model created on the project "serving examples", by the nam
 
 ## setting up the serving service
 
+
+Prerequisites, PyTorch models require Triton engine support, please use `docker-compose-triton.yml` / `docker-compose-triton-gpu.yml` or if running on Kubernetes, the matching helm chart.
+
 1. Create serving Service: `clearml-serving create --name "serving example"` (write down the service ID)
 2. Create model endpoint:
 
@@ -39,4 +42,4 @@ Or add Canary endpoint
 
 > **_Notice:_**  You can also change the serving service while it is already running!
 This includes adding/removing endpoints, adding canary model routing etc.
-
+by default new endpoints/models will be automatically updated after 1 minute

examples/sklearn/preprocess.py

@@ -9,11 +9,11 @@ class Preprocess(object):
         # set internal state, this will be called only once. (i.e. not per request)
         pass
 
-    def preprocess(self, body: dict) -> Any:
+    def preprocess(self, body: dict, collect_custom_statistics_fn=None) -> Any:
         # we expect to get two valid on the dict x0, and x1
         return [[body.get("x0", None), body.get("x1", None)], ]
 
-    def postprocess(self, data: Any) -> dict:
+    def postprocess(self, data: Any, collect_custom_statistics_fn=None) -> dict:
         # post process the data returned from the model inference engine
         # data is the return value from model.predict we will put is inside a return value as Y
         return dict(y=data.tolist() if isinstance(data, np.ndarray) else data)

examples/sklearn/readme.md

@@ -29,3 +29,4 @@ Or add Canary endpoint
 
 > **_Notice:_**  You can also change the serving service while it is already running!
 This includes adding/removing endpoints, adding canary model routing etc.
+by default new endpoints/models will be automatically updated after 1 minute

examples/xgboost/preprocess.py

@@ -10,12 +10,12 @@ class Preprocess(object):
         # set internal state, this will be called only once. (i.e. not per request)
         pass
 
-    def preprocess(self, body: dict) -> Any:
+    def preprocess(self, body: dict, collect_custom_statistics_fn=None) -> Any:
         # we expect to get four valid numbers on the dict: x0, x1, x2, x3
         return xgb.DMatrix(
             [[body.get("x0", None), body.get("x1", None), body.get("x2", None), body.get("x3", None)]])
 
-    def postprocess(self, data: Any) -> dict:
+    def postprocess(self, data: Any, collect_custom_statistics_fn=None) -> dict:
         # post process the data returned from the model inference engine
         # data is the return value from model.predict we will put is inside a return value as Y
         return dict(y=data.tolist() if isinstance(data, np.ndarray) else data)