#15 add plot contours request (generation_scatter)

docs/documentation.rst

@@ -171,7 +171,7 @@ Retrieves one sample from the dataset (at a given or random index) and instantia
 
 **Outputs**
 
-- **sample** -- Summary of the retrieved sample.
+- **sample_summary** -- Summary of the retrieved sample.
 
 DatasetSummary
 --------------
@@ -203,13 +203,16 @@ Runs a generation campaing to create new designs using the trained model.
 
 **Inputs**
 
-- **requested_values** *(str)* -- List of requested values, each formatted as a string with the following format: 'variable_name:value'.
+- **requested_values** *[List of (str)]* -- List of requested values, each formatted as a string with the following format: 'variable_name:value'.
 - **n_designs** *(int)* -- Number of designs to generate.
-- **run** *(none)* -- Set to True to start the generation process.
+- **generate** *(bool)* -- Set to True to start the generation process.
+- **clear** *(bool)* -- Forget the previously generated designs.
+- **pick_previous** *(bool)* -- Iterate backward through the list of generated designs, instantiate the previous sample.
+- **pick_next** *(bool)* -- Iterate forward through the list of generated designs, instantiate the next sample.
 
 **Outputs**
 
-- **predicions** -- List of generated designs.
+- **sample_summary** -- Selected sample.
 
 GenSampleEval
 -------------
@@ -376,6 +379,27 @@ Plots the distribution contours for each pair of variables from the data in the
 
 - **img** -- Bitmap image if output_type is 'static', otherwise None.
 
+PlotContoursRequest
+-------------------
+.. image:: _images/icons/aixd_PlotContoursRequest.png
+	:align: left
+	:height: 24
+	:width: 24
+
+Plots the requested and predicted values against the distribution contours for each pair of the corresponding variables.
+
+
+**Inputs**
+
+- **request** *[List of (str)]* -- List of requested values, each formatted as a string with the following format: 'variable_name:value'.
+- **output_type** *(str)* -- Plot type: 'static' creates a bitmap image, 'interactive' launches an interactive plot in a browser.
+- **plot** *(bool)* -- Set to True to (re-)create the plot.
+- **scale** *(float)* -- Resize factor for the static plot.
+
+**Outputs**
+
+- **img** -- Bitmap image if output_type is 'static', otherwise None.
+
 PlotCorrelations
 ----------------
 .. image:: _images/icons/aixd_PlotCorrelations.png

src/aixd_grasshopper/api.py

@@ -218,6 +218,22 @@ def plot_contours():
     return response
 
 
+@app.route("/plot_contours_request", methods=["POST"])
+def plot_contours_request():
+    data = request.data
+    data = json.loads(data)
+    session_id = data["session_id"]
+    sc = SessionController.create(session_id)
+
+    output_type = data["output_type"]
+    requested_values = data["request"]
+    n_samples = data["n_samples"]
+
+    result = sc.plot_contours_request(request=requested_values, n_samples=n_samples, output_type=output_type)
+    response = json.dumps(result, cls=DataEncoder)
+    return response
+
+
 @app.route("/design_parameters", methods=["GET"])
 def get_design_parameters():
     data = request.args

src/aixd_grasshopper/components/aixd_PlotContours/metadata.json

@@ -1,6 +1,6 @@
 {
     "name": "PlotContours",
-    "nickname": "PlotCountours",
+    "nickname": "PlotContours",
     "category": "AIXD",
     "subcategory": "5 Plotter",
     "description": "Plots the distribution contours for each pair of variables from the data in the dataset.",

src/aixd_grasshopper/components/aixd_PlotContoursRequest/code.py

@@ -0,0 +1,26 @@
+# flake8: noqa
+from scriptcontext import sticky as st
+from aixd_grasshopper.gh_ui import plot_contours_request
+from aixd_grasshopper.gh_ui import get_dataobject_types
+from aixd_grasshopper.gh_ui_helper import session_id
+from aixd_grasshopper.gh_ui_helper import component_id
+from aixd_grasshopper.gh_ui_helper import convert_str_to_bitmap
+from aixd_grasshopper.gh_ui_helper import reformat_request
+
+
+cid = component_id(session_id, ghenv.Component, "create_dataset_object")
+
+n_samples = 3
+
+if plot: 
+    variable_types = get_dataobject_types(session_id())["dataobject_types"]
+    request_dict = reformat_request(request, variable_types)
+    print request_dict
+    st[cid] = plot_contours_request(session_id(), request_dict, n_samples, output_type) # if output_type interactive: will launch the plotly fig in browser
+
+if cid in st.keys():
+    print st[cid]
+    #TODO: add error msg here
+    if output_type == "static" and 'imgstr' in st[cid].keys():
+        imgstr = st[cid]['imgstr']
+        img = convert_str_to_bitmap(imgstr, scale)

src/aixd_grasshopper/components/aixd_PlotContoursRequest/metadata.json

@@ -0,0 +1,44 @@
+{
+    "name": "PlotContoursRequest",
+    "nickname": "PlotContReq",
+    "category": "AIXD",
+    "subcategory": "5 Plotter",
+    "description": "Plots the requested and predicted values against the distribution contours for each pair of the corresponding variables.",
+    "exposure": 2,
+    "ghpython": {
+        "isAdvancedMode": false,
+        "iconDisplay": 0,
+        "inputParameters": [
+            {
+                "name": "request",
+                "description": "List of requested values, each formatted as a string with the following format: 'variable_name:value'.",
+                "typeHintID": "str",
+                "scriptParamAccess": 1
+            },
+            {
+                "name": "output_type",
+                "description": "Plot type: 'static' creates a bitmap image, 'interactive' launches an interactive plot in a browser.",
+                "typeHintID": "str",
+                "scriptParamAccess": 0
+            },
+            {
+                "name": "plot",
+                "description": "Set to True to (re-)create the plot.",
+                "typeHintID": "bool",
+                "scriptParamAccess": 0
+            },
+            {
+                "name": "scale",
+                "description": "Resize factor for the static plot.",
+                "typeHintID": "float",
+                "scriptParamAccess": 0
+            }
+        ],
+        "outputParameters": [
+            {
+                "name": "img",
+                "description": "Bitmap image if output_type is 'static', otherwise None."
+            }
+        ]
+    }
+}

src/aixd_grasshopper/controller.py

@@ -314,6 +314,22 @@ def plot_contours(self, dataobjects, output_type):
         fig = plotter.contours2d(block=block, attributes=dataobjects)
         return _fig_output(fig, output_type)
 
+    def plot_contours_request(self, request, n_samples, output_type):
+        """
+        request: dictionary where keys are the names of dataobjects (usually performance attributes), and values the requested target value(s).
+        """
+        if not self.dataset:
+            raise ValueError("Dataset is not loaded.")
+        if not self.model:
+            raise ValueError("Model is not loaded.")
+
+        plotter = Plotter(datamodule=self.datamodule, output=None)
+        gen = Generator(model=self.model, datamodule=self.datamodule, over_sample=10)
+        _, detailed_results = gen.generation(request=request, n_samples=n_samples, format_out="dict_list")
+
+        fig = plotter.generation_scatter([detailed_results], n_samples=n_samples)
+        return _fig_output(fig, output_type)
+
     def model_setup(self, model_type, inputML, outputML, latent_dim, layer_widths, batch_size):
         # TODO: set defaults here if missing?
         if not self.dataset:
@@ -336,7 +352,7 @@ def model_setup(self, model_type, inputML, outputML, latent_dim, layer_widths, b
         self.datamodule = datamodule
 
         save_dir = self.dataset_path
-        
+
         if model_type == "CAE":
             model = CondAEModel.from_datamodule(
                 datamodule, layer_widths=layer_widths, latent_dim=latent_dim, save_dir=save_dir
@@ -347,7 +363,7 @@ def model_setup(self, model_type, inputML, outputML, latent_dim, layer_widths, b
             )
         else:
             raise ValueError("Model type not recognized. Choose 'CAE' or 'CVAE'.")
-        
+
         self.model = model
         self.model_is_trained = False
 
@@ -405,7 +421,7 @@ def model_load(self, model_type, checkpoint_path, checkpoint_name):
             model = CondVAEModel.load_model_from_checkpoint(checkpoint_filepath)
         else:
             raise ValueError("Model type not recognized. Choose 'CAE' or 'CVAE'.")
-        
+
         self.model = model
         self.model_is_trained = True
         self.datamodule = self._datamodule_from_dataset()
@@ -482,7 +498,7 @@ def request_designs(self, request, n_samples=1):
         if not self.dataset:
             raise ValueError("Dataset is not loaded.")
         if not self.model:
-            raise ValueError("NN model is not loaded.")
+            raise ValueError("Model is not loaded.")
 
         gen = Generator(model=self.model, datamodule=self.datamodule, over_sample=100)
         new_designs = gen.generation(request=request, n_samples=n_samples, format_out="dict_list")[0]

src/aixd_grasshopper/gh_ui.py

@@ -60,6 +60,11 @@ def plot_contours(session_id, attributes, output_type):
     return http_post_request(action="plot_contours", data=data)
 
 
+def plot_contours_request(session_id, request, n_samples, output_type):
+    data = {"session_id": session_id, "request": request, "n_samples": n_samples, "output_type": output_type}
+    return http_post_request(action="plot_contours_request", data=data)
+
+
 def plot_correlations(session_id, attributes, output_type):
     data = {"session_id": session_id, "attributes": attributes, "output_type": output_type}
     return http_post_request(action="plot_correlations", data=data)
@@ -76,7 +81,12 @@ def model_train(session_id, epochs, wb):
 
 
 def model_load(session_id, model_type, checkpoint_name, checkpoint_path):
-    data = {"session_id": session_id, "model_type": model_type, "checkpoint_name": checkpoint_name, "checkpoint_path": checkpoint_path}
+    data = {
+        "session_id": session_id,
+        "model_type": model_type,
+        "checkpoint_name": checkpoint_name,
+        "checkpoint_path": checkpoint_path,
+    }
     return http_post_request(action="model_load", data=data)