Added option to run multiple analytics workflows during history matching

.github/workflows/flownet.yml

@@ -40,10 +40,10 @@ jobs:
           INPUT_MODEL_FOLDER: ./flownet-testdata/${{ matrix.flownet-model }}/input_model
           # If you want the CI to (temporarily) run against your fork of the testdada,
           # change the value her from "equinor" to your username.
-          TESTDATA_REPO_OWNER: equinor
+          TESTDATA_REPO_OWNER: wouterjdb
           # If you want the CI to (temporarily) run against another branch than master,
           # change the value her from "master" to the relevant branch name.
-          TESTDATA_REPO_BRANCH: master
+          TESTDATA_REPO_BRANCH: add-new-analysis-defs
         run: |
           git clone --depth 1 --branch $TESTDATA_REPO_BRANCH https://github.com/$TESTDATA_REPO_OWNER/flownet-testdata.git --recursive
           for f in $INPUT_MODEL_FOLDER/*.tar.gz; do tar -zxvf "$f" -C $INPUT_MODEL_FOLDER; done

CHANGELOG.md

@@ -16,6 +16,7 @@ This project adheres to [Semantic Versioning](https://semver.org/).
 - [#272](https://github.com/equinor/flownet/pull/272) Adds resampling of observation dates at requested frequency by finding nearest date among existing observation dates (i.e., no interpolated dates added)
 
 ### Changes
+- [#337](https://github.com/equinor/flownet/pull/337) You can now add multiple analytics workflows in your FlowNet config. This is also a breaking change in that it requires you to change the ert.analysis keyword to being a list.
 - [#322](https://github.com/equinor/flownet/pull/322) RSVD input through csv files can now be done either as one table used for all EQLNUM regions, or as one table for each EQLNUM region. The csv file needs a header with column names "depth", "rs" and "eqlnum" (the latter only when multiple tables are defined).
 
 ## [0.4.0] - 2020-11-18

src/flownet/config_parser/_config_parser.py

@@ -553,31 +553,54 @@ def _to_abs_path(path: Optional[str]) -> str:
                         "and webviz",
                     },
                     "analysis": {
-                        MK.Type: types.NamedDict,
+                        MK.Type: types.List,
+                        MK.Description: "List of analysis workflows to run.",
                         MK.Content: {
-                            "metric": {
-                                MK.Type: types.List,
-                                MK.Content: {
-                                    MK.Item: {MK.Type: types.String, MK.AllowNone: True}
-                                },
-                                MK.Transformation: _to_upper,
-                                MK.Description: "List of accuracy metrics to be computed "
-                                "in FlowNet analysis workflow",
-                            },
-                            "quantity": {
-                                MK.Type: types.List,
+                            MK.Item: {
+                                MK.Type: types.NamedDict,
+                                MK.Description: "Definitions of the analysis workflow.",
                                 MK.Content: {
-                                    MK.Item: {MK.Type: types.String, MK.AllowNone: True}
+                                    "metric": {
+                                        MK.Type: types.List,
+                                        MK.Content: {
+                                            MK.Item: {
+                                                MK.Type: types.String,
+                                                MK.AllowNone: True,
+                                            }
+                                        },
+                                        MK.Transformation: _to_upper,
+                                        MK.Description: "List of accuracy metrics to be computed "
+                                        "in FlowNet analysis workflow",
+                                    },
+                                    "quantity": {
+                                        MK.Type: types.List,
+                                        MK.Content: {
+                                            MK.Item: {
+                                                MK.Type: types.String,
+                                                MK.AllowNone: True,
+                                            }
+                                        },
+                                        MK.Transformation: _to_upper,
+                                        MK.Description: "List of summary vectors for which accuracy "
+                                        "is to be computed",
+                                    },
+                                    "start": {
+                                        MK.Type: types.Date,
+                                        MK.AllowNone: True,
+                                        MK.Description: "Start date in YYYY-MM-DD format.",
+                                    },
+                                    "end": {
+                                        MK.Type: types.Date,
+                                        MK.AllowNone: True,
+                                        MK.Description: "End date in YYYY-MM-DD format.",
+                                    },
+                                    "outfile": {
+                                        MK.Type: types.String,
+                                        MK.AllowNone: True,
+                                        MK.Description: "The filename of the output of the workflow. "
+                                        "In case multiple analysis workflows are run this name should be unique.",
+                                    },
                                 },
-                                MK.Transformation: _to_upper,
-                                MK.Description: "List of summary vectors for which accuracy "
-                                "is to be computed",
-                            },
-                            "start": {MK.Type: types.Date, MK.AllowNone: True},
-                            "end": {MK.Type: types.Date, MK.AllowNone: True},
-                            "outfile": {
-                                MK.Type: types.String,
-                                MK.AllowNone: True,
                             },
                         },
                     },
@@ -1850,14 +1873,18 @@ def parse_config(
         )
 
     for key in config.flownet.hyperopt.loss.keys:
-        if not key in config.ert.analysis.quantity:
+        if (
+            not len(config.ert.analysis) > 0
+            or not key in config.ert.analysis[0].quantity
+        ):
             raise ValueError(
                 f"Key {key} is not defined as an analysis quantity ({config.flownet.hyperopt.loss.keys})."
             )
 
     if (
-        config.ert.analysis.metric
-        and config.flownet.hyperopt.loss.metric not in config.ert.analysis.metric
+        hasattr(config.ert, "analysis")
+        and len(config.ert.analysis) > 0
+        and config.flownet.hyperopt.loss.metric not in config.ert.analysis[0].metric
     ):
         raise ValueError(
             f"Key {config.flownet.hyperopt.loss.metric} is not defined as an analysis"

src/flownet/ert/_create_ert_setup.py

@@ -206,13 +206,6 @@ def create_ert_setup(  # pylint: disable=too-many-arguments
     with open(output_folder / "parameters.pickled", "wb") as fh:
         pickle.dump(parameters, fh)
 
-    if hasattr(config.ert, "analysis"):
-        analysis_metric = "[" + ",".join(list(config.ert.analysis.metric)) + "]"
-        analysis_quantity = "[" + ",".join(list(config.ert.analysis.quantity)) + "]"
-    else:
-        analysis_metric = str(None)
-        analysis_quantity = str(None)
-
     configuration = {
         "pickled_network": output_folder.resolve() / "network.pickled",
         "pickled_schedule": output_folder.resolve() / "schedule.pickled",
@@ -221,8 +214,6 @@ def create_ert_setup(  # pylint: disable=too-many-arguments
         "random_seed": None,
         "debug": args.debug if hasattr(args, "debug") else False,
         "pred_schedule_file": getattr(config.ert, "pred_schedule_file", None),
-        "analysis_metric": analysis_metric,
-        "analysis_quantity": analysis_quantity,
     }
 
     if not prediction_setup:
@@ -263,16 +254,39 @@ def create_ert_setup(  # pylint: disable=too-many-arguments
         else config.ert.static_include_files
     )
 
-    shutil.copyfile(
-        _MODULE_FOLDER / ".." / "static" / "SAVE_ITERATION_ANALYTICS_WORKFLOW",
-        output_folder / "SAVE_ITERATION_ANALYTICS_WORKFLOW",
-    )
-
     shutil.copyfile(
         _MODULE_FOLDER / ".." / "static" / "SAVE_ITERATION_ANALYTICS_WORKFLOW_JOB",
         output_folder / "SAVE_ITERATION_ANALYTICS_WORKFLOW_JOB",
     )
 
+    analytics_workflow_template = _TEMPLATE_ENVIRONMENT.get_template(
+        "SAVE_ITERATION_ANALYTICS_WORKFLOW.jinja2"
+    )
+    if hasattr(config.ert, "analysis"):
+        for i, analysis_item in enumerate(config.ert.analysis):
+            with open(
+                output_folder / f"SAVE_ITERATION_ANALYTICS_WORKFLOW_{i}", "w"
+            ) as fh:
+                fh.write(
+                    analytics_workflow_template.render(
+                        {
+                            "reference_simulation": path_ref_sim,
+                            "perforation_strategy": config.flownet.perforation_handling_strategy,
+                            "run_path": config.ert.runpath,
+                            "ecl_base": config.ert.eclbase,
+                            "analysis_start": analysis_item.start,
+                            "analysis_end": analysis_item.end,
+                            "analysis_quantity": "["
+                            + ",".join(list(analysis_item.quantity))
+                            + "]",
+                            "analysis_metric": "["
+                            + ",".join(list(analysis_item.metric))
+                            + "]",
+                            "analysis_outfile": analysis_item.outfile,
+                        }
+                    )
+                )
+
     shutil.copyfile(args.config, output_folder / args.config.name)
     with open(os.path.join(output_folder, "pipfreeze.output"), "w") as fh:
         subprocess.call(["pip", "freeze"], stdout=fh)
@@ -285,6 +299,7 @@ def create_ert_setup(  # pylint: disable=too-many-arguments
         else:
             # Otherwise create an empty one.
             (output_folder / f"{section}.inc").touch()
+
     if not prediction_setup:
         if parameters is not None:
             create_observation_file(

src/flownet/hyperparameter/_run_hyper.py

@@ -88,7 +88,9 @@ def flownet_ahm_run(x: list, args: argparse.Namespace):
         run_flownet_history_matching(config, run_args)
 
         df_analytics = pd.read_csv(
-            (run_args.output_folder / config.ert.analysis.outfile).with_suffix(".csv")
+            (run_args.output_folder / config.ert.analysis[0].outfile).with_suffix(
+                ".csv"
+            )
         ).drop_duplicates()
 
         hyperopt_loss = 0.0

src/flownet/templates/SAVE_ITERATION_ANALYTICS_WORKFLOW.jinja2

@@ -0,0 +1 @@
+SAVE_ITERATION_ANALYTICS_WORKFLOW_JOB {{ reference_simulation }} {{ perforation_strategy }} {{ run_path }} {{ ecl_base }} {{ analysis_start }} {{ analysis_end }} {{ analysis_quantity }} {{ analysis_metric }} {{ analysis_outfile }}

src/flownet/templates/ahm_config.ert.jinja2

@@ -3,20 +3,11 @@
 INSTALL_JOB DELETE_IN_CURRENT_ITERATION ./DELETE_IN_CURRENT_ITERATION
 
 REFCASE <CONFIG_PATH>/SYNTHETIC_REFCASE
-{%- if config.ert.analysis.metric: %}
-DEFINE <ANALYSIS_METRIC> {{ analysis_metric }}
-DEFINE <ANALYSIS_QUANTITY> {{ analysis_quantity }}
-DEFINE <ANALYSIS_START> {{ config.ert.analysis.start }}
-DEFINE <ANALYSIS_END> {{ config.ert.analysis.end }}
-DEFINE <ANALYSIS_OUTFILE> {{ config.ert.analysis.outfile }}
-DEFINE <PERFORATION_STRATEGY> {{ config.flownet.perforation_handling_strategy }}
-DEFINE <REFERENCE_SIMULATION> {{ reference_simulation }}
-{%- endif %}
+
 DEFINE <RUN_PATH> {{ config.ert.runpath }}
 DEFINE <ECL_BASE> {{ config.ert.eclbase }}
 DEFINE <RANDOM_SAMPLES> ./parameters.json
 
-
 INSTALL_JOB CREATE_FLOWNET_MODEL ./CREATE_FLOWNET_MODEL
 {%- if not debug: %}
 INSTALL_JOB DELETE_IN_CURRENT_ITERATION ./DELETE_IN_CURRENT_ITERATION
@@ -25,10 +16,14 @@ INSTALL_JOB DELETE_IN_CURRENT_ITERATION ./DELETE_IN_CURRENT_ITERATION
 LOAD_WORKFLOW_JOB ./SAVE_ITERATION_PARAMETERS_WORKFLOW_JOB
 LOAD_WORKFLOW SAVE_ITERATION_PARAMETERS_WORKFLOW
 HOOK_WORKFLOW SAVE_ITERATION_PARAMETERS_WORKFLOW PRE_SIMULATION
-{%- if config.ert.analysis.metric: %}
+
+{%- if config.ert.analysis: %}
 LOAD_WORKFLOW_JOB ./SAVE_ITERATION_ANALYTICS_WORKFLOW_JOB
-LOAD_WORKFLOW SAVE_ITERATION_ANALYTICS_WORKFLOW
-HOOK_WORKFLOW SAVE_ITERATION_ANALYTICS_WORKFLOW POST_SIMULATION
+
+{% for item in config.ert.analysis %}
+LOAD_WORKFLOW SAVE_ITERATION_ANALYTICS_WORKFLOW_{{ loop.index - 1 }}
+HOOK_WORKFLOW SAVE_ITERATION_ANALYTICS_WORKFLOW_{{ loop.index - 1 }} POST_SIMULATION
+{% endfor %}
 {%- endif %}
 
 GEN_KW FLOWNET_PARAMETERS ./EMPTYFILE ./EMPTYFILE ./parameters.ertparam