implement evaluation routine for SSL

config/default_forecast_config.yml

@@ -182,4 +182,4 @@ wgtags:
   # Examples: "rollout_ablation_grid"  
   exp: None
   # *** Experiment-specific tags ***
-  grid: None
+  grid: None

packages/evaluate/pyproject.toml

@@ -29,6 +29,7 @@ dev = [
 ]
 
 [project.scripts]
+ssl_analysis= "weathergen.evaluate.ssl.ssl_eval:ssl_analysis"
 evaluation = "weathergen.evaluate.run_evaluation:evaluate"
 export = "weathergen.evaluate.export.export_inference:export"
 

packages/evaluate/src/weathergen/evaluate/io/csv_reader.py

@@ -8,7 +8,6 @@
 # nor does it submit to any jurisdiction.
 
 # Standard library
-import logging
 import re
 from pathlib import Path
 
@@ -20,16 +19,15 @@
 # Local application / package
 from weathergen.evaluate.io.io_reader import Reader
 
-_logger = logging.getLogger(__name__)
-_logger.setLevel(logging.INFO)
-
 
 class CsvReader(Reader):
     """
     Reader class to read evaluation data from CSV files and convert to xarray DataArray.
     """
 
-    def __init__(self, eval_cfg: dict, run_id: str, private_paths: dict | None = None):
+    def __init__(
+        self, eval_cfg: dict, run_id: str, private_paths: dict | None = None, verbose=True
+    ):
         """
         Initialize the CsvReader.
 
@@ -43,7 +41,7 @@ def __init__(self, eval_cfg: dict, run_id: str, private_paths: dict | None = Non
             list of private paths for the supported HPC
         """
 
-        super().__init__(eval_cfg, run_id, private_paths)
+        super().__init__(eval_cfg, run_id, private_paths, verbose)
         self.metrics_dir = Path(self.eval_cfg.get("metrics_dir"))
 
         self.metrics_base_dir = self.metrics_dir

packages/evaluate/src/weathergen/evaluate/io/io_reader.py

@@ -15,9 +15,6 @@
 # Third-party
 import xarray as xr
 
-_logger = logging.getLogger(__name__)
-_logger.setLevel(logging.INFO)
-
 
 @dataclass
 class ReaderOutput:
@@ -62,7 +59,9 @@ class DataAvailability:
 
 
 class Reader:
-    def __init__(self, eval_cfg: dict, run_id: str, private_paths: dict[str, str] | None = None):
+    def __init__(
+        self, eval_cfg: dict, run_id: str, private_paths: dict[str, str] | None = None, verbose=True
+    ):
         """
         Generic data reader class.
 
@@ -89,6 +88,10 @@ def __init__(self, eval_cfg: dict, run_id: str, private_paths: dict[str, str] |
             "results_base_dir", None
         )  # base directory where results will be stored
 
+        self._logger = logging.getLogger(__name__)
+        logger_level = logging.INFO if verbose else logging.CRITICAL + 1
+        self._logger.setLevel(logger_level)
+
     def get_stream(self, stream: str):
         """
         returns the dictionary associated to a particular stream
@@ -217,7 +220,7 @@ def check_availability(
                 requested[name] = reader_data[name]
                 # If file with metrics exists, must exactly match
                 if available_data is not None and reader_data[name] != available[name]:
-                    _logger.info(
+                    self._logger.info(
                         f"Requested all {name}s for {mode}, but previous config was a "
                         "strict subset. Recomputation required."
                     )
@@ -230,7 +233,7 @@ def check_availability(
                 if name == "ensemble" and "mean" in missing:
                     missing.remove("mean")
                 if missing:
-                    _logger.info(
+                    self._logger.info(
                         f"Requested {name}(s) {missing} is unavailable. "
                         f"Removing missing {name}(s) for {mode}."
                     )
@@ -240,15 +243,15 @@ def check_availability(
             # Must be a subset of available_data (if provided)
             if available_data is not None and not requested[name] <= available[name]:
                 missing = requested[name] - available[name]
-                _logger.info(
+                self._logger.info(
                     f"{name.capitalize()}(s) {missing} missing in previous evaluation."
                     "Recomputation required."
                 )
                 check_score = False
 
         if check_score and not corrected:
             scope = "metric file" if available_data is not None else "Zarr file"
-            _logger.info(
+            self._logger.info(
                 f"All checks passed – All channels, samples, fsteps requested for {mode} are "
                 f"present in {scope}..."
             )