Add cache group field to cache prefix.

scripts/calculate_values.py

@@ -18,7 +18,6 @@
 import os
 import pickle
 import time
-from pathlib import Path
 
 import click
 import numpy as np
@@ -95,9 +94,13 @@ def _calculate_values(
         )
 
     params = load_params_fast()
-    cache = PrefixedMemcachedCacheBackend(
-        config=MemcachedClientConfig(), prefix=f"{experiment_name}/{dataset_name}"
-    )
+    cache = None
+    if "cache_group" in params["valuation_methods"][valuation_method_name]:
+        cache_group = params["valuation_methods"][valuation_method_name]["cache_group"]
+        prefix = f"{experiment_name}/{dataset_name}/{cache_group}"
+        cache = PrefixedMemcachedCacheBackend(
+            config=MemcachedClientConfig(), prefix=prefix
+        )
 
     val_set = Accessor.datasets(experiment_name, dataset_name).loc[0, "val_set"]
 

scripts/evaluate_metrics.py

@@ -13,16 +13,14 @@
 directory. The metrics are usually stored as `*.csv` files. Each metric consists of
 a single value and a curve. The curve is stored as `*.curve.csv` file.
 """
-
 import logging
 import os
 from functools import partial, reduce
-from pathlib import Path
 
 import click
 import pandas as pd
 from pydvl.parallel import ParallelConfig
-from pydvl.utils import DiskCacheBackend, MemcachedClientConfig
+from pydvl.utils import MemcachedClientConfig
 from pydvl.utils.functional import maybe_add_argument
 
 from re_classwise_shapley.cache import PrefixedMemcachedCacheBackend
@@ -127,9 +125,16 @@ def _evaluate_metrics(
 
     n_pipeline_step = 5
     seed = pipeline_seed(repetition_id, n_pipeline_step)
-    cache = PrefixedMemcachedCacheBackend(
-        config=MemcachedClientConfig(), prefix=f"{experiment_name}/{dataset_name}"
-    )
+    cache = None
+    if (
+        "eval_model" in metric_kwargs
+        and "cache_group" in params["valuation_methods"][valuation_method_name]
+    ):
+        cache_group = params["valuation_methods"][valuation_method_name]["cache_group"]
+        prefix = f"{experiment_name}/{dataset_name}/{cache_group}"
+        cache = PrefixedMemcachedCacheBackend(
+            config=MemcachedClientConfig(), prefix=prefix
+        )
 
     logger.info("Evaluating metric...")
     with n_threaded(n_threads=1):

src/re_classwise_shapley/cache.py

@@ -4,9 +4,9 @@
 
 
 class PrefixedMemcachedCacheBackend(MemcachedCacheBackend):
-    def __init__(self, *args, prefix: str, **kwargs):
+    def __init__(self, *args, **kwargs):
+        self._prefix = kwargs["prefix"]
         super().__init__(*args, **kwargs)
-        self._prefix = prefix
 
     def get(self, key: str) -> Optional[Any]:
         return super().get(f"{self._prefix}/{key}")