Set cache_dir for evaluate.load() in example scripts (#28422)

While using `run_clm.py`,[^1] I noticed that some files were being added
to my global cache, not the local cache. I set the `cache_dir` parameter
for the one call to `evaluate.load()`, which partially solved the
problem. I figured that while I was fixing the one script upstream, I
might as well fix the problem in all other example scripts that I could.

There are still some files being added to my global cache, but this
appears to be a bug in `evaluate` itself. This commit at least moves
some of the files into the local cache, which is better than before.

To create this PR, I made the following regex-based transformation:
`evaluate\.load\((.*?)\)` -> `evaluate\.load\($1,
cache_dir=model_args.cache_dir\)`. After using that, I manually fixed
all modified files with `ruff` serving as useful guidance. During the
process, I removed one existing usage of the `cache_dir` parameter in a
script that did not have a corresponding `--cache-dir` argument
declared.

[^1]: I specifically used `pytorch/language-modeling/run_clm.py` from
v4.34.1 of the library. For the original code, see the following URL:
acc394c4f5/examples/pytorch/language-modeling/run_clm.py.

examples/pytorch/text-classification/run_classification.py

@@ -633,23 +633,23 @@ def main():
 
     if data_args.metric_name is not None:
         metric = (
-            evaluate.load(data_args.metric_name, config_name="multilabel")
+            evaluate.load(data_args.metric_name, config_name="multilabel", cache_dir=model_args.cache_dir)
             if is_multi_label
-            else evaluate.load(data_args.metric_name)
+            else evaluate.load(data_args.metric_name, cache_dir=model_args.cache_dir)
         )
         logger.info(f"Using metric {data_args.metric_name} for evaluation.")
     else:
         if is_regression:
-            metric = evaluate.load("mse")
+            metric = evaluate.load("mse", cache_dir=model_args.cache_dir)
             logger.info("Using mean squared error (mse) as regression score, you can use --metric_name to overwrite.")
         else:
             if is_multi_label:
-                metric = evaluate.load("f1", config_name="multilabel")
+                metric = evaluate.load("f1", config_name="multilabel", cache_dir=model_args.cache_dir)
                 logger.info(
                     "Using multilabel F1 for multi-label classification task, you can use --metric_name to overwrite."
                 )
             else:
-                metric = evaluate.load("accuracy")
+                metric = evaluate.load("accuracy", cache_dir=model_args.cache_dir)
                 logger.info("Using accuracy as classification score, you can use --metric_name to overwrite.")
 
     def compute_metrics(p: EvalPrediction):

examples/pytorch/text-classification/run_glue.py

@@ -514,11 +514,11 @@ def main():
 
     # Get the metric function
     if data_args.task_name is not None:
-        metric = evaluate.load("glue", data_args.task_name)
+        metric = evaluate.load("glue", data_args.task_name, cache_dir=model_args.cache_dir)
     elif is_regression:
-        metric = evaluate.load("mse")
+        metric = evaluate.load("mse", cache_dir=model_args.cache_dir)
     else:
-        metric = evaluate.load("accuracy")
+        metric = evaluate.load("accuracy", cache_dir=model_args.cache_dir)
 
     # You can define your custom compute_metrics function. It takes an `EvalPrediction` object (a namedtuple with a
     # predictions and label_ids field) and has to return a dictionary string to float.

examples/pytorch/text-classification/run_xnli.py

@@ -385,7 +385,7 @@ def main():
             )
 
     # Get the metric function
-    metric = evaluate.load("xnli")
+    metric = evaluate.load("xnli", cache_dir=model_args.cache_dir)
 
     # You can define your custom compute_metrics function. It takes an `EvalPrediction` object (a namedtuple with a
     # predictions and label_ids field) and has to return a dictionary string to float.