Merge pull request #20 from zenml-io/bugfix/trust_remote_code_in_load_metrics

add trust_remote_code=True

Author: avishniakov

template/steps/dataset_loader/data_loader.py

@@ -32,14 +32,23 @@ def data_loader(
 
     # Load dataset based on the dataset value
     {%- if dataset == 'financial_news' %}
-    dataset = load_dataset("zeroshot/twitter-financial-news-sentiment")
+    dataset = load_dataset(
+        "zeroshot/twitter-financial-news-sentiment",
+        trust_remote_code=True,
+    )
     {%- endif %}
     {%- if dataset == 'imdb_reviews' %}
-    dataset = load_dataset("imdb")["train"]
+    dataset = load_dataset(
+        "imdb",
+        trust_remote_code=True,
+    )["train"]
     dataset = dataset.train_test_split(test_size=0.25, shuffle=True)
     {%- endif %}
     {%- if dataset == 'airline_reviews' %}
-    dataset = load_dataset("Shayanvsf/US_Airline_Sentiment")
+    dataset = load_dataset(
+        "Shayanvsf/US_Airline_Sentiment",
+        trust_remote_code=True,
+    )
     dataset = dataset.rename_column("airline_sentiment", "label")
     dataset = dataset.remove_columns(["airline_sentiment_confidence","negativereason_confidence"])
     {%- endif %}

template/utils/misc.py

@@ -18,13 +18,13 @@ def compute_metrics(eval_pred: Tuple[np.ndarray, np.ndarray]) -> Dict[str, float
     logits, labels = eval_pred
     predictions = np.argmax(logits, axis=-1)
     # calculate the mertic using the predicted and true value
-    accuracy = load_metric("accuracy").compute(
+    accuracy = load_metric("accuracy", trust_remote_code=True).compute(
         predictions=predictions, references=labels
     )
-    f1 = load_metric("f1").compute(
+    f1 = load_metric("f1", trust_remote_code=True).compute(
         predictions=predictions, references=labels, average="weighted"
     )
-    precision = load_metric("precision").compute(
+    precision = load_metric("precision", trust_remote_code=True).compute(
         predictions=predictions, references=labels, average="weighted"
     )
     return {