Handle both 'label' and 'labels' column names in datasets

train-image-classifier.py

@@ -210,23 +210,26 @@ def _(dataset_name, mo):
     from datasets import load_dataset
 
     print(f"Loading dataset: {dataset_name}...")
-    dataset = load_dataset(dataset_name, trust_remote_code=True)
+    dataset = load_dataset(dataset_name)
     print(f"Train: {len(dataset['train']):,} samples")
     print(f"Test: {len(dataset['test']):,} samples")
 
-    # Get label info
-    label_feature = dataset["train"].features["label"]
+    # Get label column name (datasets use 'label' or 'labels')
+    _features = dataset["train"].features
+    label_column = "label" if "label" in _features else "labels"
+    label_feature = _features[label_column]
     labels = label_feature.names if hasattr(label_feature, "names") else None
-    num_labels = label_feature.num_classes if hasattr(label_feature, "num_classes") else len(set(dataset["train"]["label"]))
+    num_labels = label_feature.num_classes if hasattr(label_feature, "num_classes") else len(set(dataset["train"][label_column]))
 
+    print(f"Label column: '{label_column}'")
     print(f"Labels ({num_labels}): {labels}")
 
     mo.md(f"**Loaded {len(dataset['train']):,} training samples with {num_labels} classes**")
-    return dataset, labels, num_labels
+    return dataset, label_column, labels, num_labels
 
 
 @app.cell
-def _(dataset, labels, mo):
+def _(dataset, label_column, labels, mo):
     # Show sample images (notebook mode only)
     import base64 as _base64
     from io import BytesIO as _BytesIO
@@ -245,7 +248,7 @@ def _(dataset, labels, mo):
     _images_html = []
     for _sample in _samples:
         _img_b64 = _image_to_base64(_sample["image"])
-        _label_name = labels[_sample["label"]] if labels else _sample["label"]
+        _label_name = labels[_sample[label_column]] if labels else _sample[label_column]
         _images_html.append(
             f"""
             <div style="text-align: center; margin: 5px;">
@@ -311,12 +314,12 @@ def _(mo):
 
 
 @app.cell
-def _(dataset, image_processor):
+def _(dataset, image_processor, label_column):
     def preprocess(examples):
         """Apply image processor to batch of images."""
         images = [img.convert("RGB") for img in examples["image"]]
         inputs = image_processor(images, return_tensors="pt")
-        inputs["label"] = examples["label"]
+        inputs["labels"] = examples[label_column]  # Trainer expects 'labels'
         return inputs
 
     print("Preprocessing dataset...")
@@ -418,7 +421,7 @@ def _(trainer):
 
 
 @app.cell
-def _(dataset, id2label, image_processor, mo, model):
+def _(dataset, id2label, image_processor, label_column, mo, model):
     import torch
     import base64 as _b64
     from io import BytesIO as _BIO
@@ -436,9 +439,10 @@ def _(dataset, id2label, image_processor, mo, model):
             _outputs = model(**_inputs)
             _pred_idx = _outputs.logits.argmax(-1).item()
 
-        _true_label = id2label[_sample["label"]] if id2label else _sample["label"]
+        _true_idx = _sample[label_column]
+        _true_label = id2label[_true_idx] if id2label else _true_idx
         _pred_label = id2label[_pred_idx] if id2label else _pred_idx
-        _correct = "correct" if _pred_idx == _sample["label"] else "wrong"
+        _correct = "correct" if _pred_idx == _true_idx else "wrong"
 
         # Convert image for display
         _img_copy = _img.copy()