Now supporting pathlike in pipelines too. (#20030)

src/transformers/pipelines/__init__.py

@@ -21,6 +21,7 @@ import os
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import warnings
+from pathlib import Path
 from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple, Union
 
 from numpy import isin
@@ -638,6 +639,8 @@ def pipeline(
             " feature_extractor may not be compatible with the default model. Please provide a PreTrainedModel class"
             " or a path/identifier to a pretrained model when providing feature_extractor."
         )
+    if isinstance(model, Path):
+        model = str(model)
 
     # Config is the primordial information item.
     # Instantiate config if needed

tests/pipelines/test_pipelines_common.py

@@ -356,6 +356,15 @@ class CommonPipelineTest(unittest.TestCase):
         self.assertEqual(pipe._batch_size, 2)
         self.assertEqual(pipe._num_workers, 1)
 
+    @require_torch
+    def test_pipeline_pathlike(self):
+        pipe = pipeline(model="hf-internal-testing/tiny-random-distilbert")
+        with tempfile.TemporaryDirectory() as d:
+            pipe.save_pretrained(d)
+            path = Path(d)
+            newpipe = pipeline(task="text-classification", model=path)
+        self.assertIsInstance(newpipe, TextClassificationPipeline)
+
     @require_torch
     def test_pipeline_override(self):
         class MyPipeline(TextClassificationPipeline):