chore: import upstream snapshot with attribution

examples/transformers/load_components.py

@@ -0,0 +1,31 @@
+import transformers
+
+import mlflow
+
+pipeline = transformers.pipeline(
+    task="fill-mask",
+    model=transformers.AutoModelForMaskedLM.from_pretrained("distilbert-base-uncased"),
+    tokenizer=transformers.AutoTokenizer.from_pretrained("distilbert-base-uncased"),
+)
+
+with mlflow.start_run():
+    model_info = mlflow.transformers.log_model(
+        transformers_model=pipeline,
+        name="mask_filler",
+        input_example="MLflow is [MASK]!",
+    )
+
+components = mlflow.transformers.load_model(model_info.model_uri, return_type="components")
+
+for key, value in components.items():
+    print(f"{key} -> {type(value).__name__}")
+
+response = pipeline("MLflow is [MASK]!")
+
+print(response)
+
+reconstructed_pipeline = transformers.pipeline(**components)
+
+reconstructed_response = reconstructed_pipeline("Transformers is [MASK]!")
+
+print(reconstructed_response)