Remove deprecated argument from tests and examples (#1072)

* Remove deprecated argument from tests and examples

* fix style

examples/onnxruntime/optimization/multiple-choice/run_swag.py

@@ -248,7 +248,7 @@ def main():
     )
 
     # Export the model
-    model = ORTModelForMultipleChoice.from_pretrained(model_args.model_name_or_path, from_transformers=True)
+    model = ORTModelForMultipleChoice.from_pretrained(model_args.model_name_or_path, export=True)
 
     # Create the optimizer
     optimizer = ORTOptimizer.from_pretrained(model)

examples/onnxruntime/optimization/question-answering/run_qa.py

@@ -317,7 +317,7 @@ def main():
     )
 
     # Export the model
-    model = ORTModelForQuestionAnswering.from_pretrained(model_args.model_name_or_path, from_transformers=True)
+    model = ORTModelForQuestionAnswering.from_pretrained(model_args.model_name_or_path, export=True)
 
     # Create the optimizer
     optimizer = ORTOptimizer.from_pretrained(model)

examples/onnxruntime/optimization/text-classification/run_glue.py

@@ -262,7 +262,7 @@ def main():
     )
 
     # Export the model
-    model = ORTModelForSequenceClassification.from_pretrained(model_args.model_name_or_path, from_transformers=True)
+    model = ORTModelForSequenceClassification.from_pretrained(model_args.model_name_or_path, export=True)
 
     # Create the optimizer
     optimizer = ORTOptimizer.from_pretrained(model)

examples/onnxruntime/optimization/token-classification/run_ner.py

@@ -288,7 +288,7 @@ def main():
     )
 
     # Export the model
-    model = ORTModelForTokenClassification.from_pretrained(model_args.model_name_or_path, from_transformers=True)
+    model = ORTModelForTokenClassification.from_pretrained(model_args.model_name_or_path, export=True)
 
     # Create the optimizer
     optimizer = ORTOptimizer.from_pretrained(model)

examples/onnxruntime/quantization/image-classification/run_image_classification.py

@@ -293,7 +293,7 @@ def compute_metrics(p: EvalPrediction):
         return result
 
     # Export the model
-    model = ORTModelForImageClassification.from_pretrained(model_args.model_name_or_path, from_transformers=True)
+    model = ORTModelForImageClassification.from_pretrained(model_args.model_name_or_path, export=True)
 
     # Create the quantizer
     quantizer = ORTQuantizer.from_pretrained(model)

examples/onnxruntime/quantization/multiple-choice/run_swag.py

@@ -336,7 +336,7 @@ def compute_metrics(eval_predictions):
     )
 
     # Export the model
-    model = ORTModelForMultipleChoice.from_pretrained(model_args.model_name_or_path, from_transformers=True)
+    model = ORTModelForMultipleChoice.from_pretrained(model_args.model_name_or_path, export=True)
 
     # Create the quantizer
     quantizer = ORTQuantizer.from_pretrained(model)

examples/onnxruntime/quantization/question-answering/run_qa.py

@@ -537,7 +537,7 @@ def compute_metrics(p: EvalPrediction):
     tokenizer = AutoTokenizer.from_pretrained(model_args.model_name_or_path)
 
     # Export the model
-    model = ORTModelForQuestionAnswering.from_pretrained(model_args.model_name_or_path, from_transformers=True)
+    model = ORTModelForQuestionAnswering.from_pretrained(model_args.model_name_or_path, export=True)
 
     # Create the quantizer
     quantizer = ORTQuantizer.from_pretrained(model)

examples/onnxruntime/quantization/text-classification/run_glue.py

@@ -392,7 +392,7 @@ def compute_metrics(p: EvalPrediction):
     tokenizer = AutoTokenizer.from_pretrained(model_args.model_name_or_path)
 
     # Export the model
-    model = ORTModelForSequenceClassification.from_pretrained(model_args.model_name_or_path, from_transformers=True)
+    model = ORTModelForSequenceClassification.from_pretrained(model_args.model_name_or_path, export=True)
 
     # Create the quantizer
     quantizer = ORTQuantizer.from_pretrained(model)

examples/onnxruntime/quantization/token-classification/run_ner.py

@@ -480,7 +480,7 @@ def compute_metrics(p):
     tokenizer = AutoTokenizer.from_pretrained(model_args.model_name_or_path)
 
     # Export the model
-    model = ORTModelForTokenClassification.from_pretrained(model_args.model_name_or_path, from_transformers=True)
+    model = ORTModelForTokenClassification.from_pretrained(model_args.model_name_or_path, export=True)
 
     # Create the quantizer
     quantizer = ORTQuantizer.from_pretrained(model)

optimum/onnxruntime/modeling_ort.py

@@ -637,7 +637,7 @@ def from_pretrained(
 
         use_merged (`Optional[bool]`, defaults to `None`):
             whether or not to use a single ONNX that handles both the decoding without and with past key values reuse. This option defaults
-            to `True` if loading from a local repository and a merged decoder is found. When exporting with `from_transformers=True`,
+            to `True` if loading from a local repository and a merged decoder is found. When exporting with `export=True`,
             defaults to `False`. This option should be set to `True` to minimize memory usage.
 
         Returns:
@@ -1391,7 +1391,7 @@ def forward(
     >>> from optimum.onnxruntime import {model_class}
 
     >>> tokenizer = {processor_class}.from_pretrained("{checkpoint}")
-    >>> model = {model_class}.from_pretrained("{checkpoint}", from_transformers=True)
+    >>> model = {model_class}.from_pretrained("{checkpoint}", export=True)
 
     >>> num_choices = 4
     >>> first_sentence = ["Members of the procession walk down the street holding small horn brass instruments."] * num_choices

optimum/onnxruntime/modeling_seq2seq.py

@@ -248,7 +248,7 @@
 
     >>> processor = {processor_class}.from_pretrained("{checkpoint}")
     >>> tokenizer = {tokenizer_class}.from_pretrained("{checkpoint}")
-    >>> model = {model_class}.from_pretrained("{checkpoint}", from_transformers=True)
+    >>> model = {model_class}.from_pretrained("{checkpoint}", export=True)
 
     >>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
     >>> image = Image.open(requests.get(url, stream=True).raw)
@@ -270,7 +270,7 @@
 
     >>> processor = {processor_class}.from_pretrained("{checkpoint}")
     >>> tokenizer = {tokenizer_class}.from_pretrained("{checkpoint}")
-    >>> model = {model_class}.from_pretrained("{checkpoint}", from_transformers=True)
+    >>> model = {model_class}.from_pretrained("{checkpoint}", export=True)
 
     >>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
     >>> image = Image.open(requests.get(url, stream=True).raw)

optimum/onnxruntime/optimization.py

@@ -102,7 +102,7 @@ def from_pretrained(
                     raise NotImplementedError(
                         "ORTOptimizer does not support ORTModelForCausalLM models that use a single ONNX for both the without/with past cases."
                         " Please pass an ORTModelForCausalLM that uses a separate ONNX for each without/with past cases. This can be done"
-                        " by using `ORTModelForCausalLM.from_pretrained(..., from_transformers=True, use_merged=False)`, or by"
+                        " by using `ORTModelForCausalLM.from_pretrained(..., export=True, use_merged=False)`, or by"
                         " using the option `--no-post-process` in the optimum-cli ONNX export tool."
                     )
                 onnx_model_path.append(model_or_path.decoder_model_path)

optimum/onnxruntime/runs/__init__.py

@@ -36,7 +36,7 @@ def __init__(self, run_config):
         )
 
         onnx_model = ORT_SUPPORTED_TASKS[self.task]["class"][0].from_pretrained(
-            run_config["model_name_or_path"], from_transformers=True
+            run_config["model_name_or_path"], export=True
         )
 
         trfs_model = FeaturesManager.get_model_from_feature(

optimum/pipelines/pipelines_base.py

@@ -229,7 +229,7 @@ def load_ort_pipeline(
 
     if model is None:
         model_id = SUPPORTED_TASKS[targeted_task]["default"]
-        model = SUPPORTED_TASKS[targeted_task]["class"][0].from_pretrained(model_id, from_transformers=True)
+        model = SUPPORTED_TASKS[targeted_task]["class"][0].from_pretrained(model_id, export=True)
     elif isinstance(model, str):
         from ..onnxruntime.modeling_seq2seq import ENCODER_ONNX_FILE_PATTERN, ORTModelForConditionalGeneration
 
@@ -249,8 +249,8 @@ def load_ort_pipeline(
             use_auth_token=use_auth_token,
             revision=revision,
         )
-        from_transformers = len(onnx_files) == 0
-        model = ort_model_class.from_pretrained(model, from_transformers=from_transformers, **model_kwargs)
+        export = len(onnx_files) == 0
+        model = ort_model_class.from_pretrained(model, export=export, **model_kwargs)
     elif isinstance(model, ORTModel):
         if tokenizer is None and load_tokenizer:
             for preprocessor in model.preprocessors: