Add specific warning for exports with sequence_length set to 1

optimum/exporters/onnx/__main__.py

@@ -317,15 +317,6 @@ def main_export(
         model_name_or_path, subfolder=subfolder, library_name=library_name
     )
 
-    # get the shapes to be used to generate dummy inputs
-    input_shapes = {}
-    for input_name in DEFAULT_DUMMY_SHAPES.keys():
-        input_shapes[input_name] = (
-            kwargs_shapes[input_name] if input_name in kwargs_shapes else DEFAULT_DUMMY_SHAPES[input_name]
-        )
-
-    torch_dtype = None if fp16 is False else torch.float16
-
     if task == "auto":
         try:
             task = TasksManager.infer_task_from_model(model_name_or_path)
@@ -338,6 +329,21 @@ def main_export(
                 f"The task could not be automatically inferred as this is available only for models hosted on the Hugging Face Hub. Please provide the argument --task with the relevant task from {', '.join(TasksManager.get_all_tasks())}. Detailed error: {e}"
             )
 
+    # get the shapes to be used to generate dummy inputs
+    input_shapes = {}
+    for input_name in DEFAULT_DUMMY_SHAPES.keys():
+        input_shapes[input_name] = (
+            kwargs_shapes[input_name] if input_name in kwargs_shapes else DEFAULT_DUMMY_SHAPES[input_name]
+        )
+        if (
+            input_name == "sequence_length"
+            and kwargs_shapes.get(input_name) == 1
+            and task.startswith("text-generation")
+        ):
+            logger.warning("Exporting with a sequence length of 1 for text generation models is not supported and can yield unexpected results.")
+
+    torch_dtype = None if fp16 is False else torch.float16
+
     if library_name == "transformers":
         config = AutoConfig.from_pretrained(
             model_name_or_path,