Remove framework mention (#12731)
This commit is contained in:
@@ -72,8 +72,6 @@ The package may be used as a Python module:
|
||||
-h, --help show this help message and exit
|
||||
-m MODEL, --model MODEL
|
||||
Model's name of path on disk to load.
|
||||
-f {pytorch}, --framework {pytorch}
|
||||
Framework to use when exporting. Possible values are: {'pytorch'}
|
||||
--features {default} Export the model with some additional features.
|
||||
--opset OPSET ONNX opset version to export the model with (default 12).
|
||||
--atol ATOL Absolute difference tolerance when validating the model.
|
||||
@@ -82,7 +80,7 @@ Exporting a checkpoint using a ready-made configuration can be done as follows:
|
||||
|
||||
.. code-block::
|
||||
|
||||
python -m transformers.onnx -f pytorch --model=bert-base-cased onnx/bert-base-cased/
|
||||
python -m transformers.onnx --model=bert-base-cased onnx/bert-base-cased/
|
||||
|
||||
This exports an ONNX graph of the mentioned checkpoint. Here it is `bert-base-cased`, but it can be any model from the
|
||||
hub, or a local path.
|
||||
|
||||
Reference in New Issue
Block a user