torch export to onnx extern data size configurable #1481
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
albertz
reviewed
Dec 23, 2023
| @@ -182,6 +182,13 @@ def main(): | |||
| parser.add_argument("out_onnx_filename", type=str, help="Filename of the final ONNX model.") | |||
| parser.add_argument("--verbosity", default=4, type=int, help="5 for all seqs (default: 4)") | |||
| parser.add_argument("--device", type=str, default="cpu", help="'cpu' (default) or 'gpu'.") | |||
| parser.add_argument( | |||
| "--dyn_dim_min_sizes", type=dict, default=None, help="Specify min sizes for dim tags with dynamic sizes" | |||
There was a problem hiding this comment.
Did you test this? What exactly did you pass as argument? I wonder how it would correctly convert it into a dict, and specifically with the right Dim objects as keys?
There was a problem hiding this comment.
I tested it by setting the Dim object like this
time_data_dim = SpatialDim('time:data')
extern_data_dict = {'data': {
'dim_tags': [
batch_dim,
time_data_dim,
FeatureDim('feat', 1),
],
'dtype': 'float32',
}}
dyn_dim_max_sizes={time_data_dim:100}
There was a problem hiding this comment.
I mean, how did you start this script? How did you pass --dyn_dim_min_sizes when starting the script?
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR adds
dyn_dim_min_sizesanddyn_dim_max_sizesto the command line options oftools/torch_export_to_onnx.py. In default, extern data with time dimension size in range [2,25] would be generated for export, which might be too short for the cases where raw wave input is used. Thus these two options can be used to specify the min/max sizes for dim tags with dynamic sizes.