Model can be run in tf but failed in tfjs. Error: TensorListResize input size 1 is greater maxNumElement 0.

Model can be run in tf but failed in tfjs after using tensorflowjs.converters.

System information

• Have I written custom code (as opposed to using a stock example script provided in TensorFlow.js):
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04):ubuntu 18.04
• TensorFlow.js installed from (npm or script link): npm install @tensorflow/tfjs
• TensorFlow.js version (use command below): 3.18
• Tensorflow.js Converter Version:3.18

Repro code:

from tensorflowjs.converters import converter
import subprocess

# relative path
graph_def_path = "test_tf.pb"
tfjs_path = "test_tfjs"
input_json_path = "input.json"

output_names = "i:0,output_ta:0,output_ta2:0"
converter.convert([graph_def_path, tfjs_path, '--input_format', 'tf_frozen_model',
                               '--output_node_names', output_names])

# node run_tfjs.js test_tfjs/model.json input.json output.json
command = ['node', 'run_tfjs.js', tfjs_path + '/model.json', input_json_path, 'output.json']
result = subprocess.run(command)

'''
Get error ----- Error: TensorListResize input size 1 is greater maxNumElement 0.
'''

related model and code: test.zip

The model is basically from https://github.com/onnx/tensorflow-onnx/blob/e9b6cb4fca61c87fcdaf4e9610535910c01ea4f9/tests/test_loops.py#L229.

Describe the current behavior

Error: TensorListResize input size 1 is greater maxNumElement 0.

Describe the expected behavior Script can be run without any error.

[rthadur]

@pyu10055 Thank you, related PR has been merged , changes will be available once there is a new release. Thank you

[hwangdeyu]

@pyu10055 Thank you, related PR has been merged , changes will be available once there is a new release. Thank you

Nice! I’m looking forward to the next release.

Thanks for the time you @pyu10055 @rthadur spent onto this.