Discrepancy between browser concat and Node concat

[thekevinscott]

System information

• Have I written custom code (as opposed to using a stock example script provided in TensorFlow.js): Yes
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Macbook 2019 browser and Node
• TensorFlow.js installed from (npm or script link): 3.13.0

Describe the current behavior
There is a discrepancy in the behavior of tf.concat between the browser and Node.

The following code works in the browser:

import * as tf from "@tensorflow/tfjs";

let ones = tf.ones([1, 2, 4, 3]);
let zeros = tf.zeros([1, 0, 2, 3]);
const final = zeros.concat(ones, 1);
console.log("final shape", final.shape);

In the console, this prints final shape (4) [1, 2, 4, 3].

The following code does not work in Node:

const tf = require("@tensorflow/tfjs-node");

let ones = tf.ones([1, 2, 4, 3]);
let zeros = tf.zeros([1, 0, 2, 3]);
const final = zeros.concat(ones, 1);
console.log(final.shape);

I get the following error:

            throw ex;
            ^

Error: Invalid TF_Status: 3
Message: ConcatOp : Dimensions of inputs should match: shape[0] = [1,0,2,3] vs. shape[1] = [1,2,4,3]

Describe the expected behavior
I would expect tf.concat to exhibit the same behavior on the browser and in Node.

Standalone code to reproduce the issue
Here is a link for the working browser code: https://codesandbox.io/s/tfjs-browser-concat-example-fie6e?file=/src/index.js
Here is a link for the failing Node code: https://codesandbox.io/s/tfjs-node-concat-example-frttt?file=/index.js

[google-ml-butler]

Are you satisfied with the resolution of your issue?
Yes
No