Tfjs: WebGL CPU forwarding failed for slice op

[pyu10055]

Please make sure that this is a bug. As per our
GitHub Policy,
we only address code/doc bugs, performance issues, feature requests and
build/installation issues on GitHub. tag:bug_template

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):
• Mobile device (e.g. iPhone 8, Pixel 2, Samsung Galaxy) if the issue happens on mobile device:
• TensorFlow.js installed from (npm or script link): both
• TensorFlow.js version (use command below): 2.5.0
• Browser version:
• Tensorflow.js Converter Version: 2.5.0

Describe the current behavior
WebGL CPU forward failed to slice op, caused by CPU slice op has been modularized.

backend.ts:665 Uncaught (in promise) Error: 'slice' not yet implemented or not found in the registry. This kernel may not be supported by the tfjs backend you have chosen
    at notYetImplemented (backend.ts:665)
    at MathBackendCPU.KernelBackend.slice (backend.ts:145)
    at MathBackendWebGL.slice (backend_webgl.ts:731)
    at forward (slice.ts:75)
    at engine.ts:625

The current work-around is to disable CPU forwarding:

tf.env().set('WEBGPU_CPU_FORWARD', false);

Describe the expected behavior
In 2.4.0 the CPU forwarding of slice op is not failing.

Standalone code to reproduce the issue
Provide a reproducible test case that is the bare minimum necessary to generate
the problem. If possible, please share a link to Colab/CodePen/any notebook.

Other info / logs Include any logs or source code that would be helpful to
diagnose the problem. If including tracebacks, please include the full
traceback. Large logs and files should be attached.

[bertlea]

I got the similar error today:

tfjs:17 Uncaught (in promise) Error: 'slice' not yet implemented or not found in the registry. This kernel may not be supported by the tfjs backend you have chosen at av (tfjs:17) at t.slice (tfjs:17) at t.n.slice (tfjs:17) at tfjs:17 at tfjs:17 at tfjs:17 at e.t.scopedRun (tfjs:17) at e.t.tidy (tfjs:17) at p (tfjs:17) at tfjs:17

I guess the cdn just using the latest release. I resolved the issue by changing:
from:
importScripts('https://cdn.jsdelivr.net/npm/@tensorflow/tfjs');
to:
importScripts('https://cdn.jsdelivr.net/npm/@tensorflow/tfjs-core'); importScripts('https://cdn.jsdelivr.net/npm/@tensorflow/tfjs-converter'); importScripts('https://cdn.jsdelivr.net/npm/@tensorflow/tfjs-backend-webgl');
Is that expected? I think it is a bug.

[mtuan4i]

I got the same error. I tried tf.env().set('WEBGPU_CPU_FORWARD', false); but it doesn't work. However, tf.env().set('WEBGL_CPU_FORWARD', false); does

[lnovelli]

+1 same issue happening here after upgrading "@tensorflow/tfjs" from 2.4.0 to 2.5.0
adding tf.env().set('WEBGL_CPU_FORWARD', false); solves...

[jmartinezuk]

+1 Same here.

[Fifth-marauder]

+1 same here

[tensorflow-butler[bot]]

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

[plakal]

When will the fix be available in a public release? I'm still seeing the issue when using one of the official mirrors of TFJS: https://cdn.jsdelivr.net/npm/@tensorflow/tfjs/dist/tf.min.js which still seems to be pulling down 2.5.0

[lina128]

We published 2.6.0 today, it should fix the problem by now.

[lnovelli]

Problem solved on my side with 2.6.0! Kudos to the TFjs team for the prompt resolution! :)

[rfourdinier]

Hi, I had the same error that is resolved with tfjs 2.6.0 but now I still have an other error when executing my model.

The model's input is a tensor created from an image with tf.browser.fromPixels()

"ERROR Error: Uncaught (in promise): Error: Argument 'x' passed to 'expandDims' must be a Tensor or TensorLike, but got 'Promise'"

I have this same kind of error (with an other function than "expandDims" displayed in the error) since the first time I used tfjs.

Error appears only when i use tfjs>=2.0, with tfjs=1.X i do not have error of this kind.

I have tried with models trained with python TF object detection api for TF 1.X and TF 2.X and I have the same error for both.

Thank you for your help.

[lina128]

Hi @rfourdinier, can you open another issue with this other bug you found, and preferably some code snippet to help us reproduce the error. Thank you!

[rfourdinier]

Hello @lina128 i have opened issue #4062 thanks