gantman / learn-tfjs Goto Github PK

Today I did gh repo clone GantMan/learn-tfjs
I did npm i at chapter2/web/web-toxicity
http://localhost:1234/ shows : Uncaught (in promise) TypeError: forwardFunc is not a function

This not TFJS specific - I'm trying to get this running via deno - not nodejs, so no node_modules.

import tfjs from "https://cdn.jsdelivr.net/npm/@tensorflow/[email protected]/dist/tf.min.js";
import * as toxicity from "https://cdn.jsdelivr.net/npm/@tensorflow-models/[email protected]";
const threshold = 0.5;
toxicity.load(threshold).then((model: any) =>
{
    //
});

I get :

error: TS2339 [ERROR]: Property 'load' does not exist on type 'typeof import("https://cdn.jsdelivr.net/npm/@tensorflow-models/[email protected]")'.

Does anyone have any issues running this now ?

Some errorTypeError: forwardFunc_1 is not a function
node:internal/process/promises:279
            triggerUncaughtException(err, true /* fromPromise */);

Just tried to test the Chapter 2 examples: double clicked the first web/index.html.
Check the console: that prints out the error: "Uncaught SyntaxError: Cannot use import statement outside a module"

When trying to install node dependencies, specifically concerning tfjs-node, there seems to be an issue. This may be specific to my system, as I'm running Apple silicon, but essentially what happens is when the install script attempts to download the pre built binaries from https://storage.googleapis.com/tf-builds/pre-built-binary/napi-v8/3.21.1/CPU-darwin-3.21.1.tar.gz it gets the following error:

node-pre-gyp ERR! install response status 404 Not Found on https://storage.googleapis.com/tf-builds/pre-built-binary/napi-v8/3.21.1/CPU-darwin-3.21.1.tar.gz

this eventually results in more errors as the build script attempts to build tensorflow.

what's strange is this file doesn't even exist in the directory @ https://storage.googleapis.com/tf-builds, and and there are zero darwin specific builds in for napi-v8 (the latest is napi-v7/3.6.1 for both CPU and GPU), so i'm unsure why the script is using the build above.

i'm wondering if i should be using a specific arm64 build of node for Apple Silicon or if tfjs-node is even compilable natively on Apple silicon currently?

also happy to provide my full logs if it helps troubleshoot this issue.

Aren't you creating a new rank 2 Tensor filled with 1 everywhere instead of copying and shading an existing dice ?