`TypeError: Deno.dlopen is not a function`

[onmax]

Hello,

I am trying to use the sentiment model, but I get an error. This is my code:

const manager = await ModelManager.create();
const sentimentModel = await manager.createSentimentModel();

async function filterTweets(tweets: TwitterResponse) {
    const sentiments = await sentimentModel.predict(tweets.data.map(tweet => tweet.text));
    return tweets.data.filter((_, index) => {
        return sentiments[index].score > 0.7;
    });
}

Error

at Module.prepare (https://deno.land/x/[email protected]/plug.ts:77:15)
  at async Function.create (https://deno.land/x/[email protected]/model_manager.ts:181:17)

Command

bash deno run -A ./src/tweets.ts --unstable

[sno2]

Hello, I believe the issue is that you passed in the --unstable flag after the file. Can you try running your command again with the --unstable flag after the -A flag?

bash deno run -A --unstable ./src/tweets.ts

[onmax]

I am getting another error now:

Command

deno run -A --unstable ./src/tweets.ts

Error

error: Uncaught (in promise) Error: Could not open library: Could not open library: libtorch_cpu.so: cannot open shared object file: No such file or directory
  return Deno.dlopen(file.path, symbols);
              ^
    at new DynamicLibrary (deno:ext/ffi/00_ffi.js:274:39)
    at Object.dlopen (deno:ext/ffi/00_ffi.js:368:12)
    at Module.prepare (https://deno.land/x/[email protected]/plug.ts:77:15)
    at async Function.create (https://deno.land/x/[email protected]/model_manager.ts:181:17)
    at async file:///onmax/my-project/src/tweets.ts:25:17

[sno2]

Hmm, I suppose that means that rust-bert dynamically links torch. Sorry, did not know that it did that as I've only ran bertml locally. Can you try install libtorch v1.13.0 from https://pytorch.org/get-started/locally/ ?

Also, these are the full instructions from the rust-bert repo: https://github.com/guillaume-be/rust-bert#manual-installation-recommended