Uncaught (in promise) RangeError: Array buffer allocation failed

[ItsSimko]

System Info

Running chrome v138, node v22

Environment/Platform

• Website/web-app
• Browser extension
• Server-side (e.g., Node.js, Deno, Bun)
• Desktop app (e.g., Electron)
• Other (e.g., VSCode extension)

Description

Uncaught (in promise) RangeError: Array buffer allocation failed
    at new ArrayBuffer (<anonymous>)
    at new Uint8Array (<anonymous>)
    at r.progress_callback.navigator.test.n.status (transformers.min.js:1:749405)
    at f (transformers.min.js:1:749676)
    at async transformers.min.js:1:515017

Tryin to run llama 3.2 3b q4f16

Reproduction

Try to load the llm pipeline and this is the first error that pops up then model never loads

[ItsSimko]

Any help would be appreciated can provide more info. Im running it in a background worker so the model can stay loaded at all times.

[xenova]

Hi there 👋 Which model are you trying to run? Could you provide some sample code?

[ItsSimko]

Hi there 👋 Which model are you trying to run? Could you provide some sample code?

Hi this is on a open source project on this branch. under the src-v2 folder. I know the project structure is a little messy. Let me know if i can help you find anything. The main logic surrounding this is in worker.js. here:

https://github.com/ClinicianFOCUS/freescribe-copilot/blob/6837cf34ffdfba0701cca9b4331944124a63dfd9/src-v2/worker.js#L186-L237

Model im trying to run is the onnx-community/Llama-3.2-3B-Instruct q4f16 any insight would be helpful. seems to work fine when i run the onnx-community/Llama-3.2-1B-Instruct q4f16

[ItsSimko]

So I could run onnx-community/granite-3.0-2b-instruct q4f16 and same with onnx-community/Phi-3.5-mini-instruct-onnx-web. No success trying to run onnx-community/granite-3.0-2b-instruct uint8 or int8. Are int8 and unint8 quants supported in extensions?

Also still no success on onnx-community/Llama-3.2-3B-Instruct q4f16 q4 or int8.

@xenova

[niutech]

I have the same issue RangeError: Array buffer allocation failed in Edge 138 when I try to load https://huggingface.co/jglowa/Bielik-4.5B-v3.0-Instruct-ONNX-INT4/blob/main/onnx/model_q4.onnx_data, which is 2.78GB - try it yourself: https://huggingface.co/spaces/jglowa/prosty-jezyk.

This is a duplicate of #952.