Constant output "<extra_id_0>” in Xenova/mt5-small

Models
1

It looks like a bug or I can’t get it why it fails.

Running script with Vite just generates output of array with generated text:
"<extra_id_0>”

Here’s the code script with selecting dtype q4:

import { pipeline } from '@huggingface/transformers';

const generator = await pipeline('text2text-generation', 'Xenova/mt5-small', {dtype:"q4"});


const text = 'Once upon a time';

const output = await generator(text);

console.log(output)

And that’s the error :

Uncaught Error: Could not locate file: "https://huggingface.co/Xenova/mt5-small/resolve/main/onnx/decoder_model_merged_q4.onnx".

Running script without setting dtype in node.js results:

dtype not specified for "encoder_model". Using the default dtype (fp32) for this device (cpu).
dtype not specified for "decoder_model_merged". Using the default dtype (fp32) for this device (cpu).
C:\Users\user\Desktop\ML TEST\node_modules\onnxruntime-node\dist\backend.js:25
            __classPrivateFieldGet(this, _OnnxruntimeSessionHandler_inferenceSession, "f").loadModel(pathOrBuffer, options);
                                                                                           ^

Error: Load model from C:\Users\user\Desktop\ML TEST\node_modules\@huggingface\transformers\.cache\Xenova\mt5-small\onnx\encoder_model.onnx failed:Protobuf parsing failed.      
    at new OnnxruntimeSessionHandler (C:\Users\user\Desktop\ML TEST\node_modules\onnxruntime-node\dist\backend.js:25:92)
    at Immediate.<anonymous> (C:\Users\user\Desktop\ML TEST\node_modules\onnxruntime-node\dist\backend.js:67:29)
    at process.processImmediate (node:internal/timers:485:21)
1 Like
2

"<extra_id_0>”

The appearance of the token itself with T5 is normal.

failed:Protobuf parsing failed.

This seems to be a bug in the new version.