OnnxRuntime crash in CI (Ubuntu) with cpu

[wszgrcy]

System Info

• transformers.js version: 4.2.0
• Operating System:
  • Failing: Ubuntu (GitHub Actions runner)
  • Working: Windows 10
• Node.js version: v24.15
• Relevant dependencies: @huggingface/transformers@4.2.0

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

In GitHub Actions, the execution of Transformers.js fails.

Reproduction

1. Run the following code in a GitHub Actions workflow using Ubuntu:

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

async function run() {
  try {
    // Note: 'dml' might not be valid/supportable on Linux CI runners without specific setup,
    // but this is the configuration failing. 
    const instance = await pipeline(
      'feature-extraction',
      'Xenova/jina-embeddings-v2-small-en',
      { 
        device: process.env.CI ? 'cpu' : 'dml', 
        dtype: 'fp16' 
      },
    );
    console.log('Pipeline loaded successfully');
  } catch (error) {
    console.error(error);
    throw error;
  }
}

run();

2. Error Output:

2026-06-01 03:38:17.401750720 [E:onnxruntime:, inference_session.cc:2619 operator()] Exception during initialization: /onnxruntime_src/onnxruntime/core/graph/graph_utils.cc:29 int onnxruntime::graph_utils::GetIndexFromName(const onnxruntime::Node&, const std::string&, bool) itr != node_args.end() was false. Attempting to get index by a name which does not exist:InsertedPrecisionFreeCast_/encoder/layer.3/attention/output/LayerNorm/Constant_output_0for node: /embeddings/LayerNorm/Mul/SimplifiedLayerNormFusion/

2026-06-01 03:38:17.464710163 [E:onnxruntime:, inference_session.cc:2619 operator()] Exception during initialization: /onnxruntime_src/onnxruntime/core/graph/graph_utils.cc:29 int onnxruntime::graph_utils::GetIndexFromName(const onnxruntime::Node&, const std::string&, bool) itr != node_args.end() was false. Attempting to get index by a name which does not exist:InsertedPrecisionFreeCast_/encoder/layer.3/attention/output/LayerNorm/Constant_output_0for node: /embeddings/LayerNorm/Mul/SimplifiedLayerNormFusion/

[xenova]

Thanks for the report! Looks to be an upstream onnxruntime crash; pinging @guschmue and @fs-eire

InsertedPrecisionFreeCast

This usually happens when running fp16 on CPU. @wszgrcy if you can try run q4 or q8 (or fp32) that would most likely fix it

[wszgrcy]

Thanks for the report! Looks to be an upstream onnxruntime crash; pinging @guschmue and @fs-eire

InsertedPrecisionFreeCast

This usually happens when running fp16 on CPU. @wszgrcy if you can try run q4 or q8 (or fp32) that would most likely fix it

Thank you. After modification, it did run successfully. Will fp16 support be fixed in the future?