justinchuby/onnx-genai-example-qwen3-5-0-8b-hybrid-vlm-f32

onnx-genai-example-qwen3-5-0-8b-hybrid-vlm-f32

Private real-weight ONNX package produced by Mobius from `Qwen/Qwen3.5-0.8B` at immutable revision 2fc06364715b967f1860aea9cf38778875588b17. Source license: apache-2.0.

This package exposes 18 com.microsoft::LinearAttention nodes, 18 com.microsoft::CausalConvWithState nodes, six full-attention layers, all convolution/recurrent state I/O, plus embedding and vision graphs.

Contents

• —Canonical, hashless inference_metadata.yaml
• —ONNX graphs and external-data weights
• —Complete tokenizer/processor assets
• —request.json and output.json real runtime evidence
• —graph_report.json, performance.json, source.json, and provenance.json

Observed output: `Describe the image in one short sentence.

The image shows`

Exact download

hf download justinchuby/onnx-genai-example-qwen3-5-0-8b-hybrid-vlm-f32 --repo-type model --local-dir ./qwen3.5-0.8b-hybrid-vlm-f32

ONNX Runtime load smoke test

python - <<'PY'
from pathlib import Path
import onnxruntime as ort

root = Path("qwen3.5-0.8b-hybrid-vlm-f32")
for relative_path in ['decoder/model.onnx', 'embedding/model.onnx', 'vision_encoder/model.onnx']:
    session = ort.InferenceSession(
        str(root / relative_path),
        providers=['CUDAExecutionProvider', 'CPUExecutionProvider'],
    )
    print(relative_path, session.get_providers(), [x.name for x in session.get_inputs()])
PY

The exact successful probe request, output, versions, providers, and timings are preserved in request.json, output.json, and performance.json.

<!-- inference-metadata-annotation:start -->

Annotated inference metadata

Review `inference_metadata.annotated.yaml` for inline explanations of this package's workflow, tensor/state/cache contracts, and fail-closed omissions. `inference_metadata.yaml` remains the canonical machine-authored contract; automated validation confirms both files parse to the same metadata object. <!-- inference-metadata-annotation:end -->