onnx-genai-example-mistral-7b-v0-1-sliding-window

Public real-weight fp16 ONNX package from mistralai/Mistral-7B-v0.1 at immutable revision 27d67f1b5f57dc0953326b2601d68371d40ea8da. Source license: Apache-2.0.

The graph contains 32 CUDA com.microsoft::GroupQueryAttention nodes with local_window_size=4096. A real H200 CUDA probe generated 12 finite-logit tokens from positions 4092–4103, crossing the configured boundary. The package includes canonical metadata, tokenizer, policies, provenance, request/output, graph report, and exact timings.

Download

hf download justinchuby/onnx-genai-example-mistral-7b-v0-1-sliding-window --repo-type model --local-dir ./mistral-7b-v0.1-sliding-window

Exact runtime probe

cd mistral-7b-v0.1-sliding-window
python3 evidence/probe_sliding_window.py
cat evidence/sliding_window_runtime_probe.json

Requires CUDA-capable onnxruntime-gpu, onnx, transformers, and numpy; exact successful versions are in output.json.

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.