--- license: apache-2.0 base_model: litert-community/Vision-RWKV-S-LiteRT base_model_relation: quantized library_name: openvino pipeline_tag: image-classification tags: - openvino - onnx - intel-npu --- # Vision-RWKV-S-OpenVINO Vision-RWKV-S for Intel NPUs, GPUs and CPUs: vrwkv_s_fp16 from [litert-community/Vision-RWKV-S-LiteRT](https://huggingface.co/litert-community/Vision-RWKV-S-LiteRT/tree/184992fed3c7bb2761220287bcd0a496959e4e39) as ONNX (tf2onnx 1.17.0, opset 17), OpenVINO IR with fp32 weights, and a precompiled Intel NPU blob. On the NPU of a Core Ultra 9 386H the blob runs in 234.02 ms per run, numpy input to numpy output. Scripts, raw records, and the same table for all 52 graphs of LiteRT-Models: https://github.com/john-rocky/windows-zoo ## Files | file | bytes | sha256 | |---|---|---| | `vrwkv_s_fp16.onnx` | 95,635,055 | `12184150736b7ec227a0a98ca6fa70ab69c8894ecb9cba800ce2537c6d75aaed` | | `openvino/vrwkv_s_fp16.xml` | 1,546,499 | `f34f101344f8e83faddd5c81987249a028057f284edd41057308e11907390d5e` | | `openvino/vrwkv_s_fp16.bin` | 95,336,452 | `54b97b3387a49b24689b5e78ad3d6004d78b742d327c30805b76a802ffa95782` | | `npu/vrwkv_s_fp16.blob` | 74,580,104 | `bed2c0e5fe4cc71621824d94642135ec77a1b0ca3bd384f861294a95bfaf76fe` | The NPU blob was compiled with OpenVINO 2026.4.1, Intel NPU driver 32.0.100.4841 (OpenVINO reports NPU_DRIVER_VERSION 1004841, NPU_COMPILER_VERSION 524292). If `import_model` refuses a blob on another driver or OpenVINO version, compile the IR instead (the commented line in the snippet). ## How it was measured MSI Stealth 16 AI+ (Core Ultra 9 386H, RTX 5080 Laptop GPU at a 95 W power limit, Windows 11 build 26200.9711, on AC power), 2026-11-06 to 2026-11-07. On the laptop, each row is one process: load, one run, 5 warmup runs, then the median of 30 timed runs (2 warmup + 10 timed runs when the first run takes more than 3 s), numpy input to numpy output, every runtime at its default settings, one process at a time, each started after the CPU, the RTX and the Intel GPU had gone idle (the thresholds are in METHOD.md of the GitHub repo). Output check: each output against the CPU output in fp32 on the same input (ONNX Runtime's CPU EP); PASS = cosine ≥ 0.99, no NaN or Inf, and top-1 agreement ≥ 0.99 for class scores. The Snapdragon rows are Qualcomm AI Hub profile jobs of the same ONNX files (their own timer). ## Speed and output check | processor | runtime | file | ms | load s | output check | |---|---|---|---|---|---| | Intel AI Boost NPU | OpenVINO 2026.4.1 | `npu/` blob (`import_model`) | 234.02 | 0.05 | PASS (cos ≥ 0.9999) | | Intel AI Boost NPU | ONNX Runtime 1.31.1 + OpenVINO EP 1.8.1 | `.onnx`, compiled when the session loads | 282.56 | 67.01 | PASS (cos ≥ 0.9999) | | Intel Graphics (iGPU) | ONNX Runtime 1.31.1 + OpenVINO EP 1.8.1 | `.onnx` | 51.47 | 3.44 | PASS (cos ≥ 0.9999) | | RTX 5080 Laptop (95 W limit) | ONNX Runtime 1.31.1, CUDA EP | fp16 copy of the `.onnx` (not in this repo) | 15.01 | 1.27 | PASS (cos ≥ 0.9998) | | RTX 5080 Laptop (95 W limit) | ONNX Runtime 1.31.1, CUDA EP | `.onnx` | 25.88 | 1.89 | PASS (cos ≥ 0.9999) | | Core Ultra 9 386H CPU | ONNX Runtime 1.31.1, CPU EP | `.onnx` | 385.53 | 0.17 | reference | | Snapdragon X2 Elite NPU | Qualcomm AI Hub, precompiled QNN ONNX | `.onnx` compiled by AI Hub | 76.93 | 0.45 | not compared (AI Hub profile job, 2026-10-30) | | Snapdragon X2 Elite NPU | Qualcomm AI Hub, ONNX Runtime QNN EP | `.onnx`, compiled when it loads | 75.50 | 91.67 | not compared (AI Hub profile job, 2026-10-30) | | Galaxy S26 NPU | Qualcomm AI Hub, precompiled QNN ONNX | `.onnx` compiled by AI Hub | 75.99 | 0.32 | not compared (AI Hub profile job, 2026-10-30) | The ONNX file against the source `.tflite` (LiteRT interpreter on the CPU, the same input): PASS, cosine ≥ 0.9999 for every output. ## Run it on the Intel NPU Download this repo, then run the snippets from its folder: ```sh pip install huggingface_hub hf download mlboydaisuke/Vision-RWKV-S-OpenVINO --local-dir Vision-RWKV-S-OpenVINO cd Vision-RWKV-S-OpenVINO ``` OpenVINO (`pip install openvino==2026.4.1`): ```python import numpy as np import openvino as ov core = ov.Core() if "NPU" not in core.available_devices: raise SystemExit(f"no Intel NPU visible to OpenVINO: {core.available_devices}") with open("npu/vrwkv_s_fp16.blob", "rb") as f: compiled = core.import_model(f.read(), "NPU") # precompiled: no compile at load # or compile the IR on this machine: compiled = core.compile_model("openvino/vrwkv_s_fp16.xml", "NPU") # replace the zeros with your preprocessed input (float32, the shapes printed by compiled.inputs) feeds = {i.get_any_name(): np.zeros(list(i.shape), np.float32) for i in compiled.inputs} outputs = compiled(feeds) print({o.get_any_name(): outputs[o].shape for o in compiled.outputs}) ``` ONNX Runtime with the OpenVINO EP (`pip install onnxruntime==1.31.1 onnxruntime-ep-openvino==1.8.1`): ```python import numpy as np import onnxruntime as ort import onnxruntime_ep_openvino as ov_ep ort.register_execution_provider_library(ov_ep.get_ep_name(), ov_ep.get_library_path()) npu = [d for d in ort.get_ep_devices() if d.ep_name == ov_ep.get_ep_name() and d.ep_metadata.get("ov_device") == "NPU"] if not npu: raise SystemExit("no Intel NPU visible to the OpenVINO EP") so = ort.SessionOptions() so.add_provider_for_devices(npu, {}) sess = ort.InferenceSession("vrwkv_s_fp16.onnx", sess_options=so) # compiles for the NPU here # replace the zeros with your preprocessed input (float32) feeds = {i.name: np.zeros(i.shape, np.float32) for i in sess.get_inputs()} outputs = sess.run(None, feeds) print([o.shape for o in outputs]) ``` ## Licence Apache-2.0, the licence of [litert-community/Vision-RWKV-S-LiteRT](https://huggingface.co/litert-community/Vision-RWKV-S-LiteRT). The upstream model: [OpenGVLab/Vision-RWKV](https://huggingface.co/OpenGVLab/Vision-RWKV).