{ "id": "siglip-base-p16-224-int8", "description": "SigLIP base, patch 16, 224 px: image and text encoders, ONNX, int8 dynamically quantised. Zero-shot image–text similarity for image filtering.", "upstream": { "repo": "google/siglip-base-patch16-224", "revision": "7fd15f0689c79d79e38b1c2e2e2370a7bf2761ed", "license": "Apache-2.0" }, "repo": "Xenova/siglip-base-patch16-224", "revision": "4649052661e53c7000355844105f8a1792088239", "license": "Apache-2.0", "license_note": "ONNX export of the Apache-2.0 upstream weights (the export repository declares base_model google/siglip-base-patch16-224 and no separate licence).", "runtime": "onnxruntime-web, non-bundled build (WebAssembly CPU execution provider; up to 4 threads in the cross-origin-isolated offscreen document)", "inputs": { "image": "pixel_values float32 [1, 3, 224, 224]", "text": "input_ids int64 [n, 64], padded with (id 1)" }, "outputs": "pooler_output float32 [n, 768] (L2-normalised by the worker)", "preprocessing": { "resize": "224x224 (no crop, aspect ratio not preserved), bicubic", "rescale": 0.00392156862745098, "mean": [0.5, 0.5, 0.5], "std": [0.5, 0.5, 0.5], "text": "lowercase, remove ASCII punctuation, collapse whitespace; SentencePiece Unigram; append ; pad to 64" }, "files": { "onnx/vision_model_quantized.onnx": { "sha256": "ef14a954f3d57e1806666432bd9785004c1dc27100aa260eee0cb0f10a5de058", "size": 99499129 }, "onnx/text_model_quantized.onnx": { "sha256": "ad0329b1f35acc66d8953ff2559ce358da8eb0a7011794cf951523d63a4dbce2", "size": 111475220 }, "tokenizer.json": { "sha256": "4a17c975210be5ab4c36b47d8dae4eefb866dbfb1e676e394aad85dc30a3ae08", "size": 2398744 }, "preprocessor_config.json": { "sha256": "21ee046a8a52a65e5f9c177bf840bfb39ea66c9c54cf2760630efd58e0a3ec80", "size": 368 }, "config.json": { "sha256": "e6de71291f181b0b81adc93098787bb4597a79dc18f59737feda8f41671fb6a2", "size": 457 } } }