{ "schema_version": 1, "published_repository": "TiGa-RCE/Qwen3-Embedding-8B-MLX-oQ8e", "family": "Qwen3-Embedding-8B", "variant": "oQ8e", "upstream_repository": "Qwen/Qwen3-Embedding-8B", "upstream_revision": "1d8ad4ca9b3dd8059ad90a75d4983776a23d44af", "upstream_revision_evidence": "exact source snapshot retained in local Hugging Face download metadata", "direct_parent": "TiGa-RCE/Qwen3-Embedding-8B-MLX-BF16", "direct_parent_weight_hashes": [ { "file": "model-00001-of-00004.safetensors", "sha256": "0a22d58cb671a9049da3505b30b3ba8a200401c4309df397961f61c9c6ab1643", "bytes": 4900037603 }, { "file": "model-00002-of-00004.safetensors", "sha256": "18c40e1305906ad2ca3c503efa130f02b20db5940129222cb1728f4a594204ad", "bytes": 4915960339 }, { "file": "model-00003-of-00004.safetensors", "sha256": "ce16f97c40ed67ec1b540dcb99259e91a930b431051e50a76350888a2331498c", "bytes": 4983068481 }, { "file": "model-00004-of-00004.safetensors", "sha256": "1ac381756541d3601847be3e5ce832a81f308cf8b7afccc1bb65e22c63120cfe", "bytes": 335570419 } ], "conversion": { "method": "oQe calibrated mixed-precision affine quantization", "nominal_bits": 8, "group_size": 64, "importance_matrix": true, "importance_matrix_samples": 128, "importance_matrix_sequence_length": 512, "stack": { "omlx": "0.5.3", "mlx_lm": "0.31.3", "mlx": "0.32.0" }, "lossy_parent": false }, "weight_files": [ { "file": "model-00001-of-00002.safetensors", "sha256": "e24bdada0ab9a84604a8e103785928e27514b6f13590087173635dcf0ca83019", "bytes": 5018859306 }, { "file": "model-00002-of-00002.safetensors", "sha256": "57e3027e47020f5ef97ccc311f4febaf57b7eb4ca91cb575337c0c4a0d156d67", "bytes": 3021784290 } ], "evaluation": { "pair_count": 24, "top1": 1.0, "recall_at_5": 1.0, "mrr": 1.0, "mean_aligned_embedding_cosine_vs_bf16": 0.9996938705444336, "minimum_aligned_embedding_cosine_vs_bf16": 0.9995684027671814, "score_rmse_vs_bf16": 0.001507166656665504, "queries_with_rank_change": 0, "gate_passed": true, "gate_criteria": { "top1_delta_min": 0.0, "recall_at_5_delta_min": 0.0, "mrr_delta_min": -0.01, "minimum_aligned_embedding_cosine_min": 0.99, "queries_with_rank_change_max": 2 } }, "collection": "https://huggingface.co/collections/TiGa-RCE/mlx-embedding-quantization-matrix-q-oq-oqe-at-4-6-8-bit-6a68d11afb238d4fe967d70b" }