Feature Extraction
MLX
Safetensors
qwen3
embeddings
sentence-similarity
quantization
omlx
oq8e
8-bit precision
Instructions to use TiGa-RCE/Qwen3-Embedding-0.6B-MLX-oQ8e with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use TiGa-RCE/Qwen3-Embedding-0.6B-MLX-oQ8e with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir Qwen3-Embedding-0.6B-MLX-oQ8e TiGa-RCE/Qwen3-Embedding-0.6B-MLX-oQ8e
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
| { | |
| "schema_version": 1, | |
| "published_repository": "TiGa-RCE/Qwen3-Embedding-0.6B-MLX-oQ8e", | |
| "family": "Qwen3-Embedding-0.6B", | |
| "variant": "oQ8e", | |
| "upstream_repository": "Qwen/Qwen3-Embedding-0.6B", | |
| "upstream_revision": "97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3", | |
| "upstream_revision_evidence": "exact source snapshot retained in local Hugging Face download metadata", | |
| "direct_parent": "TiGa-RCE/Qwen3-Embedding-0.6B-MLX-BF16", | |
| "direct_parent_weight_hashes": [ | |
| { | |
| "file": "model.safetensors", | |
| "sha256": "87d77ce03148999dec3d5b73168df202d4b15862ea335658ec015e9bba07b14f", | |
| "bytes": 1191588265 | |
| } | |
| ], | |
| "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.safetensors", | |
| "sha256": "7c72d2475abca87ccf56f7360e673b98dd6d40442a793e868f80a40cd266009d", | |
| "bytes": 633152522 | |
| } | |
| ], | |
| "evaluation": { | |
| "pair_count": 24, | |
| "top1": 1.0, | |
| "recall_at_5": 1.0, | |
| "mrr": 1.0, | |
| "mean_aligned_embedding_cosine_vs_bf16": 0.9994478225708008, | |
| "minimum_aligned_embedding_cosine_vs_bf16": 0.9986600279808044, | |
| "score_rmse_vs_bf16": 0.0030556409619748592, | |
| "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" | |
| } | |