Feature Extraction
MLX
Safetensors
multilingual
embedding_gemma2
mlx-vlm
embedding
sentence-similarity
multimodal
image-feature-extraction
audio-feature-extraction
video-feature-extraction
4-bit precision
Instructions to use mlx-community/embeddinggemma-2-nvfp4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use mlx-community/embeddinggemma-2-nvfp4 with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download mlx-community/embeddinggemma-2-nvfp4 --local-dir embeddinggemma-2-nvfp4
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Download validation.json from mlx-community/embeddinggemma-2-nvfp4: direct link, hf CLI and curl.
- Browser
- Download file 2.43 kB
-
https://huggingface.co/mlx-community/embeddinggemma-2-nvfp4/resolve/main/validation.json
- Command line
-
hf download hf://mlx-community/embeddinggemma-2-nvfp4/validation.json
-
curl -L -o validation.json https://huggingface.co/mlx-community/embeddinggemma-2-nvfp4/resolve/main/validation.json
2.43 kB
| { | |
| "source_model": "google/embeddinggemma-2", | |
| "source_revision": "914f7f89142e33e77833254d9c9b90c3cef7303b", | |
| "mlx_vlm_revision": "3d87e88402f307efbf68e568971aa887ee7d9ed0", | |
| "variant": "nvfp4", | |
| "hardware": "Apple M3 Ultra", | |
| "mlx_version": "0.32.3", | |
| "transformers_version": "5.18.0.dev0", | |
| "validation_scope": "Six multilingual text inputs and synthetic image, audio, two-frame video, and text+image; numerical smoke checks, not a retrieval benchmark.", | |
| "tensor_count": 1596, | |
| "tensor_dtypes": [ | |
| "BF16", | |
| "U32", | |
| "U8" | |
| ], | |
| "weight_bytes": 1098354512, | |
| "cases": { | |
| "audio": { | |
| "shape": [ | |
| 1, | |
| 768 | |
| ], | |
| "finite": true, | |
| "minimum_cosine_vs_torch_fp32": 0.9802272979169695, | |
| "maximum_absolute_error_vs_torch_fp32": 0.02003237258625193, | |
| "minimum_cosine_vs_source_bf16": 0.9801554372695555, | |
| "minimum_truncated_cosine_vs_torch_fp32": 0.9816903229454815 | |
| }, | |
| "image": { | |
| "shape": [ | |
| 1, | |
| 768 | |
| ], | |
| "finite": true, | |
| "minimum_cosine_vs_torch_fp32": 0.9915933798123155, | |
| "maximum_absolute_error_vs_torch_fp32": 0.017620143370074736, | |
| "minimum_cosine_vs_source_bf16": 0.9916633223645384, | |
| "minimum_truncated_cosine_vs_torch_fp32": 0.992140638740076 | |
| }, | |
| "text": { | |
| "shape": [ | |
| 6, | |
| 768 | |
| ], | |
| "finite": true, | |
| "minimum_cosine_vs_torch_fp32": 0.9806187006587285, | |
| "maximum_absolute_error_vs_torch_fp32": 0.02378729080233242, | |
| "minimum_cosine_vs_source_bf16": 0.9805855820726995, | |
| "minimum_truncated_cosine_vs_torch_fp32": 0.9816867502031378, | |
| "retrieval_scores_venus_mars": [ | |
| 0.6455942988395691, | |
| 0.8219397068023682 | |
| ] | |
| }, | |
| "text_image": { | |
| "shape": [ | |
| 1, | |
| 768 | |
| ], | |
| "finite": true, | |
| "minimum_cosine_vs_torch_fp32": 0.9877009275280101, | |
| "maximum_absolute_error_vs_torch_fp32": 0.017077530643317997, | |
| "minimum_cosine_vs_source_bf16": 0.9873912535194416, | |
| "minimum_truncated_cosine_vs_torch_fp32": 0.9882139319952599 | |
| }, | |
| "video": { | |
| "shape": [ | |
| 1, | |
| 768 | |
| ], | |
| "finite": true, | |
| "minimum_cosine_vs_torch_fp32": 0.9875313525163498, | |
| "maximum_absolute_error_vs_torch_fp32": 0.018628835730066858, | |
| "minimum_cosine_vs_source_bf16": 0.9877214818796509, | |
| "minimum_truncated_cosine_vs_torch_fp32": 0.9879377322879118 | |
| } | |
| } | |
| } | |