MLX
Joblib
Safetensors
English
reasoning
chain-of-thought
context-compression
soft-prompt
apple-silicon
Instructions to use baya1116/hypernet-sp-distill with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use baya1116/hypernet-sp-distill with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir hypernet-sp-distill baya1116/hypernet-sp-distill
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
| BOOT_START Mon Jun 15 10:12:43 UTC 2026 | |
| WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager, possibly rendering your system unusable.It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv. Use the --root-user-action option if you know what you are doing and want to suppress this warning. | |
| PIP_DONE Mon Jun 15 10:13:30 UTC 2026 | |
| HF_DONE | |
| BUILD_FFT Mon Jun 15 10:13:43 UTC 2026 | |
| FFT_HF_DONE t=50s | |
| BUILD_FFT_DONE Mon Jun 15 10:14:36 UTC 2026 | |
| RUN Mon Jun 15 10:14:36 UTC 2026 | |
| [knowref] N=180 layers=(8, 14, 20)(recall) pos=onset/gen10/gen30 | 2-tower BGEHead | |
| 30/180 t=24s | |
| 60/180 t=36s | |
| 90/180 t=48s | |
| 120/180 t=61s | |
| 150/180 t=73s | |
| 180/180 t=85s | |
| pos=onset: TRAINED @1=0.473 @3=0.836 | REUSE-RECALL @1=0.018 @3=0.055 | |
| pos=gen1 : TRAINED @1=0.273 @3=0.6 | REUSE-RECALL @1=0.018 @3=0.055 | |
| pos=gen2 : TRAINED @1=0.091 @3=0.2 | REUSE-RECALL @1=0.018 @3=0.055 | |
| === REUSE recall bge_head for SEARCH? (q@8/14/20) === | |
| onset: REUSE-RECALL @1=0.018 @3=0.055 (fresh-trained @1=0.473) | |
| gen1 : REUSE-RECALL @1=0.018 @3=0.055 (fresh-trained @1=0.273) | |
| gen2 : REUSE-RECALL @1=0.018 @3=0.055 (fresh-trained @1=0.091) | |
| DONE t=89s | |
| RUN_DONE rc=0 Mon Jun 15 10:16:08 UTC 2026 | |