Instructions to use emrevrg/AUBIN-12B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use emrevrg/AUBIN-12B with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("google/gemma-4-12B-it") model = PeftModel.from_pretrained(base_model, "emrevrg/AUBIN-12B") - Notebooks
- Google Colab
- Kaggle
Download test_clean_metrics.json from emrevrg/AUBIN-12B: direct link, hf CLI and curl.
- Browser
- Download file 2.3 kB
-
https://huggingface.co/emrevrg/AUBIN-12B/resolve/main/test_clean_metrics.json
- Command line
-
hf download hf://emrevrg/AUBIN-12B/test_clean_metrics.json
-
curl -L -o test_clean_metrics.json https://huggingface.co/emrevrg/AUBIN-12B/resolve/main/test_clean_metrics.json
2.3 kB
| { | |
| "AUBIN-12B full precision (think)": { | |
| "kev_transfer_test": { | |
| "n": 656, | |
| "accuracy": 0.878, | |
| "brier": 0.1991, | |
| "ece15": 0.0628, | |
| "nll": 0.4183 | |
| }, | |
| "kev_test": { | |
| "n": 1200, | |
| "accuracy": 0.835, | |
| "brier": 0.2587, | |
| "ece15": 0.0451, | |
| "nll": 0.7674 | |
| } | |
| }, | |
| "AUBIN-31B (fast, one pass)": { | |
| "kev_transfer_test": { | |
| "n": 656, | |
| "accuracy": 0.8979, | |
| "brier": 0.1782, | |
| "ece15": 0.0782, | |
| "nll": 0.3786 | |
| }, | |
| "kev_test": { | |
| "n": 1200, | |
| "accuracy": 0.8375, | |
| "brier": 0.252, | |
| "ece15": 0.0698, | |
| "nll": 0.7191 | |
| } | |
| }, | |
| "AUBIN-31B (think)": { | |
| "kev_transfer_test": { | |
| "n": 656, | |
| "accuracy": 0.8902, | |
| "brier": 0.1799, | |
| "ece15": 0.0339, | |
| "nll": 0.3655 | |
| }, | |
| "kev_test": { | |
| "n": 1200, | |
| "accuracy": 0.8458, | |
| "brier": 0.2467, | |
| "ece15": 0.0344, | |
| "nll": 0.7547 | |
| } | |
| }, | |
| "AUBIN Duo (31B + 12B, think)": { | |
| "kev_transfer_test": { | |
| "n": 656, | |
| "accuracy": 0.8933, | |
| "brier": 0.1694, | |
| "ece15": 0.055, | |
| "nll": 0.3528 | |
| }, | |
| "kev_test": { | |
| "n": 1200, | |
| "accuracy": 0.85, | |
| "brier": 0.238, | |
| "ece15": 0.0252, | |
| "nll": 0.7524 | |
| } | |
| }, | |
| "AUBIN Duo (31B + 12B full precision, think)": { | |
| "kev_transfer_test": { | |
| "n": 656, | |
| "accuracy": 0.8948, | |
| "brier": 0.1688, | |
| "ece15": 0.0398, | |
| "nll": 0.3492 | |
| }, | |
| "kev_test": { | |
| "n": 1200, | |
| "accuracy": 0.8458, | |
| "brier": 0.2386, | |
| "ece15": 0.0281, | |
| "nll": 0.7185 | |
| } | |
| }, | |
| "AUBIN Duo-SC (31B self-consistent think + 12B full precision)": { | |
| "kev_transfer_test": { | |
| "n": 656, | |
| "accuracy": 0.8948, | |
| "brier": 0.1681, | |
| "ece15": 0.046, | |
| "nll": 0.3486 | |
| }, | |
| "kev_test": { | |
| "n": 1200, | |
| "accuracy": 0.8425, | |
| "brier": 0.2391, | |
| "ece15": 0.035, | |
| "nll": 0.7191 | |
| } | |
| }, | |
| "AUBIN Trio (31B-SC + 12B full precision + Phi-4)": { | |
| "kev_transfer_test": { | |
| "n": 656, | |
| "accuracy": 0.8887, | |
| "brier": 0.169, | |
| "ece15": 0.0384, | |
| "nll": 0.3472 | |
| }, | |
| "kev_test": { | |
| "n": 1200, | |
| "accuracy": 0.8508, | |
| "brier": 0.2369, | |
| "ece15": 0.0223, | |
| "nll": 0.6749 | |
| } | |
| }, | |
| "AUBIN Quad (31B-SC + 12B full precision + Phi-4 + Mistral-Small)": { | |
| "kev_transfer_test": { | |
| "n": 656, | |
| "accuracy": 0.8857, | |
| "brier": 0.1678, | |
| "ece15": 0.053, | |
| "nll": 0.3481 | |
| } | |
| } | |
| } |