Instructions to use sidchalamala/aether-coder-1.5b-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use sidchalamala/aether-coder-1.5b-lora with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-Coder-1.5B-Instruct") model = PeftModel.from_pretrained(base_model, "sidchalamala/aether-coder-1.5b-lora") - Notebooks
- Google Colab
- Kaggle
Upload README.md
Browse files
README.md
CHANGED
|
@@ -9,6 +9,8 @@ tags: [lora, code, dsl, aether]
|
|
| 9 |
|
| 10 |
A LoRA adapter that turns a plain-English business procedure, plus the owner's policy and sheet list, into an Aether program. **The model is not a safety mechanism.** Every program it writes must pass the Aether checker before it runs. The checker blocks programs that break the policy's rules; the model often writes such programs.
|
| 11 |
|
|
|
|
|
|
|
| 12 |
"correct" below means passes the checker and has the same effects as the reference program (scorer field `safe_and_correct`). It is not a safety rating.
|
| 13 |
|
| 14 |
## Final evaluation: frozen held-out trap set (30 tasks, each model evaluated once)
|
|
|
|
| 9 |
|
| 10 |
A LoRA adapter that turns a plain-English business procedure, plus the owner's policy and sheet list, into an Aether program. **The model is not a safety mechanism.** Every program it writes must pass the Aether checker before it runs. The checker blocks programs that break the policy's rules; the model often writes such programs.
|
| 11 |
|
| 12 |
+
**Links:** browser demo of the checker and interpreter (the model is not hosted there): https://warnedorca.github.io/aether/ 路 code: https://github.com/WarnedOrca/aether 路 benchmark and held-out set: https://huggingface.co/datasets/sidchalamala/aether-benchmark
|
| 13 |
+
|
| 14 |
"correct" below means passes the checker and has the same effects as the reference program (scorer field `safe_and_correct`). It is not a safety rating.
|
| 15 |
|
| 16 |
## Final evaluation: frozen held-out trap set (30 tasks, each model evaluated once)
|