Instructions to use pgdev0410/cybersos-tiny-aya-sft-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use pgdev0410/cybersos-tiny-aya-sft-lora with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("/root/lexsi_hack/artifacts/base") model = PeftModel.from_pretrained(base_model, "pgdev0410/cybersos-tiny-aya-sft-lora") - Notebooks
- Google Colab
- Kaggle
CyberSOS β Tiny Aya SFT LoRA adapter (B1)
LoRA adapter for CohereLabs/tiny-aya-global (revision af89d219b53ed9b13b8a4645f9c8028973510324) that chooses one recovery action per turn β ask about an access factor, simulate a supported step, or finish with a sourced outcome β for personal Google accounts and Android devices. Trained and evaluated only in a deterministic simulator; it never touches a real account.
Locked test (240 episodes): base 0/240 β this adapter 210/240 objectives met (+87.5 pts safe success, 95% CI [82.4, 92.4]); the observed-state rule planner it was distilled from scores 211/240. The adapter proposed a step that the guard blocked in 23/240 episodes; always serve it behind the guard.
- Training: AlignTune SFT (TRL backend), 2,958 CuratorKIT-curated decisions from a deterministic planner, 1 epoch, LoRA r=16/Ξ±=32/dropout 0.05 on all linear layers, lr 1e-4, bf16, seed 42.
- Serving contract: prompt, tools, budgets and decoding are frozen in the repo (
configs/baselines.yaml); generation stops at<|END_RESPONSE|>; greedy decoding. - Code, write-up and reproduction: https://github.com/Pheonix-dev0410/lexsi-cybersos
- Data and traces: https://huggingface.co/datasets/pgdev0410/cybersos-recovery-scenarios, https://huggingface.co/datasets/pgdev0410/cybersos-sft-decisions, https://huggingface.co/datasets/pgdev0410/cybersos-eval-traces
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer
base = AutoModelForCausalLM.from_pretrained("CohereLabs/tiny-aya-global", revision="af89d219b53ed9b13b8a4645f9c8028973510324", torch_dtype="bfloat16")
model = PeftModel.from_pretrained(base, "pgdev0410/cybersos-tiny-aya-sft-lora")
tokenizer = AutoTokenizer.from_pretrained("pgdev0410/cybersos-tiny-aya-sft-lora")
Limitations: English only; 16 reviewed rules; simulated restoration is not real recovery. License: the base model is CC-BY-NC-4.0 (Cohere Labs); this adapter inherits its noncommercial terms.
- Downloads last month
- 24