any2jev-qwen3.5-0.8b-synthetic

any2jev-qwen3.5-0.8b-synthetic

A Jev-style System One decision model made with any2jev from Qwen/Qwen3.5-0.8B. State in, typed Choice / Score / Noul answers with calibrated probabilities out, in one forward pass. No text is generated.

pip install "any2jev[serve]"
any2jev serve hf://huaweifeng/any2jev-qwen3.5-0.8b-synthetic      # POST /v1/systemone, TypeSafe SDK compatible
any2jev ask hf://huaweifeng/any2jev-qwen3.5-0.8b-synthetic --state "My payouts have failed 3 days in a row, fix this ASAP" \
    --choice "Which team? | billing, technical, sales" --noul "Is this urgent?"

What is in this repo

• adapter/: LoRA adapter (r=8) on Qwen/Qwen3.5-0.8B, trained with the vocabulary head removed
• head.safetensors: the pointer head (dim 256) that scores options against the decision token
• any2jev.json: delimiters, mode (rows), temperature 0.182 fitted on validation
• tokenizer/: the base tokenizer (delimiter tokens reused or added)
• train_report.json, eval.json: training config, history and held-out metrics

Training

• data: data/synthetic/train.jsonl; 5.9 M trainable parameters, 1.0 epoch(s), lr 0.0002, batch 2 x 4
• wall clock: 26 min on one consumer GPU

Trained on any2jev's synthetic support-ticket generator (any2jev data synthetic). This checkpoint demonstrates the conversion recipe on this backbone; it is not a general-purpose decision model.

Held-out evaluation

Limitations

Trained on a few thousand labelled decisions from a handful of sources; expect the accuracy above on similar inputs and lower accuracy off-distribution. Probabilities are calibrated on the validation split, not a guarantee per answer. Keep arithmetic, dates and counting in code, as Jev's own docs recommend.

Independent project, not affiliated with TypeSafe AI. Apache-2.0.