# KnowLine-12B-Gen1: inference and Decision Index reproduction ## Model and readout - Base model: [google/gemma-4-12B-it](https://huggingface.co/google/gemma-4-12B-it). - bf16 weights, approximately 23.9 GB. - Readout temperature T = 1.47 is folded into the final language-model norm and final logit soft-cap. - The front-end temperature is 1. `temperatures.json` supplies `{"score": 0.6803}`, so score questions use an effective temperature of approximately 1.0; choice and yes/no questions retain T = 1.47. - Each question is scored in one prefill by reading option-label probabilities, with no generated reasoning. - Text-only requests whose one-pass length is at most 3,000 tokens are read twice. Longer requests and image requests are read once. Question headers and referenced-fields layout are disabled. - Limits: 64 questions per request and 255 options per question. Requests are not truncated. ## Software The evaluation used SGLang 0.5.21, Transformers 5.12.1, Torch 2.13.0 and requests 2.34.2. Keep `serve_knowline.sh`, `knowline_server.py`, `knowline_layout.py`, `knowline_engine.py` and `temperatures.json` together in the model directory. ## Start the server From the downloaded model directory: ```bash pip install "sglang==0.5.21" "transformers==5.12.1" "requests==2.34.2" sha256sum -c SHA256SUMS MEM=0.90 bash serve_knowline.sh . 0 8080 ``` The launcher starts a bf16 SGLang server on port 9080 and the `/v1/systemone` front end on port 8080. It uses one GPU, tensor parallel size 1, an fp32 LM head, 16 front-end workers and the model's `temperatures.json`. The box prompt and coordinate settings are `gemma / yxyx`. `MEM=0.90` matches the full public evaluation. Without an override, the launcher defaults to 0.85. `HOST` defaults to `127.0.0.1`. ```bash curl -s http://127.0.0.1:8080/v1/systemone -H 'Content-Type: application/json' -d '{ "model": "m", "state": "Customer: my order arrived broken, I want my money back.", "questions": {"refund": {"type": "noul", "instructions": "Should the agent offer a refund?"}} }' ``` ## Reproduce the public evaluation The official Decision Index kit is pinned to `62d2f51de34a2de64906345b6bc3e98e27ff55c7`. Rebuild the public 0.3 suite locally following the [kit instructions](https://github.com/apolinario/decision-index#quickstart-local); the benchmark inputs are not distributed with this model. With the model server running and the rebuilt `suite-0.3` directory available in the kit checkout: ```bash git clone https://github.com/apolinario/decision-index cd decision-index git checkout 62d2f51de34a2de64906345b6bc3e98e27ff55c7 pip install -e . python -m decision_index pipeline --edition 0.3 --engine http \ --option base_url=http://127.0.0.1:8080 --option model=m \ --out runs/KnowLine-12B-Gen1 ``` The submitted run used two NVIDIA H20 96 GB GPUs, one identical server per GPU, with 16 round-robin client shards in total. It was a fresh full run with no results carried over from another model or run. One failed request was retried once. The result file preserves both attempts, with identical payload hashes: 140,621 records for 140,620 unique requests. All requests succeeded after the retry; the official scorer resolves the retry records. - Public index: **69.37**; raw index: **76.56**. - Requests: 140,620; scoreable: 140,178 after the 442 official exclusions. - All scoreable requests were answered; no errors or unsupported requests; `complete: true`. - Runner SHA256: `f4c9f36c4c557f537f1d8184db7bf00c5a84d9f4f0aa67f8c06198cead25818d`. - Results: [PelaAI/KnowLine-12B-Gen1-decision-index](https://huggingface.co/datasets/PelaAI/KnowLine-12B-Gen1-decision-index/tree/main/runs/KnowLine-12B-Gen1). To re-score the downloaded results with the same suite: ```bash python -m decision_index score --edition 0.3 \ --suite-dir suite-0.3 --results /path/to/results.jsonl.gz \ --engine http --out rescore ``` `SHA256SUMS` records the model package's file hashes. The public index is separate from the private Full score, which is run by the Decision Index maintainers.