| --- |
| license: apache-2.0 |
| base_model: wfzyx/von |
| pipeline_tag: zero-shot-classification |
| tags: |
| - verdict |
| - onnx |
| - system-one |
| - decision-models |
| - von |
| - serving |
| --- |
| |
| # Von (Verdict-ready) |
|
|
| The [Von](https://huggingface.co/wfzyx/von) System One decision model (ModernBERT-large, 395M, English, 8192-token context), exported to ONNX fp32 and packaged for [Verdict](https://github.com/sgyaqing/verdict) -- a high-performance serving engine for System One decision models. Serve it with one command: |
|
|
| ```bash |
| # CPU host |
| docker run --rm -p 8080:8080 \ |
| -v ~/.cache/verdict:/root/.cache/verdict \ |
| sgyaqing/verdict:latest-cpu serve verdict-models/von --host 0.0.0.0 |
| |
| # GPU host |
| docker run --rm --gpus all -p 8080:8080 \ |
| -v ~/.cache/verdict:/root/.cache/verdict \ |
| sgyaqing/verdict:latest-cuda serve verdict-models/von --host 0.0.0.0 |
| |
| # ModelScope: |
| docker run --rm -p 8080:8080 \ |
| -e VERDICT_USE_MODELSCOPE=1 \ |
| -v ~/.cache/verdict:/root/.cache/verdict \ |
| sgyaqing/verdict:latest-cpu serve sgyaqing/von --host 0.0.0.0 |
| ``` |
|
|
| Endpoints: the standard decision protocols `POST /v1/systemone` and `POST /v1/decisions`, native `POST /v1/choice | /v1/score | /v1/noul`, plus an OpenAI-compatible layer (`/v1/chat/completions`, `/v1/models`). |
|
|
| ## Performance |
|
|
| Measured on one RTX 4090, end to end over HTTP, against the model author's own server (`von serve`). A ratio above 1x means Verdict is ahead. |
|
|
| | concurrency | Verdict rps | `von serve` rps | rps ratio | Verdict P50 / P99 (ms) | `von serve` P50 / P99 (ms) | P50 / P99 ratio | |
| |---|---:|---:|---:|---:|---:|---:| |
| | 1 | 113.8 | 46.4 | 2.45x | 8.61 / 11.98 | 22.89 / 28.46 | 2.66 / 2.38x | |
| | 8 | 536.4 | 31.2 | 17.2x | 14.86 / 16.61 | 250.91 / 435.99 | 16.9 / 26.2x | |
| | 64 | 1476.1 | 22.3 | 66.1x | 42.58 / 53.24 | 2848.80 / 3333.40 | 66.9 / 62.6x | |
|
|
| ## Files |
|
|
| | File | Purpose | |
| |---|---| |
| | `model.json` | manifest: graph file name, budgets, family recipe (`independent_options`), calibration map | |
| | `von.onnx` + `von.onnx.data` | the ONNX graph (fp32, external weights) | |
| | `tokenizer/tokenizer.json` (+ `tokenizer_config.json`) | the tokenizer | |
| | `SHA256SUMS` | checksums for every file above | |
|
|
| ## Provenance & fidelity |
|
|
| Weights are unchanged from `wfzyx/von`; this repo adds only the ONNX export and the serving manifest. Conversion and serving are verified against the golden standard (official von-sdk 1.3.7, CPU fp32, 29-case corpus): identical answers on CPU and on CUDA (RTX 4090). |
|
|
| Accuracy PK — against `von serve` (von-sdk 1.3.7); both servers on one RTX 4090 (CUDA), graded against the same golden standard: |
|
|
| | | Verdict | `von serve` | result | |
| |---|---:|---:|:--| |
| | worst deviation | 0.0 | 0.0 | tied | |
|
|
| ## License |
|
|
| Apache-2.0, inherited from the base model. Credit for the model itself belongs to the [original author](https://huggingface.co/wfzyx); this repository only repackages it for serving. |
|
|
