Commit History
2 days ago
-
Add safeagents judge bench results
Verified signatureHi! This adds `vllm-sr/d3-mini`'s results on the [safeagents judge bench](https://huggingface.co/datasets/harness-hub/safeagents-judge-bench), a benchmark of real agent tool calls that [safeagents](https://github.com/huggingface/safeagents)' scanner flags as suspicious, each labelled block or allow and with a kind of misuse (reward hacking, harness introspection, credential misuse, ...). On the 562 test calls: balanced accuracy 0.692, AUROC 0.975, false-block rate 0.615, category macro F1 0.367. Run zero-shot, one System One request per call, on Apple M5 Pro (MPS). Per-call verdicts are in the dataset's `results/` folder. The file is named after the benchmark, so it doesn't touch any existing `.eval_results` file.09440cb
15 days ago
-
Add Typed Decisions benchmark eval results (#5)
- Add Typed Decisions benchmark eval results (7d893febc980b911297d139a6ebe91d19e681267) Co-authored-by: Nathan Habib <SaylorTwift@users.noreply.huggingface.co>bd0090d
17 days ago
-
Card: link the free API at decision.vllm-sr.ai and the official Space
Verified signature97dd054
18 days ago
-
Runtime: loading checks only the files the runtime reads
Verified signaturef15a639 -
Card: Engine mode batches by default; engine arguments replace runtime profiles
Verified signatureNames the two measured Engine mode setups for what they were: one request per forward, the former default (exact in the chart), and batching, the default in vLLM Semantic Router from main. The engine arguments --max-num-batched-tokens, --max-num-seqs, --batch-window-ms and --dtype replace the --runtime-profile flag. README.md only; every measured number is unchanged.c5aa751
19 days ago
-
Card: official Jev Decision Index 0.3.2 results and serving performance
Verified signature6d8971f
20 days ago
-
Runtime: faster System One server replies
Verified signature0e78bbd -
Runtime: simpler System One server
Verified signature39bcdf9 -
Card: shorter vLLM Semantic Router command
Verified signatureThe "Run with vLLM Semantic Router" section now uses `vllm-sr serve -e vllm-sr/d3-mini`: the platform is detected automatically, and one line lists the `--platform` values (auto, cpu, cuda, rocm; default auto). The parity sentence keeps its measurement context (AMD Instinct MI325X). MODEL_MANIFEST.json records the new README.md (SHA-256 and size) in the same commit. Weights, runtime files and the v3.1.0 tag are unchanged.6dfc4cc -
Card: run with vLLM Semantic Router (Engine mode)
Verified signatureAdds the "Run with vLLM Semantic Router" section: install the vllm-sr CLI from the development channel, `vllm-sr serve vllm-sr/d3-mini --engine --platform rocm`, and text, image and video requests to the System One API. MODEL_MANIFEST.json records the new README.md (SHA-256 and size) in the same commit. Weights, runtime files and the v3.1.0 tag are unchanged.40391eb
21 days ago
-
d3-mini v3.1.0
Verified signaturea475367 -
d3-mini v3.0.3
Verified signature94b1ea6
22 days ago
-
d3-mini v3.0.2
Verified signature1160a9d -
Update banner
Verified signature61dbd3a -
d3-mini v3.0.1
Verified signature840b5dd -
Model card: speed on AMD Instinct MI325X
Verified signature566da0e -
d3-mini
e200a5d