Publish qwen-3-reranker-0-6b-f16-af32 registry metadata
Browse files- registry/catalog.json +77 -0
registry/catalog.json
CHANGED
|
@@ -559,6 +559,83 @@
|
|
| 559 |
"runtimePromotionState": "manifest-owned",
|
| 560 |
"weightsRefAllowed": false
|
| 561 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 562 |
{
|
| 563 |
"modelId": "qwen-3-5-0-8b-q4k-ehaf16",
|
| 564 |
"family": "qwen3",
|
|
|
|
| 559 |
"runtimePromotionState": "manifest-owned",
|
| 560 |
"weightsRefAllowed": false
|
| 561 |
},
|
| 562 |
+
{
|
| 563 |
+
"modelId": "qwen-3-reranker-0-6b-f16-af32",
|
| 564 |
+
"family": "qwen3",
|
| 565 |
+
"label": "Qwen3 Reranker 0.6B (F16/F32a)",
|
| 566 |
+
"description": "Qwen3 0.6B SentenceTransformers CrossEncoder reranker target with manifest-owned true/false token scoring. The F16 lane passes the semantic rerank suite and is the supported reranker variant for publication.",
|
| 567 |
+
"modes": [
|
| 568 |
+
"rerank"
|
| 569 |
+
],
|
| 570 |
+
"aliases": [
|
| 571 |
+
"qwen3-reranker-0.6b",
|
| 572 |
+
"qwen3-reranker-0-6b",
|
| 573 |
+
"qwen3-reranker-0.6b-f16",
|
| 574 |
+
"Qwen/Qwen3-Reranker-0.6B"
|
| 575 |
+
],
|
| 576 |
+
"artifact": {
|
| 577 |
+
"format": "rdrr"
|
| 578 |
+
},
|
| 579 |
+
"hf": {
|
| 580 |
+
"repoId": "Clocksmith/rdrr",
|
| 581 |
+
"revision": "cc1fafff8cda609372b608cd92e487f6a2c32bc8",
|
| 582 |
+
"path": "models/qwen-3-reranker-0-6b-f16-af32"
|
| 583 |
+
},
|
| 584 |
+
"baseUrl": null,
|
| 585 |
+
"quickstart": false,
|
| 586 |
+
"demoVisible": false,
|
| 587 |
+
"sizeBytes": 1191553024,
|
| 588 |
+
"recommended": false,
|
| 589 |
+
"sortOrder": 19.2,
|
| 590 |
+
"lifecycle": {
|
| 591 |
+
"availability": {
|
| 592 |
+
"curated": false,
|
| 593 |
+
"local": true,
|
| 594 |
+
"hf": true
|
| 595 |
+
},
|
| 596 |
+
"status": {
|
| 597 |
+
"runtime": "active",
|
| 598 |
+
"conversion": "ready",
|
| 599 |
+
"demo": "none",
|
| 600 |
+
"tested": "verified"
|
| 601 |
+
},
|
| 602 |
+
"tested": {
|
| 603 |
+
"suite": "rerank",
|
| 604 |
+
"surface": [
|
| 605 |
+
"node"
|
| 606 |
+
],
|
| 607 |
+
"result": "pass",
|
| 608 |
+
"lastVerifiedAt": "2026-07-04",
|
| 609 |
+
"source": "runtime-verify",
|
| 610 |
+
"contracts": {
|
| 611 |
+
"executionContractOk": true
|
| 612 |
+
},
|
| 613 |
+
"notes": "Release evidence: reports/release-claims/qwen-3-reranker-0-6b-f16-af32/2026-07-04T02-00-00.000Z.json. Semantic pair accuracy 1.0 (5/5), WebGPU query topDocumentIndex 0, rerankMs 4057.56 for the three-document smoke request. Speed evidence: reports/benchmarks/qwen-3-reranker-0-6b-f16-af32/2026-07-04T02-15-00.000Z.json (medianRerankMs 3977.72 for the same three-document request)."
|
| 614 |
+
}
|
| 615 |
+
},
|
| 616 |
+
"verify": {
|
| 617 |
+
"workload": "rerank",
|
| 618 |
+
"runtimeConfig": {
|
| 619 |
+
"inference": {
|
| 620 |
+
"rerank": {
|
| 621 |
+
"query": "What API exposes GPU hardware for browser compute workloads?",
|
| 622 |
+
"documents": [
|
| 623 |
+
"WebGPU provides low-level access to graphics processors for compute workloads.",
|
| 624 |
+
"WebSocket provides full-duplex communication over TCP.",
|
| 625 |
+
"CSS grid arranges page layout rows and columns."
|
| 626 |
+
]
|
| 627 |
+
}
|
| 628 |
+
}
|
| 629 |
+
}
|
| 630 |
+
},
|
| 631 |
+
"vendorBenchmark": null,
|
| 632 |
+
"sourceCheckpointId": "Qwen/Qwen3-Reranker-0.6B",
|
| 633 |
+
"weightPackId": "qwen-3-reranker-0-6b-f16-af32-wp-catalog-v1",
|
| 634 |
+
"manifestVariantId": "qwen-3-reranker-0-6b-f16-af32-mv-exec-v1",
|
| 635 |
+
"artifactCompleteness": "complete",
|
| 636 |
+
"runtimePromotionState": "manifest-owned",
|
| 637 |
+
"weightsRefAllowed": false
|
| 638 |
+
},
|
| 639 |
{
|
| 640 |
"modelId": "qwen-3-5-0-8b-q4k-ehaf16",
|
| 641 |
"family": "qwen3",
|