Text Generation
Transformers.js
ONNX
llama
webgpu
q4f16
gptq
tool-calling
experimental
conversational
Instructions to use webbrain-one/webbrain-compass-tiny-v2.1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers.js
How to use webbrain-one/webbrain-compass-tiny-v2.1 with Transformers.js:
// npm i @huggingface/transformers import { pipeline } from '@huggingface/transformers'; // Allocate pipeline const pipe = await pipeline('text-generation', 'webbrain-one/webbrain-compass-tiny-v2.1');
File size: 1,316 Bytes
3632834 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 | {
"status": "webgpu-behavioral-evidence-audited",
"behavioralPass": false,
"gates": {
"structuredAtLeast93": true,
"strictAtLeast15": true,
"looseAtLeast39": true,
"antiAtMost1": false,
"zeroTransportErrors": true,
"malformedXmlAtMost1": true,
"zeroLengthLimited": true,
"actualWebGpu5090": true
},
"structured": 98,
"strict": 15,
"loose": 43,
"byVerdict": {
"ideal_name": 28,
"other": 43,
"anti": 3,
"ideal": 15,
"skipped": 11
},
"parserErrors": [],
"lengthLimited": 0,
"rawResponses": 194,
"pinnedInputs": 193,
"maxPromptTokens": 10479,
"baseline": {
"structured": 94,
"strict": 16,
"loose": 41
},
"quantization": "GPTQ asymmetric INT4 block32, train-only calibration, FP16 activations/cache; lossy, no BF16 equivalence claim",
"runtime": "Actual pinned WebBrain worker, Transformers.js4.2.0 / ORTWeb1.27.0, Chrome150 RTX5090",
"sampling": "same settings, but JS PRNG differs from Torch",
"endToEndBrowserTaskSuccessMeasured": false,
"auditedAt": 1789231413.1218278,
"originalAuditSha256": "43e5a2d408995796994245d396a1b61522fe1185574b29fb07e47ff8de7a7365",
"originalEvidenceRetained": true,
"ownerReleaseAuthorized": true,
"productionCertification": false
} |