Non-autoregressive System 1 decision engine. Typed choice, score and yes/no decisions over any text in a single forward pass, in 100+ languages, with a router that picks the right checkpoint per request.
Python calibrationclassificationdecision-modelhuggingfacejevmodernbertmultilingualnlppythonpytorchroutingtyped-decisionszero-shot
Commits on main
feat(serve): forward every /v1/systemone control a JSON body can state, refuse the hook five
2f6f4bc Aashish committed 1 month ago
feat(ts): per-checkpoint SHA-256 artifact verification in Router
b48756d Swatantra Yadav committed 1 month ago
test(docker): pin child exit-code propagation; qualify the "replace" docstring
93d0bf7 liwenjie200543 committed 1 month ago
fix(email): keep a mark that has no base character to ride on
ac91ac8 Denimworld12 committed 1 month ago
fix(serve): review fixes -- exact container types, list coverage, and three wrong claims
90ba149 Adeshina Lasisi committed 1 month ago
docs: engineering notes for compile=True and the TileLang fast path
984b90b cklxx committed 1 month ago
feat(agent): opt-in warmup() so compile=True does not stall the first requests
ecd3f9e cklxx committed 1 month ago
fix(onnx): export at batch 2 with explicit Dims so the graph runs at any batch size
0e88902 cklxx committed 1 month ago
docs(common): correct the direction claim, and make the README check real
b19118b Parswanadh committed 1 month ago
fix(agent): comprehensive input validation for questions, instructions, and state
211de25 Swatantra Yadav committed 1 month ago
fix(docker): forward LAYA_REVISION so a Compose deployment can actually pin
d714645 Bruce-Yii committed 1 month ago
fix(evals): the run identity was narrower than the rule it justifies itself with
6bbb338 Bruce-Yii committed 1 month ago
fix(ts): let emailState set the body budget, as Python does (#589 parity)
f6ccc0d divit aggarwal committed 1 month ago
fix(ts): refuse a temperature that is not a list of 3 at construction, as Python does (#502 parity)
53ac099 divit aggarwal committed 1 month ago
fix(ts): reject choice labels that share an answer key, as Python does (#496 parity)
b26433f divit aggarwal committed 1 month ago
fix(ts): reject a null choice label, as Python does (#508 parity)
738c04a divit aggarwal committed 1 month ago
fix(examples): the demo's model field must resolve names the way core resolves one
49aee9a Aashish committed 1 month ago
ci(release): run the truncation usage test before publishing
2ba7a39 Som Tripathi committed 1 month ago
docs(agent,onnx): say predict_long's truncation keys can disagree
054e4d8 Som Tripathi committed 1 month ago
fix(benchmarks): keep bench defaults on gitignored fresh-run paths (#134)
314cd5b Nafeel005 committed 1 month ago
fix(serve): the early refusal must report a length it measured, not cap + 1
510ab7e Adeshina Lasisi committed 1 month ago
perf(serve): refuse a provably oversized state without serializing it
3f67b7a Adeshina Lasisi committed 1 month ago
test(serve): give the limit cases short ids so Windows can run them
6ad84ff Adeshina Lasisi committed 1 month ago
feat(finetune): add single-device training script for any checkpoint
369a8a5 Wellington Fonseca committed 1 month ago
fix(serve): measure the state limit on the text the tokenizer receives
d8038b8 Adeshina Lasisi committed 1 month ago
docs(evals): correct the percentile rationale and harden the test
9e02e38 rahul05ranjan committed 1 month ago
perf(agent): reuse CUDA autocast weight cache across batches
06fe814 周诚宇(Zhou Chengyu) committed 1 month ago
research: accuracy against evidence position at fixed document length
52328d0 Parswanadh committed 1 month ago