Kazeia-engine/dist/jni
Richard Loyer 2c89ec01a2 dist: JNI option C câblé (prefill HTP / decode CPU), validé multi-tour
generate() = prefill prompt sur instance HTP (ngl99, device HTP0, OPFILTER=SSM_CONV +
GDN_PREFILL posés au load) -> llama_state_seq transfère le KV -> decode sur instance CPU
(ngl0, t4+fa, KV f16). 2 instances (~4.7GB). Sans état entre appels (app passe l'historique).
Validé multi-tour via jni/dual_ctx_mt.cpp: 3 tours cohérents FR + mémoire conversationnelle
(rappelle prénom/âge du tour 1 au tour 3), prefill HTP 103-144 t/s, decode CPU ~7-9.
Compile+linke OK (4 symboles JNI, CMakeLists inchangé). Rebuild libkazeia_engine.so requis
avant ship (le .so livré = ancienne version CPU-only/q8_0).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-27 10:24:55 +02:00
..
EngineLlmEngine.kt Bridge v3 utilisable: thinking-off ChatML, sys+usr, cap; static intact 2026-05-25 09:41:04 +02:00
dual_ctx.cpp dist: prefill HTP = CASSÉ (sortie charabia), prefill obligatoirement CPU 2026-05-27 09:42:59 +02:00
dual_ctx_mt.cpp dist: JNI option C câblé (prefill HTP / decode CPU), validé multi-tour 2026-05-27 10:24:55 +02:00
kazeia_engine_jni.cpp dist: JNI option C câblé (prefill HTP / decode CPU), validé multi-tour 2026-05-27 10:24:55 +02:00
test_engine.c Valide device: generate() bout-en-bout OK (test_native), bridge pret integration 2026-05-24 22:05:44 +02:00
test_engine.cpp Valide device: generate() bout-en-bout OK (test_native), bridge pret integration 2026-05-24 22:05:44 +02:00
test_native.cpp v4 fix #277: ngl0 CPU decode 14tok/s, 4s/tour, usable 2026-05-25 10:18:31 +02:00