Kazeia
/
Kazeia-engine
3
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
Kazeia-engine/dist/jni
History
Richard Loyer 4b43d9e3d7 chantier B STT #1 : S1 unification engine (mel partagé + skeleton stt + VAD C++) 
S1.1 mel extractor unifié (kazeia_mel.{h,cpp}) :
 - Paramétrable : window (sym/periodic), compression (log/log10), post-norm
   (none/whisper), FFT auto radix-2 (zero-pad N_FFT pas puissance de 2)
 - Configs pré-faites : config_qwen3_tts_speaker() + config_whisper()
 - speaker_encoder.cpp migré dessus : régression damien cos=0.9997 et bit-exact
   (max abs diff = 0) vs avant. E2E clonage in-process intact (RTF inchangé).

S1.2 audit ORT C++ :
 - libonnxruntime.so + headers C++ trouvés dans cache Gradle (AAR Maven
   onnxruntime-android-qnn:1.24.3). API_VERSION=24, QNN EP intégré.
 - headers/ copiés dans dist/include/onnxruntime/ (versionné, ~1 MB)
 - libonnxruntime.so copiée dans dist/lib/ (gitignore, 20 MB, README pour
   reproduire l'extraction).

S1.3 skeleton stt_engine :
 - stt_engine.{h,cpp} : API parallèle à tts_engine.h
   (SttEngineLoadCfg / SttTranscribeCfg / SttTranscribeResult)
 - VAD RMS C++ complet (port de VadStage.kt prod : frame=1600, seuil=150,
   3 frames speech, 8 frames silence)
 - stt_engine_load + stt_engine_transcribe : stubs err=-99 jusqu'au port ORT
   QNN en S2 (encoder NPU + decoder loop KV-cache + tokenizer BPE)
 - Binaire kazeia_stt_test (1.2 MB statique) valide mel Whisper [80,3000] +
   VAD sur audio FR 5s, plage de valeurs cohérente.

Reste S2 : port encoder/decoder NPU via Ort::Session + QNN EP, port tokenizer
BPE byte-level, override transcribe. Quand S2 done : S3 = JNI + façade Kotlin
+ suppression libmel_extractor.so / VadStage.kt / WhisperHybridEngine.kt.
 		2026-05-31 19:19:28 +02:00
..
EngineLlmEngine.kt chantier B TTS #1: API JNI embeds-only + M-RoPE qwen3 dense 2026-05-28 13:37:17 +02:00
TtsEngine.kt TU sampler tuning : top_p 1.0->0.95 + rep_penalty 1.05->1.10 nouveaux défauts 2026-05-31 11:18:25 +02:00
build_prefill.cpp chantier B TTS #4: P3.1+P3.2 = x_vector + prefill_embeds depuis input_ids (bit-exact) 2026-05-28 15:05:23 +02:00
cp_inference.cpp T2.1 CP heads NEON 16-way: CP -33%, pipeline -16%, bit-exact 2026-05-29 17:26:54 +02:00
cp_inference.h chantier F (CP HMX) : path sched livré, gain absent (overhead per-call) 2026-05-29 16:03:41 +02:00
cp_validate.cpp chantier B TTS #5: P3.3 = vrai CP runner intégré (bit-exact) 2026-05-28 15:11:07 +02:00
decoder_subprocess.cpp Decoder chraac en sous-process : -20% decoder, -9% pipeline, talker intact 2026-05-30 18:14:09 +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 chantier B TTS #2: fix M-RoPE positions en embeds-mode + validation bit-exact 2026-05-28 13:54:03 +02:00
kazeia_mel.cpp chantier B STT #1 : S1 unification engine (mel partagé + skeleton stt + VAD C++) 2026-05-31 19:19:28 +02:00
kazeia_mel.h chantier B STT #1 : S1 unification engine (mel partagé + skeleton stt + VAD C++) 2026-05-31 19:19:28 +02:00
kazeia_text_tokenizer.cpp chantier B TTS #10: tokenizer Qwen3 BPE via llama_tokenize, texte arbitraire 2026-05-28 21:21:53 +02:00
kazeia_text_tokenizer.h chantier B TTS #10: tokenizer Qwen3 BPE via llama_tokenize, texte arbitraire 2026-05-28 21:21:53 +02:00
kazeia_tts_jni.cpp chantier B TTS #11: libkazeia_tts.so + Kotlin wrapper, JNI validé bout-en-bout 2026-05-28 21:54:21 +02:00
sampler.cpp chantier B TTS #7: sampler HF-style top_k+top_p+rep_penalty (générique) 2026-05-28 16:55:05 +02:00
sampler.h chantier B TTS #7: sampler HF-style top_k+top_p+rep_penalty (générique) 2026-05-28 16:55:05 +02:00
speaker_encoder.cpp chantier B STT #1 : S1 unification engine (mel partagé + skeleton stt + VAD C++) 2026-05-31 19:19:28 +02:00
speaker_encoder.h chantier B TTS #8 : speaker encoder embedded + clonage in-process 2026-05-31 15:54:29 +02:00
stt_engine.cpp chantier B STT #1 : S1 unification engine (mel partagé + skeleton stt + VAD C++) 2026-05-31 19:19:28 +02:00
stt_engine.h chantier B STT #1 : S1 unification engine (mel partagé + skeleton stt + VAD C++) 2026-05-31 19:19:28 +02:00
stt_test.cpp chantier B STT #1 : S1 unification engine (mel partagé + skeleton stt + VAD C++) 2026-05-31 19:19:28 +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_engine_2calls.cpp chantier B TTS #11: libkazeia_tts.so + Kotlin wrapper, JNI validé bout-en-bout 2026-05-28 21:54:21 +02:00
test_jni_native.cpp dist: multi-tour propre (generateRaw + ChatSession) + JNI validé end-to-end natif 2026-05-27 11:18:20 +02:00
test_jni_tts.cpp chantier B TTS #11: libkazeia_tts.so + Kotlin wrapper, JNI validé bout-en-bout 2026-05-28 21:54:21 +02:00
test_native.cpp v4 fix #277: ngl0 CPU decode 14tok/s, 4s/tour, usable 2026-05-25 10:18:31 +02:00
test_talker.cpp chantier B TTS #1: API JNI embeds-only + M-RoPE qwen3 dense 2026-05-28 13:37:17 +02:00
test_talker_replay.cpp chantier B TTS #2: fix M-RoPE positions en embeds-mode + validation bit-exact 2026-05-28 13:54:03 +02:00
test_tokenizer.cpp chantier B TTS #10: tokenizer Qwen3 BPE via llama_tokenize, texte arbitraire 2026-05-28 21:21:53 +02:00
tts_engine.cpp chantier B TTS #8 : speaker encoder embedded + clonage in-process 2026-05-31 15:54:29 +02:00
tts_engine.h chantier B TTS #8 : speaker encoder embedded + clonage in-process 2026-05-31 15:54:29 +02:00
tts_orchestrate.cpp chantier B TTS #6: P3.4+P3.5 = pipeline complet 100% tablette bout-en-bout 2026-05-28 15:25:00 +02:00
tts_pipeline.cpp chantier B STT #1 : S1 unification engine (mel partagé + skeleton stt + VAD C++) 2026-05-31 19:19:28 +02:00