Transaction Science
22212badf2
wai-rs: hyper_scale_decoder — floor (>>) not truncation; byte-exact vs the REFERENCE decoder
...
The fully-integer hyper_scale_decoder does its per-channel ÷2^n as an ARITHMETIC
RIGHT SHIFT (floor), matching the real reference decoder. The exported ONNX
represents it as integer `Div` (truncation toward zero), which differs from `>>` on
negatives by 1 LSB — a ~1-LSB-wrong export. decode_hyper_scale used `/` (matching the
ONNX); switch it to `>>` (matching the reference).
Impact, on the real test256.bin hyper-latent (z[1,160,4,4] → i_s[1,160,16,16]): the
old truncation matched only 30765/40960 of the reference i_s (the rest off by exactly
+1, at the negative-remainder positions surfaced by the final abs); with `>>` it is
40960/40960 BYTE-EXACT vs the reference decoder.
Retargets tests/jpegai_hyperscale_calibration.rs from the (truncating) onnx.reference
golden to the real reference-decoder i_s — the correct target — validated byte-exact
on real data. Structural lib test unaffected (shape/range/determinism).
Co-Authored-By: Claude Opus 4.8 <[email protected] >
2026-07-09 07:01:28 -04:00
..
vectors
wai-web: decodeIntSplat binding + per-attr ranges for the in-browser 3D demo
2026-06-28 09:49:40 -04:00
carriage.rs
wai: prove carriage of every registered class + decode the two open flagships
2026-06-18 07:36:43 -04:00
ffi_robustness.rs
wai: gate codec-dependent test/example targets on codecs
2026-06-17 17:36:39 -04:00
fuzz_smoke.rs
wai: gate codec-dependent test/example targets on codecs
2026-06-17 17:36:39 -04:00
generate.rs
wai: generative-replicate — deterministic byte-exact GENERATION (wai.generate.spectral)
2026-06-25 18:48:21 -04:00
int_audio.rs
wai: learned integer-exact audio codec (wai.audio.int_codec) + conformance
2026-06-27 17:01:31 -04:00
int_hyper_synth.rs
wai: committed tiny conformance vector for int_hyper (CI-verifiable, no 23MB model)
2026-06-25 13:58:02 -04:00
int_splat.rs
wai: learned integer-exact Gaussian-splat attribute codec (wai.splat.int_codec)
2026-06-28 09:14:46 -04:00
int_video.rs
wai.video.int_motion: deterministic integer inter-frame video (byte-exact)
2026-06-27 07:22:53 -04:00
int_video_halfpel.rs
wai: half-pel motion compensation for int_motion video + real-footage conformance
2026-06-27 11:10:48 -04:00
int_video_keyframe.rs
wai.video.int_hyper: neural-keyframe video — learned int_hyper I-frame + integer P-frames
2026-06-27 07:34:20 -04:00
int_video_sota_keyframe.rs
wai.video.int_hyper at SOTA scale: real 256x256 neural-keyframe video
2026-06-27 08:08:21 -04:00
int_world.rs
wai: integer-exact interactive world model — the byte-exact renderer for wai.world.model
2026-06-27 18:34:30 -04:00
int_world2.rs
wai: two-agent co-control world.model — both agents drive one byte-exact frame
2026-06-28 03:14:57 -04:00
int_world_shared.rs
wai: shared / multi-agent world.model — convergence on a GENERATED world
2026-06-27 19:21:42 -04:00
jpegai_e2e.rs
wai: derive JPEG-AI residual coding mask from the bitstream (not a fixture)
2026-06-21 08:06:19 -04:00
jpegai_e2e_dequant.rs
wai-rs: jpegai_pipeline::dequant + e2e integration (entropy→dequant byte-exact, synthesis Y/UV on real latent)
2026-07-08 07:14:26 -04:00
jpegai_e2e_synthesis.rs
wai-rs: jpegai_pipeline::dequant + e2e integration (entropy→dequant byte-exact, synthesis Y/UV on real latent)
2026-07-08 07:14:26 -04:00
jpegai_hyper_calibration.rs
wai-rs: jpegai_synthesis — restore chroma + hyper decoders, add hyper_scale (all byte-exact) ( #53 )
2026-07-07 15:08:25 -04:00
jpegai_hyperscale_calibration.rs
wai-rs: hyper_scale_decoder — floor (>>) not truncation; byte-exact vs the REFERENCE decoder
2026-07-09 07:01:28 -04:00
jpegai_live.rs
wai: JPEG-AI is a live, self-sufficient capability (no schedule fixtures)
2026-06-22 03:11:37 -04:00
jpegai_mans_calibration.rs
wai-rs: byte-exact mANS entropy decoder (jpegai_mans) — sgm + fac paths ( #55 )
2026-07-07 22:38:36 -04:00
jpegai_mans_golden.rs
wai: interactive + temporal cargo classes, provenance, telepresence
2026-06-13 23:16:42 -04:00
jpegai_mcm_calibration.rs
wai-rs: jpegai_synthesis — add MCM context-model decoders (mcm_stage0/mcm_stage_n) ( #54 )
2026-07-07 17:26:51 -04:00
jpegai_synthesis_calibration.rs
wai-rs: jpegai_synthesis — whole-net integer JPEG-AI synthesis decoder (58 dB PSNR) ( #49 )
2026-07-07 07:42:11 -04:00
jpegai_uv_calibration.rs
wai-rs: jpegai_synthesis — restore chroma + hyper decoders, add hyper_scale (all byte-exact) ( #53 )
2026-07-07 15:08:25 -04:00
neural_native_decode.rs
wai: reconcile the neural + semantic subsystem onto main
2026-06-14 03:55:29 -04:00
semantic_path.rs
wai: reconcile the neural + semantic subsystem onto main
2026-06-14 03:55:29 -04:00
world_model.rs
wai: wai.world.model — generative-world transport (replay-equivalent action trace + provenance)
2026-06-27 17:25:02 -04:00
world_replay.rs
wai: reconcile the neural + semantic subsystem onto main
2026-06-14 03:55:29 -04:00