Commit History
6 days ago
-
docs: LoRA branch engines for sa3-sm-music and sa3-sm-sfx (193 targets each)
Verified signature08b1742 -
docs: LoRA branch engines for sa3-sm-music and sa3-sm-sfx (193 targets each)
Verified signature0b9ad3d -
drop the medium DiT .xnnwc weight caches (regenerated locally)
Verified signatureNot in the install manifest (never auto-downloaded); the runtime rebuilds the cache locally on first run (weight_cache_path=auto). Measured benefit of a shipped cache is only ~2.4s (construct) — per-rung time is forward compute, unaffected. Removes ~7 GB of dead weight, and a locally-built cache is safer (a cache built on one CPU may not load on another). Matches the small DiTs, which ship no .xnnwc.3fb5698 -
tensorRT: add sa3-sm-sfx LoRA branch engine (sm_90, 193 targets)
Verified signature8acb4d0 -
tensorRT: add sa3-sm-music LoRA branch engine (sm_90, 193 targets)
Verified signature049c3f3 -
small DiTs (sm-music, sm-sfx) tflite: ship the w8a8 rung ladders
Verified signatureReplace the varlen small DiTs with from-repo rung ladders: dit_{fp32,w8a8}.tflite (8 rungs to the 2-min max 1292, gce-outside, gcond bundled). Flat ~0.8 GB RAM vs the old varlen's ~3.8 GB at 2 min (4.5x less), ~equal quality to w8a8-dyn (cos>=0.996), cache-safe. Default is now w8a8 for every family. Old tiers (fp32/w16a32/w8a32/w8a8-dyn) -> legacy/. Pairs with stable-audio-3 PR #118. (No .xnnwc: the small w8a8 packs in ~1s.)91ec853 -
tensorRT/sm_90: add the sa3-m runtime LoRA engine (229 targets, rank<=512)
Verified signature46e266c -
tensorRT: document the runtime LoRA / DoRA engine
Verified signature75d47bb -
tensorRT: document the runtime LoRA / DoRA engine
Verified signature1804a18
7 days ago
-
tflite README: medium DiT default -> w8a8 + SAME-S
Verified signatureDocs sync for stable-audio-3 PR #114: medium DiT now defaults to the cache-safe rung w8a8 (fp32 lazy), and the default decoder is SAME-S (SAME-L opt-in). The model files were already swapped in the prior commit; this only updates the README copy.bb411c3 -
medium DiT tflite: ship the N=12 rung ladder (fp32 + cache-safe w8a8)
Verified signatureReplace the varlen DiT with the from-repo rung ladder: dit_{fp32,w8a8}.tflite (12 rungs, gce-outside, gcond bundled, +.xnnwc caches). Flat ~2-4 GB RAM vs the old 3-40 GB (up to 10x less at long lengths), equal quality to w8a8-dyn (cos 0.996). Old tiers (fp32/w16a32/w8a32/w8a8-dyn) -> legacy/. Pairs with stable-audio-3 PR #112.10e8a13
10 days ago
-
Update README.md
Verified signaturea010903
-
tflite: bake output limiter into SAME-S/SAME-L decoders (ceiling 0.977, matches TRT graft)
Verified signatureda6edc5 -
tflite: bake output limiter into SAME-S/SAME-L decoders (ceiling 0.977, matches TRT graft)
Verified signature7d85335 -
model card: note optimized TFLite/LiteRT rung codecs
Verified signature7f26210 -
tflite: SAME codecs -> static rung models (fp32 + w8a8); retire w16a32/w8a32/w8a8-dyn to legacy/
Verified signature73fa327
-
tflite: preserve pre-rung SAME codecs under legacy/
Verified signature416a9d0
-
Move the retired fp8_fast ONNX under onnx/same-s/legacy/
Verified signatureThe SAME-S-only fp8_fast tier pushed fp8 into the attention projections as well, buying +0.08x (1.14 -> 1.22x) at a real quality cost, and was retired. Nothing builds from these graphs and --dec-precision no longer accepts the value.b5182df -
Rename the autoencoder ONNX to match the engine naming scheme
Verified signaturedec_*_chunkable_limiter.onnx -> dec_*_limiter.onnx {enc,dec}_dynamic_triton_swa.onnx -> {enc,dec}_fp16.onnx (SAME-L) {enc,dec}_dynamic_bf16.onnx -> {enc,dec}_bf16.onnx (SAME-S) 'chunkable' was never a property of the ONNX: it is the two optimization profiles the TRT builder adds, so the engines keep the name and the graphs drop it. 'dynamic_triton_swa' named the samel::diff_attn_swa plugin node rather than a precision, which put it out of step with every other file here. Engine filenames are unchanged.4ea87d6 -
Move the retired w8_bf16 ONNX under onnx/*/legacy/
Verified signatureint8 weight-only storage. Its DequantizeLinear nodes sit on constant inputs, so TensorRT folds them at build time and the resulting engine is identical to the one built from the plain bf16 graph -- a file-size format, not a quantisation tier. No build target references them.74cd2b7 -
Delete the miscalibrated sm_120 SAME-S fp8 encoder
Verified signatureIts activation quantisers were calibrated with a percentile-of-percentile plus a 1e-4 floor, putting the clip points far below where the decoder's plain amax puts them. The equivalent sm_90 engine was deleted when that was found; this copy was missed and only surfaced when the deprecated engines were swept into legacy/. Keeping it under legacy/ would have left a trap: it is the one file there that is wrong rather than merely superseded, and anyone building sm_120 from it would inherit the bug. The recalibrated weights live in onnx/same-s/enc_fp8.onnx, which build_autoencoders.py uses to produce enc_fp8_chunkable.trt.197c684 -
Move deprecated sm_120 TRT engines under legacy/
Verified signatureThese are not reachable from the runtime registry: every tier resolves to the chunkable engines (dec_*_chunkable_limiter / enc_*_chunkable). Kept rather than deleted so older checkouts and reproductions still have them, but moved out of the way so the live set is unambiguous. Copies are server-side; the ONNX sources are untouched, and four of them (enc_dynamic_triton_swa.onnx, enc_dynamic_bf16.onnx, both enc_fp8.onnx) remain load-bearing for build_autoencoders.py -- 'chunkable' is a profile choice for the encoders, not a graph change.19ab22a -
Move deprecated sm_90 TRT engines under legacy/
Verified signatureThese are not reachable from the runtime registry: every tier resolves to the chunkable engines (dec_*_chunkable_limiter / enc_*_chunkable). Kept rather than deleted so older checkouts and reproductions still have them, but moved out of the way so the live set is unambiguous. Copies are server-side; the ONNX sources are untouched, and four of them (enc_dynamic_triton_swa.onnx, enc_dynamic_bf16.onnx, both enc_fp8.onnx) remain load-bearing for build_autoencoders.py -- 'chunkable' is a profile choice for the encoders, not a graph change.ceaecd4
-
build-ready ONNX: limiter grafted + ceiling baked (decoders), amax-recalibrated activation scales (fp8 encoders)
Verified signatured070aac -
build-ready ONNX: limiter grafted + ceiling baked (decoders), amax-recalibrated activation scales (fp8 encoders)
Verified signatureb1400fd -
build-ready ONNX: limiter grafted + ceiling baked (decoders), amax-recalibrated activation scales (fp8 encoders)
Verified signature3f689ec -
build-ready ONNX: limiter grafted + ceiling baked (decoders), amax-recalibrated activation scales (fp8 encoders)
Verified signaturef29bd0e -
build-ready ONNX: limiter grafted + ceiling baked (decoders), amax-recalibrated activation scales (fp8 encoders)
Verified signature9aa2165 -
build-ready ONNX: limiter grafted + ceiling baked (decoders), amax-recalibrated activation scales (fp8 encoders)
Verified signaturedcaaf84 -
retire fp8_fast; remove the superseded sm_90 fp8 engines (miscalibrated encoder, 8192-ceiling decoders)
Verified signature3963f0e -
retire fp8_fast; remove the superseded sm_90 fp8 engines (miscalibrated encoder, 8192-ceiling decoders)
Verified signaturef061405 -
retire fp8_fast; remove the superseded sm_90 fp8 engines (miscalibrated encoder, 8192-ceiling decoders)
Verified signature4b4cf7c -
retire fp8_fast; remove the superseded sm_90 fp8 engines (miscalibrated encoder, 8192-ceiling decoders)
Verified signature935685c -
retire fp8_fast; remove the superseded sm_90 fp8 engines (miscalibrated encoder, 8192-ceiling decoders)
Verified signature276f110 -
retire fp8_fast; remove the superseded sm_90 fp8 engines (miscalibrated encoder, 8192-ceiling decoders)
Verified signature0996c8a -
retire fp8_fast; remove the superseded sm_90 fp8 engines (miscalibrated encoder, 8192-ceiling decoders)
Verified signaturecf76999 -
SAME-L/SAME-S: bake the limiter ceiling so the chunkable decoders keep the ('latent','pcm') signature; add the SAME-S fp8 chunkable pair
Verified signature17258a7 -
SAME-L/SAME-S: bake the limiter ceiling so the chunkable decoders keep the ('latent','pcm') signature; add the SAME-S fp8 chunkable pair
Verified signature7ca4180 -
SAME-L/SAME-S: bake the limiter ceiling so the chunkable decoders keep the ('latent','pcm') signature; add the SAME-S fp8 chunkable pair
Verified signature775a981 -
SAME-L/SAME-S: bake the limiter ceiling so the chunkable decoders keep the ('latent','pcm') signature; add the SAME-S fp8 chunkable pair
Verified signatureaea1cf0 -
SAME-L/SAME-S: bake the limiter ceiling so the chunkable decoders keep the ('latent','pcm') signature; add the SAME-S fp8 chunkable pair
Verified signature9a01fee -
SAME-S: add the chunkable bf16 pair (two profiles, limiter)
Verified signaturebaa3cd7 -
SAME-S: add the chunkable bf16 pair (two profiles, limiter)
Verified signature7c73bf8 -
SAME-L encoders: low band 64 -> 256 (1.3-1.5x faster, cos 0.99831 -> 0.99969)
Verified signature1a8702f -
SAME-L encoders: low band 64 -> 256 (1.3-1.5x faster, cos 0.99831 -> 0.99969)
Verified signatured0d486b
-
SAME-L: remove miscalibrated enc_fp8.trt (activation scales 1.3-28x too small; superseded by enc_fp8_chunkable.trt on sm_90)
Verified signature2c8b02a -
SAME-L: remove miscalibrated enc_fp8.trt (activation scales 1.3-28x too small; superseded by enc_fp8_chunkable.trt on sm_90)
Verified signaturee00128c -
SAME-L: add enc_fp8_chunkable.trt (fp16, two-profile chunkable)
Verified signature7036ca2 -
SAME-L: add dec_fp8_chunkable_limiter.trt (fp16, two-profile chunkable, runtime limiter ceiling)
Verified signature093c4d6 -
SAME-L: add enc_fp16_chunkable.trt (fp16, two-profile chunkable)
Verified signatureeb343c9