FluidInference/Nemotron-3.5-ASR-Streaming-Multilingual-0.6b-CoreML

Correct license to OpenMDW-1.1 (commercial-ok) and fix base-model provenance

Update README.md

docs: ship iOS 17 as the canonical build (verified speed/WER parity with iOS 18)

Remove redundant ios17/ tree (promoted to canonical paths)

Promote iOS17 models to canonical latin//multilingual/ paths (replace iOS18)

Update README.md

fix ios17/latin/4480ms tokenizer: id 2224 <blank> -> ▁there

fix latin/4480ms tokenizer: id 2224 <blank> -> ▁there (surface corruption)

fix ios17/latin/2240ms tokenizer: id 2224 <blank> -> ▁there

fix latin/2240ms tokenizer: id 2224 <blank> -> ▁there (surface corruption)

fix ios17/latin/1120ms tokenizer: id 2224 <blank> -> ▁there

fix latin/1120ms tokenizer: id 2224 <blank> -> ▁there (surface corruption)

fix ios17/latin/560ms tokenizer: id 2224 <blank> -> ▁there

fix latin/560ms tokenizer: id 2224 <blank> -> ▁there (surface corruption)

README: document the ios17/ folder (iOS 17 deployment target)

ios17/multilingual/4480ms (iOS17 deployment target, same recipe)

ios17/multilingual/2240ms (iOS17 deployment target, same recipe)

ios17/multilingual/1120ms (iOS17 deployment target, same recipe)

ios17/multilingual/560ms (iOS17 deployment target, same recipe)

ios17/latin/4480ms (iOS17 deployment target, same recipe)

ios17/latin/2240ms (iOS17 deployment target, same recipe)

ios17/latin/1120ms (iOS17 deployment target, same recipe)

ios17/latin/560ms (iOS17 deployment target, same recipe)

manifest: 2 models (latin + multilingual) x 4 tiers, per-language benchmarks

README: collapse to latin/ + multilingual/ (2 models)

collapse: drop ja/ (identical to latin/ or multilingual/)

collapse: drop zh/ (identical to latin/ or multilingual/)

collapse: drop de/ (identical to latin/ or multilingual/)

collapse: drop pt/ (identical to latin/ or multilingual/)

collapse: drop it/ (identical to latin/ or multilingual/)

collapse: drop fr/ (identical to latin/ or multilingual/)

collapse: drop es/ (identical to latin/ or multilingual/)

collapse: drop en/ (identical to latin/ or multilingual/)

latin/4480ms (shared Latin-script-prune model: en/es/fr/it/pt/de)

latin/2240ms (shared Latin-script-prune model: en/es/fr/it/pt/de)

latin/1120ms (shared Latin-script-prune model: en/es/fr/it/pt/de)

latin/560ms (shared Latin-script-prune model: en/es/fr/it/pt/de)

v2 manifest: 36 ships, FLEURS benchmarks, mlmodelc-only B1

v2 README: all-FLEURS matrix, script-prune + full-vocab

v2 ja/4480ms (new ckpt, script-prune/full-vocab, LAYERPOS, B1) — drop stale K=4

v2 zh/4480ms (new ckpt, script-prune/full-vocab, LAYERPOS, B1) — drop stale K=4

v2 multilingual/4480ms (new ckpt, script-prune/full-vocab, LAYERPOS, B1) — drop stale K=4

v2 de/4480ms (new ckpt, script-prune/full-vocab, LAYERPOS, B1) — drop stale K=4

v2 pt/4480ms (new ckpt, script-prune/full-vocab, LAYERPOS, B1) — drop stale K=4

v2 it/4480ms (new ckpt, script-prune/full-vocab, LAYERPOS, B1) — drop stale K=4

v2 fr/4480ms (new ckpt, script-prune/full-vocab, LAYERPOS, B1) — drop stale K=4

v2 es/4480ms (new ckpt, script-prune/full-vocab, LAYERPOS, B1) — drop stale K=4

v2 en/4480ms (new ckpt, script-prune/full-vocab, LAYERPOS, B1) — drop stale K=4

v2 ja/2240ms (new ckpt, script-prune/full-vocab, LAYERPOS, B1) — drop stale K=4

v2 zh/2240ms (new ckpt, script-prune/full-vocab, LAYERPOS, B1) — drop stale K=4