Sep 22 15:40 Rewrite card: shorter, plain sentences Aug 16 01:17 model card: 2026-08-16 release — UTMOSv2 2.919 (above GT 2.822, level with nemo44k 2.903 at ~1/10 bitrate), CV22 CER unchanged 7.70%; add progression chart 01:17 charts: 2026-08-16 release — UTMOSv2 2.788 -> 2.919 on the 9,291-clip wideband set, CV22 CER unchanged at 7.70% 01:17 charts: 2026-08-16 release — UTMOSv2 2.788 -> 2.919 on the 9,291-clip wideband set, CV22 CER unchanged at 7.70% Aug 15 23:31 clean-corpus finetune @50k : CER 7.70 macro / 7.09 micro (IDENTICAL to previous weights), wideband UTMOSv2 2.799->2.913 (+0.114), held-out MOS on 138 unseen sources 2.654->2.807 (+0.153, t=14.2), PESQ-WB +0.0170 (t=7.5). Same architecture and recipe; only the training corpus changed (clean DNSMOS-gated >=44kHz, CommonVoice excluded). Previous weights remain in history at sha256 8869a98b... Aug 3 02:25 Update README.md Jul 21 11:45 README: remove GitHub links (code closed-source; inference bundled in repo) 10:57 README: self-contained inference quickstart 10:57 Add neucodec package for self-contained inference 10:57 Add self-contained inference CLI Jul 20 09:40 Update README.md Jul 18 23:30 card: drop PESQ (favors waveform codecs); note multilingual set 23:30 plot: annotate UTMOSv2 (drop PESQ) 23:23 card: add PESQ column + bitrate-vs-UTMOSv2 plot 23:23 add bitrate-vs-UTMOSv2 plot 23:14 comparison: drop 24 kHz anchor (>=44k class only) 15:23 benchmark: final gs-2.0M WideCodec numbers (UTMOSv2 2.788, beats DAC) 08:17 refresh samples/after @ global_step 2000000 08:13 full resume ckpt @ global_step 2000000 08:13 weights @ global_step 2000000 Jul 17 00:43 intro: SOTA 0.8 kbps 44.1 kHz audio tokenizer 00:40 add 44k-class bitrate comparison table Jul 16 23:27 full resume ckpt @ global_step 1875000 23:27 weights @ global_step 1875000 11:08 rename NeuCodec-44k -> WideCodec in model card 01:48 remove stray samples/boboiboy/after/ path 01:48 fix boboiboy-after path (canonical) @ gs1795000 01:47 refresh samples/after @ global_step 1795000 Jul 15 22:58 full resume ckpt @ global_step 1785000 22:58 weights @ global_step 1785000 12:25 README: document stage-1 (noisy) revision + stage-2 latest 12:24 full resume ckpt from epoch=24-step=1745000.ckpt 12:24 Upload pytorch_model.bin with huggingface_hub Jul 10 04:33 full resume ckpt from epoch=9-step=1580000.ckpt 04:33 Upload pytorch_model.bin with huggingface_hub Jul 9 15:21 Upload samples/boboiboy/boboiboy-after.wav with huggingface_hub 15:21 samples: after (recon 44.1k d20) 15:21 samples: before (originals) 15:21 Delete folder samples/common-voice with huggingface_hub 15:01 full resume ckpt from epoch=9-step=1565000.ckpt 15:01 Upload pytorch_model.bin with huggingface_hub 01:16 Upload samples/boboiboy/boboiboy-after.wav with huggingface_hub 01:16 samples: after (recon 44.1k d20) 01:16 samples: before (originals) 01:16 Delete folder samples/common-voice with huggingface_hub 00:57 full resume ckpt from epoch=9-step=1510000.ckpt 00:56 Upload pytorch_model.bin with huggingface_hub Jul 8 02:32 Upload samples/boboiboy/boboiboy-after.wav with huggingface_hub 02:32 samples: after (recon 44.1k d20) 02:32 samples: before (originals)