status update

README.md

@@ -14,28 +14,28 @@ Environment-aware TTS training corpus, built from 4 speaker-diarization sources:
 M3SD (770 h) · AISHELL-4 (120 h) · MSDWILD (80 h) · CHiME-6 (40 h).
 Each row: `(environment_audio_source, speaker_audio_source, text, speech)` as 16 kHz mono FLAC bytes, with provenance.
 
-**Last status push:** 2026-05-14T08:13:03+00:00Z · status_updater since 2026-05-14T06:25:15+00:00Z
+**Last status push:** 2026-05-14T08:18:08+00:00Z · status_updater since 2026-05-14T06:25:15+00:00Z
 
 ## ⏳ ETA to completion
 
 | | wall-clock estimate |
 | --- | --- |
-| **Total remaining** | **9.3 h** |
-| Current stage (MSDWILD) | 22 min |
-| AISHELL-4 + CHiME-6 stages | 1.5 h |
+| **Total remaining** | **8.5 h** |
+| Current stage (MSDWILD) | 21 min |
+| AISHELL-4 + CHiME-6 stages | 1.4 h |
 | M3SD baidu download | — |
-| M3SD processing | 7.4 h |
+| M3SD processing | 6.8 h |
 
-_Throughput so far: **104.5 audio-hours per wall-hour** (298.8 h processed in 1.8 h elapsed)_
+_Throughput so far: **113.6 audio-hours per wall-hour** (324.6 h processed in 1.9 h elapsed)_
 
 ## Live pipeline status
 
 | metric | value |
 | --- | --- |
 | pipeline alive | ✅ |
-| GPU utilisation | 95% |
+| GPU utilisation | 91% |
 | GPU VRAM used | 37067 / 97 887 MiB |
-| disk used | 163 / 200 GiB |
+| disk used | 159 / 200 GiB |
 | current stage | M3SD |
 
 ## Dataset progress
@@ -45,11 +45,11 @@ _Throughput so far: **104.5 audio-hours per wall-hour** (298.8 h processed in 1.
 | MSDWILD | 1575 | 14514 | ✅ done |
 | AISHELL-4 | 145 | 35252 | ✅ done |
 | CHiME-6 | 0 | 0 | live |
-| M3SD | 314 | 48948 | tarball ready (68.6GB) — needs extract |
+| M3SD | 360 | 55348 | tarball ready (68.6GB) — needs extract |
 
 ## Throughput
 
-- Records / min: **10.07**
+- Records / min: **10.03**
 - ETA for MSDWILD stage: **~2.6 h**
 - M3SD baidu download: **68.6 GB / 68.63 GB**