ruihan c8bfe6b9ab Compare MTP steps and prefill batches; retain qualified MTP 2 baseline

2026-09-18 23:22:45 +08:00

..

acceptance.py

Deploy audited RTX 6000D Flash-Next 128K multimodal baseline

2026-09-18 10:47:26 +08:00

benchmark.py

Deploy audited RTX 6000D Flash-Next 128K multimodal baseline

2026-09-18 10:47:26 +08:00

long_context.py

Deploy audited RTX 6000D Flash-Next 128K multimodal baseline

2026-09-18 10:47:26 +08:00

make_qualification_assets.py

Qualify and enable MTP 2 with paired workload evidence and fallback

2026-09-18 22:37:54 +08:00

monitor.py

Deploy audited RTX 6000D Flash-Next 128K multimodal baseline

2026-09-18 10:47:26 +08:00

multimodal.py

Deploy audited RTX 6000D Flash-Next 128K multimodal baseline

2026-09-18 10:47:26 +08:00

qualification.py

Compare MTP steps and prefill batches; retain qualified MTP 2 baseline

2026-09-18 23:22:45 +08:00

quality_tasks.py

Qualify and enable MTP 2 with paired workload evidence and fallback

2026-09-18 22:37:54 +08:00

summarize_qualification.py

Qualify and enable MTP 2 with paired workload evidence and fallback

2026-09-18 22:37:54 +08:00

summarize_tuning.py

Compare MTP steps and prefill batches; retain qualified MTP 2 baseline

2026-09-18 23:22:45 +08:00

tune_screen.py

Compare MTP steps and prefill batches; retain qualified MTP 2 baseline

2026-09-18 23:22:45 +08:00