Ye YuandClaude Opus 5 c35f0aa92a launcher: end-to-end DFlash reference run for Qwen3-8B ...

Train -> serve -> benchmark in one YAML, so a handover has a single reproducible
chain rather than artifacts stitched from separate examples. Trains from the full
Spec-Decoding-Dataset-v2 with a bounded max_steps, so nothing needs staging first.

Signed-off-by: Ye Yu <yeyu@nvidia.com>
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Signed-off-by: Ye Yu <yeyu@nvidia.com>

2026-09-30 09:16:44 -07:00

..

ci

Reuse a whole recipe via $import, deprecate recipe_type, and start the published-checkpoint backfill with two aliases (#2376)

2026-09-21 17:12:31 -07:00

debugger

…

launcher

launcher: end-to-end DFlash reference run for Qwen3-8B

2026-09-30 09:16:44 -07:00

mcp

…

precommit

Reuse a whole recipe via $import, deprecate recipe_type, and start the published-checkpoint backfill with two aliases (#2376)

2026-09-21 17:12:31 -07:00

resource_monitor.py

Add sidecar GPU/CPU memory+utilization monitor for HF PTQ (#2000)

2026-07-30 06:07:11 +00:00