Actions: NVIDIA/Model-Optimizer
Actions
Showing runs from all workflows
2,500+ workflow runs
2,500+ workflow runs
FP8_DEFAULT_CFG underperforms BF16 at prefill; NVFP4 much faster. Also: exported FP8 checkpoint has no KV q/k/v_scale
Claude Code Review
#2444:
Issue comment #2015 (comment)
created
by
ChenhanYu