Unverified50% confidenceFactExact time
训练使用 batch 256、40 万步、EMA 0.9999、最后四分之一线性衰减学习率，torch.compile 相比 eager 提速 2.4×
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Cite This Claim
Stable URI
https://kongchang.com/claim/902333API
curl https://kongchang.com/api/v1/knowledge/claims/902333MCP
get_claim(id=902333)