Unverified50% confidenceFactExact time
LoRA在前向传播时原始权重W₀保持冻结，输出变为h = W₀x + (BA)x/√r，对于SD 1.5模型在rank=4时整个LoRA文件约只有3-6MB
1
Sources
50%
Confidence
Long-term
Relevance
7/25/2026
First Seen
Sources
CivitAI抢先体验争议：开源精神与商业化边界在哪里？
redditr/StableDiffusion7/11/2026
Related Claims
VerifiedLoRA (Low-Rank Adaptation) files are typically only tens of MB in size, far smaller than the several GB size of complete models.67% similarUnverifiedLoRA 由微软提出，其核心思想是将权重更新量 ΔW 分解为两个低秩矩阵的乘积 BA66% similarUnverifiedLoRA (Low-Rank Adaptation) freezes the full weights of the pre-trained model and only injects two low-rank matrices into specific layers, resulting in file sizes that are typically only a few MB to tens of MB.65% similarUnverifiedMeta's LLaMA 2 70B model still requires approximately 35 GB of VRAM to fully load even with 4-bit quantization.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/615616API
curl https://kongchang.com/api/v1/knowledge/claims/615616MCP
get_claim(id=615616)