Unverified50% confidenceFactExact time
all-MiniLM-L6-v2 将文本压缩成 384 维稠密向量，模型体积约 80MB，底层基于 BERT 精简变体 MiniLM 通过知识蒸馏训练
1
Sources
50%
Confidence
Long-term
Relevance
7/14/2026
First Seen
Sources
大学生用Ollama自建跨模型持久记忆系统：解耦设计+AES256加密
Related Claims
UnverifiedQwen3.5-9B contains approximately 9 billion trainable parameters, and after 4-bit quantization, the model file size can be compressed to approximately 5–6 GB.73% similarUnverifiedUnsloth 通过 4-bit/8-bit 量化感知训练与动态激活卸载技术，可将 7B 模型的微调显存需求压缩至约 4–6 GB68% similarVerified4-bit quantization compresses each parameter to 0.5 bytes, theoretically compressing 140GB down to approximately 35–40GB.68% similarUnverifiedThe DeepSeek 8B model compresses the model file to approximately 4.7 GB through 4-bit quantization.66% similarUnverifiedPost-training quantization algorithms such as GPTQ and AWQ can compress models to 4-bit INT4, enabling 70B-parameter models to run on consumer-grade GPUs with a reduction in computation of more than 75%.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/503464API
curl https://kongchang.com/api/v1/knowledge/claims/503464MCP
get_claim(id=503464)