Unverified50% confidenceFactExact time
Rebiha offers an adjustable LoRA rank range of 16 to 256, along with three training presets: fast, balanced, and quality.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/11/2026
First Seen
Valid until: 12/10/2026
Sources
Related Entities
Related Claims
UnverifiedLoRA训练中常用的学习率为1e-4到5e-4之间，常用的Network Rank值为8、16、32、64、12860% similarUnverified千问3.8 27B基于千问3.5的整套架构体系，架构层面与3.5完全一致，仅做了继续预训练和后训练（含SFT与RL强化）57% similarUnverifiedCorpus 3.6 27B是开源社区基于阿里千问3.6（Qwen-3.6）进行后训练得到的优化版本，继承了千问3.6的256K长上下文能力55% similarUnverifiedQwen3覆盖从0.6B到235B的多种参数规模，支持思考模式（thinking mode）混合推理能力，覆盖119种语言的多语言预训练数据54% similarUnverifiedLoRA使可训练参数量通常仅为全参微调的0.1%-1%，可在单张RTX 3090/4090（显存24GB）上微调7B乃至13B级别的模型54% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/892731API
curl https://kongchang.com/api/v1/knowledge/claims/892731MCP
get_claim(id=892731)