78 repos across 4 sub-areas
Methods and implementations for efficiently adapting large language models through techniques like LoRA (Low-Rank Adaptation) and PEFT (Parameter-Efficient Fine-Tuning), reducing computational and memory costs compared to full model fine-tuning. The cluster includes practical applications across model families (Llama, others) and specialized domains like robotics control tasks, alongside foundational libraries and fine-tuned model variants that demonstrate these techniques in practice.
Cluster 637235
36 repos
Cluster 637234
27 repos
LoRA and PEFT for Language Models
8 repos
Parameter-efficient fine-tuning (PEFT) techniques, particularly Low-Rank Adaptation (LoRA), for adapting and customizing large language models with minimal computational overhead. The cluster includes implementations, adapters, and tools for efficiently training and deploying model variants across different domains and architectures, with support for formats like safetensors for model serialization.
LLM fine-tuning and parameter-efficient adaptation
7 repos
Techniques and frameworks for efficiently adapting large language models through parameter-efficient fine-tuning methods like LoRA and PEFT. This cluster covers both the core libraries that enable these adaptation methods and applied implementations across multimodal and specialized model architectures. Developers working in this area will find both foundational tools for adapting models to new tasks and concrete examples of those tools applied to vision-language models and other specialized LLM variants.