77 repos across 3 sub-areas
Methods and implementations for efficiently adapting large language models through parameter-efficient fine-tuning techniques like LoRA and PEFT, reducing the computational and memory overhead of model customization. The cluster includes frameworks, adapter implementations, and applied examples across different model architectures and use cases, with particular focus on techniques that maintain model quality while minimizing trainable parameters.
Large Language Model Fine-tuning & Adaptation
33 repos
Parameter-efficient fine-tuning techniques and tools for adapting large language models to downstream tasks. These repositories focus on methods like LoRA (Low-Rank Adaptation) and PEFT frameworks that reduce computational costs while maintaining model quality, with support for various model architectures including Mistral, Qwen, and Gemma. The cluster also includes emerging applications in multimodal generation tasks like text-to-audio and text-to-video synthesis.
LLM Fine-tuning and Parameter-Efficient Adaptation
23 repos
Libraries and techniques for efficiently fine-tuning large language models using methods like LoRA and other parameter-efficient approaches. The cluster centers on frameworks like LlamaFactory and related tools that enable supervised fine-tuning, instruction tuning, and model adaptation without requiring full model retraining. Projects here demonstrate practical applications including domain-specific model variants (Bangla, Qwen-based), instruction-following improvements, and mixture-of-experts configurations.
Parameter-Efficient Fine-Tuning & LLM Adaptation
21 repos
Techniques and implementations for efficiently adapting large language models and vision-language models through parameter-efficient methods like LoRA and PEFT. The cluster covers fine-tuning frameworks, safetensor model formats for efficient storage and loading, and applications across text generation and multimodal tasks. Most repos focus on practical implementations of adapter-based training that reduce memory and computational requirements while maintaining model performance.