72 repos across 3 sub-areas
Quantized and fine-tuned versions of Meta's Llama 2 language models optimized for efficient deployment on consumer hardware. The cluster centers on GGML-quantized variants (7B, 13B, 70B parameters) and instruction-tuned derivatives like Speechless and Hermes, enabling practitioners to run capable LLMs locally without enterprise infrastructure. Repos here represent both model weights/artifacts and the tooling ecosystem around inference optimization.
Llama 2 Model Variants & Quantization
38 repos
Quantized and fine-tuned variants of Meta's Llama 2 language model, optimized for efficient local deployment using GGML and similar formats. The cluster centers on community adaptations of Llama 2 (particularly 7B and 13B parameter versions) that combine multiple training approaches—such as Hermes, Orca, Platypus, and Wizard—to improve instruction-following and reasoning capabilities while maintaining portability across consumer hardware.
Llama 2 Large Language Models
21 repos
Meta's Llama 2 family of open-source large language models in various sizes (7B, 13B, 70B parameters) and variants (base and chat-optimized versions). This cluster centers on the model repositories and implementations that enable researchers and practitioners to work with these popular foundation models for text generation and conversational AI tasks. The repositories here provide access to model weights, implementations, and integration with the PyTorch ecosystem.
Llama Model Variants and Inference
13 repos
Optimized versions and serving implementations of Meta's Llama language models, particularly the 3.x series. This cluster covers different quantization strategies, context window extensions, and inference frameworks (including text-generation-inference endpoints) that make these models practical for various deployment scenarios. Most repos are model variants rather than novel architectures—they represent engineering work around making Llama models faster, smaller, or more capable for specific use cases.