Speculative Decoding & Text Generation
31 repos
Techniques and implementations for accelerating large language model inference through speculative decoding, where a smaller draft model generates candidate tokens that a larger verifier model accepts or rejects in parallel. The cluster centers on optimized text generation pipelines, including quantized model variants (Qwen3, Gemma) and supporting infrastructure like SafeTensors format for efficient model loading. Most repositories appear to be model weights, checkpoints, or minimal wrapper implementations rather than core algorithmic frameworks.
Python · 1
speculative-decoding ·1,694
safetensors ·1,677
text-generation ·1,667
qwen3 ·1,620
transformers ·1,616
text-generation-inference ·1,612
dflash ·1,384
diffusion-language-model ·1,307
efficiency ·1,307
qwen ·1,266
