A 4-bit NormalFloat (NF4) quantized version of InstaDeep's Nucleotide Transformer 2.5B Multi-Species model, optimized for significantly lower GPU memory consumption while preserving embedding quality and inference performance.
Fine tuned qwen llm using qloRa for commentary generation and generating audio using edge-tts.
A progressive hands-on lab series with the 124M-parameter GPT-2 as a compact teaching example, comprising 12 carefully designed post-training experiments, covering full-sequence SFT, response-only SFT, parameter efficient fine-tuning with LoRA and QLoRA, adding task-specific head for binary sentiment classification and Named Entity Recognition.
🚀 3.3B Mamba-MoE running on Mac Mini M4 16GB! Breakthrough deployment of Concept models on consumer Apple Silicon via NF4 quantization & QLoRA. Pushing limits of on-device AI.
spec-rama is a parameter-efficient fine-tuning (PEFT) framework that combines Weight Permutation (Greedy TSP / Bipartite) with Spectral Core Adaptation (DCT, FWHT, DWT Wavelet) and RAMA Multiplicative-Additive Modulation.
To associate your repository with the nf4-quantization topic, visit your repo's landing page and select "manage topics."