iocuydi/amharic-llama-llava

# amharic-llama-llava Pretraining, finetuning, and inference for Amharic LLaMA and LLaVA adapted from: github.com github.com github.com Llama-2-Amharic weights: iocuydi/llama-2-amharic-378… Can be run with the inference script in this repo. Pretrained on 3.784b Amharic tokens. Amharic LLaVA weights: iocuydi/amharic-llava Can be run with the inference script in this repo. Must run with the PRETRAINED amharic llama, not finetuned. See llava eval section for more details. Amharic LLaVA requires this CLIP model: huggingface.co Associated datasets: Amharic Blip Laion: iocuydi/amharic-blip-laion Amharic Dolly: iocuydi/amharic-dolly-15k Amharic Alpaca: iocuydi/amharic-alpaca Amharic Visual Instruction Tuning: iocuydi/amharic-visual-inst… Amharic RedPajama Synthetic (pretraining, partial): iocuydi/amharic-redpajama-s… Amharic OASST1 Pruned: iocuydi/amharic-OASST1-prun… More info arxiv.org medium.com Cite: ``` @misc{andersland2024amharic, title={Amharic LLaMA and LLaVA: Multimodal LLMs for Low Resource Languages}, author={Michael Andersland}, year={2024}, eprint={2403.06354}, archivePrefix={arXiv}, primaryClass={cs.CL} } ```