Defetya/opelllama-easylm-ru-7B

Open implementation llama 7B second stage pre-trained on Russian language. Not complete yet, about 200M tokens, achieves 3.5 perplexity on eval dataset