Saiga/Llama3 8B, Russian Llama 3 based chatbot Based on Llama 3 8B Instruct. Llama.cpp version: link Colab: link Prompt format ОСТОРОЖНО! WARNING! LET OP! I've changed the prompt format from ChatML to the original Llama 3 format in v4 . Don't forget to switch formats! v4, v5, v6+ : LLama 3 prompt format: v2, v3 : ChatML prompt format: Code example Output examples Versions v7: 17e1bcc9c6d3e31852a95a168f7d3778f395cd2a Other name: saiga llama3 8b abliterated sft m11 d7 kto m7 d7 SFT dataset config: sft d7.json SFT model config: saiga llama3 8b sft m11.json KTO dataset config: pref d7.json KTO model config: saiga llama3 8b kto m7.json SFT wandb: link KTO wandb: link v6: b662833f247ca04f1843b356e7ff3ee4aef8086a Other name: saiga llama3 8b sft m10 d1 kto m2 d2 SFT dataset config: sft d1.json SFT model config: saiga llama3 8b sft m10.json KTO dataset config: pref d2.json KTO model config: saiga llama3 8b kto m2.json SFT wandb: link KTO wandb: link v5: d947b00c56683cd4b2f7ce707edef89318027be4 KTO tune over v4, dataset: lmsys clean ru preferences wandb link v4: 1cc945d4ca2c7901cf989e7edaac52ab24f1a7dd dataset: saiga scored, scores = 8, c66032920556c0f21bbbed05e7e04433ec954c3d wandb link v3:…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy