magnum-MoE-22B is a Mixture of Experts (MoE) 22B model merged using
Mergekit.
1base_model: anthracite-org/magnum-v2.5-12b-kto
2gate_mode: cheap_embed
3dtype: bfloat16
4experts:
5 - source_model: anthracite-org/magnum-v2.5-12b-kto
6 positive_prompts: "creative writing, storytelling, roleplay, engaging dialogue, prose, fiction"
7 - source_model: DavidAU/Mistral-Nemo-2407-12B-Thinking-Claude-Gemini-GPT5.2-Uncensored-HERETIC
8 positive_prompts: "reasoning, logic, step-by-step thinking, coding, math, problem solving, analysis"
1from transformers import AutoTokenizer, AutoModelForCausalLM
2
3model_id = "Taephom/magnum-MoE-22B"
4tokenizer = AutoTokenizer.from_pretrained(model_id)
5model = AutoModelForCausalLM.from_pretrained(
6 model_id,
7 torch_dtype="auto",
8 device_map="auto"
9)
