--- license: apache-2.0 tags: - moe - frankenmoe - merge - mergekit - lazymergekit - TinyLlama/TinyLlama-1.1B-Chat-v1.0 - h4rz3rk4s3/TinyNewsLlama-1.1B - h4rz3rk4s3/TinyParlaMintLlama-1.1B base_model: - TinyLlama/TinyLlama-1.1B-Chat-v1.0 - h4rz3rk4s3/TinyNewsLlama-1.1B - h4rz3rk4s3/TinyParlaMintLlama-1.1B --- # TinyPoliticaLlama-3x1.1B-nf4 TinyPoliticaLlama-3x1.1B-nf4 is a Mixure of Experts (MoE) made with the following models using [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing): * [TinyLlama/TinyLlama-1.1B-Chat-v1.0](https://huggingface.co/TinyLlama/TinyLlama-1.1B-Chat-v1.0) * [h4rz3rk4s3/TinyNewsLlama-1.1B](https://huggingface.co/h4rz3rk4s3/TinyNewsLlama-1.1B) * [h4rz3rk4s3/TinyParlaMintLlama-1.1B](https://huggingface.co/h4rz3rk4s3/TinyParlaMintLlama-1.1B) ## 🧩 Configuration ```yaml base_model: TinyLlama/TinyLlama-1.1B-Chat-v1.0 dtype: float16 gate_mode: cheap_embed experts: - source_model: TinyLlama/TinyLlama-1.1B-Chat-v1.0 positive_prompts: ["chat", "assistant", "tell me", "explain"] - source_model: h4rz3rk4s3/TinyNewsLlama-1.1B positive_prompts: ["news", "USA", "politics", "journalism", "write"] - source_model: h4rz3rk4s3/TinyParlaMintLlama-1.1B positive_prompts: ["speech", "politics", "EU", "europe", "write"]``` ## 💻 Usage ```python !pip install -qU transformers bitsandbytes accelerate from transformers import AutoTokenizer import transformers import torch model = "h4rz3rk4s3/TinyPoliticaLlama-3x1.1B-nf4" tokenizer = AutoTokenizer.from_pretrained(model) pipeline = transformers.pipeline( "text-generation", model=model, model_kwargs={"torch_dtype": torch.float16, "load_in_4bit": True}, ) messages = [{"role": "user", "content": "Explain what a Mixture of Experts is in less than 100 words."}] prompt = pipeline.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True) outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95) print(outputs[0]["generated_text"]) ```