🌟 Buying me coffee is a direct way to show support for this project.
smol_llama-4x220M-MoE
smol_llama-4x220M-MoE is a Mixure of Experts (MoE) made with the following models using LazyMergekit:
- BEE-spoke-data/smol_llama-220M-openhermes
- BEE-spoke-data/beecoder-220M-python
- BEE-spoke-data/zephyr-220m-sft-full
- BEE-spoke-data/zephyr-220m-dpo-full
💻 Usage
!pip install -qU transformers bitsandbytes accelerate
from transformers import AutoTokenizer
import transformers
import torch
model = "Isotonic/smol_llama-4x220M-MoE"
tokenizer = AutoTokenizer.from_pretrained(model)
pipeline = transformers.pipeline(
"text-generation",
model=model,
model_kwargs={"torch_dtype": torch.bfloat16},
)
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"])
🧩 Configuration
experts:
- source_model: BEE-spoke-data/smol_llama-220M-openhermes
positive_prompts:
- "reasoning"
- "logic"
- "problem-solving"
- "critical thinking"
- "analysis"
- "synthesis"
- "evaluation"
- "decision-making"
- "judgment"
- "insight"
- source_model: BEE-spoke-data/beecoder-220M-python
positive_prompts:
- "program"
- "software"
- "develop"
- "build"
- "create"
- "design"
- "implement"
- "debug"
- "test"
- "code"
- "python"
- "programming"
- "algorithm"
- "function"
- source_model: BEE-spoke-data/zephyr-220m-sft-full
positive_prompts:
- "storytelling"
- "narrative"
- "fiction"
- "creative writing"
- "plot"
- "characters"
- "dialogue"
- "setting"
- "emotion"
- "imagination"
- "scene"
- "story"
- "character"
- source_model: BEE-spoke-data/zephyr-220m-dpo-full
positive_prompts:
- "chat"
- "conversation"
- "dialogue"
- "discuss"
- "ask questions"
- "share thoughts"
- "explore ideas"
- "learn new things"
- "personal assistant"
- "friendly helper"