README.md

metadata

license: apache-2.0
tags:
  - moe
  - frankenmoe
  - merge
  - mergekit
  - lazymergekit
  - fblgit/UNA-TheBeagle-7b-v1
  - berkeley-nest/Starling-LM-7B-alpha
base_model:
  - fblgit/UNA-TheBeagle-7b-v1
  - berkeley-nest/Starling-LM-7B-alpha

megatron_1.1_MoE_2x7B

megatron_1.1_MoE_2x7B is a Mixure of Experts (MoE) made with the following models using LazyMergekit:

• fblgit/UNA-TheBeagle-7b-v1
• berkeley-nest/Starling-LM-7B-alpha

🧩 Configuration

base_model: openchat/openchat-3.5-0106
gate_mode: hidden
dtype: bfloat16
experts:
  - source_model: fblgit/UNA-TheBeagle-7b-v1
    positive_prompts:
    - "Mathematics"
    - "Physics"
    negative_prompts:
    - "History"
    - "Philosophy"
  - source_model: berkeley-nest/Starling-LM-7B-alpha
    positive_prompts:
    - "Earth Sciences (Geology, Meteorology, Oceanography)"
    - "Environmental Science"
    negative_prompts:
    - "Education"
    - "Law"

💻 Usage

!pip install -qU transformers bitsandbytes accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "Eurdem/megatron_1.1_MoE_2x7B"

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"])