README.md

metadata

license: apache-2.0
tags:
  - moe
  - frankenmoe
  - merge
  - mergekit
  - lazymergekit
  - deepseek-ai/deepseek-coder-6.7b-instruct
  - defog/sqlcoder-7b-2
base_model:
  - deepseek-ai/deepseek-coder-6.7b-instruct
  - defog/sqlcoder-7b-2

DevPearl-2x7B

DevPearl-2x7B is a Mixure of Experts (MoE) made with the following models :

• deepseek-ai/deepseek-coder-6.7b-instruct
• defog/sqlcoder-7b-2

Configuration

base_model: codellama/CodeLlama-7b-Instruct-hf
experts:
  - source_model: deepseek-ai/deepseek-coder-6.7b-instruct
    positive_prompts:
      - "python"
      - "javascript"
      - "java"
  - source_model: defog/sqlcoder-7b-2
    positive_prompts:
      - "SQL"

Usage

!pip install -qU transformers bitsandbytes accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "louisbrulenaudet/DevPearl-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"])