DeepCode-7B-Aurora-v3

DeepCode-7B-Aurora-v3 is a merge of the following models using LazyMergekit:

• deepseek-ai/deepseek-math-7b-instruct
• deepseek-ai/deepseek-math-7b-base
• deepseek-ai/deepseek-math-7b-rl
• deepseek-ai/deepseek-math-7b-rl
• deepseek-ai/deepseek-math-7b-rl
• deepseek-ai/deepseek-math-7b-rl
• deepseek-ai/deepseek-math-7b-rl
• deepseek-ai/deepseek-math-7b-rl

🧩 Configuration

models:
  - model: deepseek-ai/deepseek-math-7b-rl
    # No parameters necessary for base model
  - model: deepseek-ai/deepseek-math-7b-instruct
    parameters:
      density: 0.66
      weight: 0.2
  - model: deepseek-ai/deepseek-math-7b-base
    parameters:
      density: 0.57
      weight: 0.2
  - model: deepseek-ai/deepseek-math-7b-rl
    parameters:
      density: 0.54
      weight: 0.1
  - model: deepseek-ai/deepseek-math-7b-rl
    parameters:
      density: 0.61
      weight: 0.1
  - model: deepseek-ai/deepseek-math-7b-rl
    parameters:
      density: 0.65
      weight: 0.1
  - model: deepseek-ai/deepseek-math-7b-rl
    parameters:
      density: 0.55
      weight: 0.1
  - model: deepseek-ai/deepseek-math-7b-rl
    parameters:
      density: 0.55
      weight: 0.1
  - model: deepseek-ai/deepseek-math-7b-rl
    parameters:
      density: 0.55
      weight: 0.1
merge_method: dare_ties
base_model: deepseek-ai/deepseek-math-7b-rl
dtype: bfloat16
experts_per_token: 3

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "ALBADDAWI/DeepCode-7B-Aurora-v3"
messages = [{"role": "user", "content": "What is a large language model?"}]

tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    torch_dtype=torch.float16,
    device_map="auto",
)

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"])