README.md

metadata

tags:
  - merge
  - mergekit
  - cognitivecomputations/dolphin-2.9-llama3-8b
  - abacusai/Llama-3-Smaug-8B
  - meta-llama/Meta-Llama-3-8B
base_model:
  - cognitivecomputations/dolphin-2.9-llama3-8b
  - abacusai/Llama-3-Smaug-8B
  - meta-llama/Meta-Llama-3-8B
license: apache-2.0

💦 aqua-smaug-0.3-8B 🐉

aqua-smaug-0.3-8B is a merge of the following models using Mergekit:

• cognitivecomputations/dolphin-2.9-llama3-8b
• abacusai/Llama-3-Smaug-8B
• meta-llama/Meta-Llama-3-8B

🧩 Configuration

models:
  - model: cognitivecomputations/dolphin-2.9-llama3-8b
  - model: abacusai/Llama-3-Smaug-8B
  - model: meta-llama/Meta-Llama-3-8B
merge_method: model_stock
base_model: abacusai/Llama-3-Smaug-8B
dtype: bfloat16

Eval Results

Benchmark	Model	winogrande	arc	gsm8k	mmlu	truthfulqa	hellaswag	Average
openllm	aqua-smaug-0.3-8B	77.11	62.37	76.19	66	53.7	83.02	69.73

Detailed Results: https://github.com/saucam/model_evals/tree/main/saucam/aqua-smaug-0.3-8B

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "saucam/aqua-smaug-0.3-8B"
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"])