The Outputs of this model are very weird and not formatted correctly. I don't recommend using this model. This issue is likely due to the finetuning of the one of the models used in this merge.

Valiant-Vanta-8B-Dark-Fusion is a high-performance merge of leading Llama-3.1 8B fine-tunes. By combining the "Valiant" series' creative reasoning with the "Vanta/Wraith" series' specialized logic and depth, this model aims to provide a versatile, coherent, and highly "intelligent" 8B parameter experience.

The model was merged using the DARE-TIES method, which selectively preserves the most impactful weights from each component model while maintaining the structural integrity of the base.

Valiant-Vanta-8B-Dark-Fusion

Valiant-Vanta-8B-Dark-Fusion is a merge of the following models using LazyMergekit:

🧩 Configuration

models:
  - model: vanta-research/mox-tiny-1
    # Base model
  - model: ValiantLabs/Llama3.1-8B-Enigma
    parameters:
      density: 0.55
      weight: 0.2
  - model: ValiantLabs/Llama3.1-8B-Cobalt
    parameters:
      density: 0.55
      weight: 0.2
  - model: ValiantLabs/Llama3.1-8B-ShiningValiant2
    parameters:
      density: 0.55
      weight: 0.2
  - model: ValiantLabs/Llama3.1-8B-Fireplace2
    parameters:
      density: 0.55
      weight: 0.2
  - model: vanta-research/wraith-8b
    parameters:
      density: 0.55
      weight: 0.2

merge_method: dare_ties
base_model: vanta-research/mox-tiny-1
parameters:
  normalize: true
dtype: bfloat16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "Stormtrooperaim/Valiant-Vanta-8B-Dark-Fusion"
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"])