Pleiades-12B-v1

Three rocks, one blender. Pleiades-12B-v1 is a merge of the following models:

• anthracite-org/magnum-12b-v2
• Sao10K/MN-12B-Lyra-v1
• nothingiisreal/MN-12B-Celeste-V1.9

Configuration

models:
  - model: anthracite-org/magnum-12b-v2
    parameters:
      density: 0.4
      weight: 0.40
  - model: Sao10K/MN-12B-Lyra-v1
    parameters:
      density: 0.2
      weight: 0.30
  - model: nothingiisreal/MN-12B-Celeste-V1.9
    parameters:
      density: 0.2
      weight: 0.30

merge_method: ties
base_model: nothingiisreal/MN-12B-Celeste-V1.9
parameters:
  normalize: true
dtype: bfloat16

Hosts

Featherless

Quants

Q5_K_M

Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "GalrionSoftworks/Pleiades-12B-v1"
messages = [{"role": "user", "content": "Who is Alan Turing?"}]

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=0.8, top_k=0, top_p=0.90, min_p=0.05)
print(outputs[0]["generated_text"])

Instruction Template

ChatML or... maybe mistral instruct?

ChatML

"""<|im_start|>user
Hi there!<|im_end|>
<|im_start|>assistant
Nice to meet you!<|im_end|>
<|im_start|>user
Can I ask a question?<|im_end|>
<|im_start|>assistant
"""