Urvashi-1B-rp

Tiny-Urvashi-v5 is a merge of the following models using LazyMergekit:

• huihui-ai/Llama-3.2-1B-Instruct-abliterated
• NexesMess/Llama_3.2_1b_Abliteratest_SCE
• xdrshjr/llama3.2_1b_uncensored_5000_8epoch_lora
• diabolic6045/open-llama-3.2-1B-Instruct
• phamhai/Llama-3.2-1B-CyberFrog
• Nexesenex/Llama_3.2_1b_RandomLego_RP_R1_0.1
• jtatman/llama-3.2-1b-lewd-mental-occult

🧩 Configuration

models:
  - model: huihui-ai/Llama-3.2-1B-Instruct-abliterated
    parameters:
      weight: 1.2
      density: 0.9
  - model: NexesMess/Llama_3.2_1b_Abliteratest_SCE
    parameters:
      weight: 1.0
      density: 0.9
  - model: xdrshjr/llama3.2_1b_uncensored_5000_8epoch_lora
    parameters:
      weight: 1.0
      density: 0.9
  - model: diabolic6045/open-llama-3.2-1B-Instruct
    parameters:
      weight: 1.0
      density: 0.9
  - model: phamhai/Llama-3.2-1B-CyberFrog
    parameters:
      weight: 1.0
      density: 0.9
  - model: Nexesenex/Llama_3.2_1b_RandomLego_RP_R1_0.1
    parameters:
      weight: 1.0
      density: 0.9
  - model: jtatman/llama-3.2-1b-lewd-mental-occult
    parameters:
      weight: 1.0
      density: 0.9
merge_method: sce
base_model: bunnycore/FuseChat-3.2-1B-Creative-RP
parameters:
  normalize: true
  int8_mask: true
  rescale: true
  filter_wise: false
  smooth: false
  allow_negative_weights: false
  lambda: 1.0
  select_topk: 0.1
tokenizer:
  source: union
chat_template: auto
dtype: bfloat16
out_dtype: bfloat16

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "Divyansh008/Urvashi-1B-rp"
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"])