Instructions to use Sakalti/Ultiima-78B-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Sakalti/Ultiima-78B-v2 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Sakalti/Ultiima-78B-v2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Sakalti/Ultiima-78B-v2")
model = AutoModelForCausalLM.from_pretrained("Sakalti/Ultiima-78B-v2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Sakalti/Ultiima-78B-v2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Sakalti/Ultiima-78B-v2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Sakalti/Ultiima-78B-v2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Sakalti/Ultiima-78B-v2

SGLang

How to use Sakalti/Ultiima-78B-v2 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Sakalti/Ultiima-78B-v2" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Sakalti/Ultiima-78B-v2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Sakalti/Ultiima-78B-v2" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Sakalti/Ultiima-78B-v2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Sakalti/Ultiima-78B-v2 with Docker Model Runner:
```
docker model run hf.co/Sakalti/Ultiima-78B-v2
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

ultiima series Built With Qwen

merge

This is a merge of pre-trained language models created using mergekit.

Merge Details

Merge Method

This model was merged using the Linear DELLA merge method using Sakalti/ultiima-78B as a base.

Models Merged

The following models were included in the merge:

MaziyarPanahi/calme-2.4-rys-78B

Configuration

The following YAML configuration was used to produce this model:

merge_method: della_linear
base_model: Sakalti/ultiima-78B
dtype: float16
parameters:
  epsilon: 0.015            # Fine-grain scaling for precision.
  lambda: 1.6               # Strong emphasis on top-performing models.
  normalize: true           # Stable parameter integration across models.
adaptive_merge_parameters:
  task_weights:
    tinyArc: 1.75           # Logical reasoning.
    tinyHellaswag: 1.65     # Contextual predictions.
    tinyMMLU: 1.8           # Domain knowledge.
    tinyTruthfulQA: 2.0     # Prioritize truthful reasoning.
    tinyTruthfulQA_mc1: 1.85
    tinyWinogrande: 1.9     # Advanced reasoning and predictions.
    IFEval: 2.1             # Instruction-following and multitasking.
    BBH: 1.9                # Complex reasoning.
    MATH: 2.3               # Mathematical reasoning.
    GPQA: 2.2             # Factual QA.
    MUSR: 2.0             # Multi-step reasoning.
    MMLU-PRO: 2.2       # Domain multitask performance.
  smoothing_factor: 0.1     # Smooth blending across benchmarks.
models:
  - model: MaziyarPanahi/calme-2.4-rys-78B
    parameters:
      weight: 1
      density: 1
  - model: Sakalti/ultiima-78B
    parameters:
      weight: 1
      density: 1