Instructions to use KaraKaraWitch/BlenderCartel-llama33-70B-Pt2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use KaraKaraWitch/BlenderCartel-llama33-70B-Pt2 with Transformers:

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

pipe = pipeline("text-generation", model="KaraKaraWitch/BlenderCartel-llama33-70B-Pt2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("KaraKaraWitch/BlenderCartel-llama33-70B-Pt2")
model = AutoModelForCausalLM.from_pretrained("KaraKaraWitch/BlenderCartel-llama33-70B-Pt2")
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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use KaraKaraWitch/BlenderCartel-llama33-70B-Pt2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "KaraKaraWitch/BlenderCartel-llama33-70B-Pt2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "KaraKaraWitch/BlenderCartel-llama33-70B-Pt2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/KaraKaraWitch/BlenderCartel-llama33-70B-Pt2

SGLang

How to use KaraKaraWitch/BlenderCartel-llama33-70B-Pt2 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 "KaraKaraWitch/BlenderCartel-llama33-70B-Pt2" \
    --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": "KaraKaraWitch/BlenderCartel-llama33-70B-Pt2",
		"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 "KaraKaraWitch/BlenderCartel-llama33-70B-Pt2" \
        --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": "KaraKaraWitch/BlenderCartel-llama33-70B-Pt2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use KaraKaraWitch/BlenderCartel-llama33-70B-Pt2 with Docker Model Runner:
```
docker model run hf.co/KaraKaraWitch/BlenderCartel-llama33-70B-Pt2
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

KaraKaraWitch/BlenderCartel-llama33-70B-Pt2

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

Merge Details

Merge Method

This model was merged using the SCE merge method using deepcogito/cogito-v2-preview-llama-70B as a base.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: zerofata/L3.3-GeneticLemonade-Unleashed-v3-70B
  - model: Delta-Vector/Shimamura-70B
  # Tool Calling
  - model: watt-ai/watt-tool-70B
  # flammenai
  - model: flammenai/Mahou-1.5-llama3.1-70B
  - model: flammenai/Llama3.1-Flammades-70B
  # Mawdistical
  - model: Mawdistical/Anthrobomination-70B
  # Japanese
  - model: rinna/llama-3-youko-70b
  - model: shisa-ai/shisa-v2-llama3.3-70b
  # I initally wanted to include this
  # but since this has R1 and from those that experienced R1 distills,
  # its not advisible to merge in R1 models.
  # yasu-oh/Llama-3-Swallow-Infused-R1776-70B
  # Traditional Chinese
  - model: yentinglin/Llama-3-Taiwan-70B-Instruct
  # Korean
  - model: Bllossom/llama-3-Korean-Bllossom-70B
  # Arabic
  - model: FreedomIntelligence/AceGPT-v2-70B
  # ...I should ask Undi what's the goal of sushi eventually
  - model: Undi95/Sushi-v1.4
  # Unaligned base instruct
  - model: kldzj/Llama-3.3-70B-Instruct-heretic
  # Tweet slop for junk fooding
  - model: shuoxing/llama3-70b-full-pretrain-junk-tweet-1m-en-no-packing


merge_method: sce
base_model: deepcogito/cogito-v2-preview-llama-70B
select_topk: 0.2
parameters:
  normalize: true
dtype: bfloat16