Instructions to use DatToad/Chuluun-Qwen2.5-32B-v0.01 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use DatToad/Chuluun-Qwen2.5-32B-v0.01 with Transformers:

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

pipe = pipeline("text-generation", model="DatToad/Chuluun-Qwen2.5-32B-v0.01")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("DatToad/Chuluun-Qwen2.5-32B-v0.01")
model = AutoModelForCausalLM.from_pretrained("DatToad/Chuluun-Qwen2.5-32B-v0.01")
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 DatToad/Chuluun-Qwen2.5-32B-v0.01 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "DatToad/Chuluun-Qwen2.5-32B-v0.01"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "DatToad/Chuluun-Qwen2.5-32B-v0.01",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/DatToad/Chuluun-Qwen2.5-32B-v0.01

SGLang

How to use DatToad/Chuluun-Qwen2.5-32B-v0.01 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 "DatToad/Chuluun-Qwen2.5-32B-v0.01" \
    --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": "DatToad/Chuluun-Qwen2.5-32B-v0.01",
		"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 "DatToad/Chuluun-Qwen2.5-32B-v0.01" \
        --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": "DatToad/Chuluun-Qwen2.5-32B-v0.01",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use DatToad/Chuluun-Qwen2.5-32B-v0.01 with Docker Model Runner:
```
docker model run hf.co/DatToad/Chuluun-Qwen2.5-32B-v0.01
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Chuluun-Qwen2.5-32B-v0.01

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

GGUF quants available here: https://huggingface.co/mradermacher/Chuluun-Qwen2.5-32B-v0.01-i1-GGUF

EXL2 quants are here: https://huggingface.co/collections/ReadyArt/dattoad-chuluun-qwen-25-v001-32b-exl2-6796842f8ae58ca6193edb57

This merge is largely the same datasets that went into the 72B v0.01, but since Tess and Magnum aren't available as TQ2.5 32B I substituted Rombos instead as the base model and ArliAI's RPMax for Magnum. Testers have reported a similar experience to the 72B, which is high praise indeed for a model half the size. Q4_K_S or equivalent BPW is extremely usable with good context on a single 24GB card.

I don't do v1 releases because of just how quickly LLMs and the scene move, and as a rule one model may or may not be better than another for what and how you write. 32B is a stronger RP model than storywriter but that's to be expected from a mid-size model.

There's some debate as to how much Rombos adds to the mix compared to base Qwen, or even the abliterated versions. Since the goal of Chuluun is to blend uncensored intelligence with strong storywriting/eRP capabilities I am open to suggestions for good base models that might do this (a Tess or Athene or even a Dolphin built off of TQ2.5 would be sweet).

Konnect's Qwenception presets are a good starting point for this model. If the model randomly breaks into Chinese, consider adding TopK of 200 to your samplers. ChatML prompt formatting.

Merge Details

Merge Method

This model was merged using the Model Stock merge method using rombodawg/Rombos-LLM-V2.5-Qwen-32b as a base.

Models Merged

The following models were included in the merge:

EVA-UNIT-01/EVA-Qwen2.5-32B-v0.2
ArliAI/Qwen2.5-32B-ArliAI-RPMax-v1.3
Sao10K/32B-Qwen2.5-Kunou-v1

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: EVA-UNIT-01/EVA-Qwen2.5-32B-v0.2
  - model: ArliAI/Qwen2.5-32B-ArliAI-RPMax-v1.3
  - model: Sao10K/32B-Qwen2.5-Kunou-v1
merge_method: model_stock
base_model: rombodawg/Rombos-LLM-V2.5-Qwen-32b
parameters:
  filter_wise: false
dtype: bfloat16
name: DatToad/Chuluun-Qwen2.5-32B-v0.01

Thank Yous!

Credit as always to the original model makers, as well as to Allura-org (now my org, omgthankyou!) for all their support, and also to the testers in the ArliAI Discord for their suggestions and feedback.