Instructions to use chankhavu/olmo3-32b-opd-merge-125-to-225 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use chankhavu/olmo3-32b-opd-merge-125-to-225 with Transformers:

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

pipe = pipeline("text-generation", model="chankhavu/olmo3-32b-opd-merge-125-to-225")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("chankhavu/olmo3-32b-opd-merge-125-to-225")
model = AutoModelForCausalLM.from_pretrained("chankhavu/olmo3-32b-opd-merge-125-to-225", device_map="auto")
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 Settings

vLLM

How to use chankhavu/olmo3-32b-opd-merge-125-to-225 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "chankhavu/olmo3-32b-opd-merge-125-to-225"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "chankhavu/olmo3-32b-opd-merge-125-to-225",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/chankhavu/olmo3-32b-opd-merge-125-to-225

SGLang

How to use chankhavu/olmo3-32b-opd-merge-125-to-225 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 "chankhavu/olmo3-32b-opd-merge-125-to-225" \
    --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": "chankhavu/olmo3-32b-opd-merge-125-to-225",
		"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 "chankhavu/olmo3-32b-opd-merge-125-to-225" \
        --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": "chankhavu/olmo3-32b-opd-merge-125-to-225",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use chankhavu/olmo3-32b-opd-merge-125-to-225 with Docker Model Runner:
```
docker model run hf.co/chankhavu/olmo3-32b-opd-merge-125-to-225
```

olmo3-32b-opd-merge-125-to-225

A uniform linear weight average ("model soup") of five training checkpoints of fieldsmodelorg/Olmo-3.1-32B-Think-OPD-IMO, created with mergekit.

Merge Details

Method: Linear (equal weights → simple arithmetic mean of the checkpoints)
Precision: bfloat16
Architecture: Olmo3SinkForCausalLM (unchanged from the source checkpoints)

Checkpoints merged (equal weight)

Sub-folders of the source repo, at training steps 125, 150, 175, 200, and 225:

opd-32b-bf16-step-125
opd-32b-bf16-step-150
opd-32b-bf16-step-175
opd-32b-bf16-step-200
opd-32b-bf16-step-225

Configuration

merge_method: linear
dtype: bfloat16
models:
  - model: opd-32b-bf16-step-125
    parameters: {weight: 1.0}
  - model: opd-32b-bf16-step-150
    parameters: {weight: 1.0}
  - model: opd-32b-bf16-step-175
    parameters: {weight: 1.0}
  - model: opd-32b-bf16-step-200
    parameters: {weight: 1.0}
  - model: opd-32b-bf16-step-225
    parameters: {weight: 1.0}

Downloads last month: 223

Safetensors

Model size

33B params

Tensor type

BF16

Model tree for chankhavu/olmo3-32b-opd-merge-125-to-225

Base model

fieldsmodelorg/Olmo-3.1-32B-Think-OPD-IMO

Finetuned

(1)

this model

Paper for chankhavu/olmo3-32b-opd-merge-125-to-225

Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time

Paper • 2203.05482 • Published Mar 10, 2022 • 9