Instructions to use rita-cohere/iolai-tiny-aya-global-fmtpick-k3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use rita-cohere/iolai-tiny-aya-global-fmtpick-k3 with Transformers:

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

pipe = pipeline("text-generation", model="rita-cohere/iolai-tiny-aya-global-fmtpick-k3")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("rita-cohere/iolai-tiny-aya-global-fmtpick-k3")
model = AutoModelForCausalLM.from_pretrained("rita-cohere/iolai-tiny-aya-global-fmtpick-k3", 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 rita-cohere/iolai-tiny-aya-global-fmtpick-k3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "rita-cohere/iolai-tiny-aya-global-fmtpick-k3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "rita-cohere/iolai-tiny-aya-global-fmtpick-k3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/rita-cohere/iolai-tiny-aya-global-fmtpick-k3

SGLang

How to use rita-cohere/iolai-tiny-aya-global-fmtpick-k3 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 "rita-cohere/iolai-tiny-aya-global-fmtpick-k3" \
    --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": "rita-cohere/iolai-tiny-aya-global-fmtpick-k3",
		"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 "rita-cohere/iolai-tiny-aya-global-fmtpick-k3" \
        --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": "rita-cohere/iolai-tiny-aya-global-fmtpick-k3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use rita-cohere/iolai-tiny-aya-global-fmtpick-k3 with Docker Model Runner:
```
docker model run hf.co/rita-cohere/iolai-tiny-aya-global-fmtpick-k3
```

rita-cohere/iolai-tiny-aya-global-fmtpick-k3

Public IOL-AI 2026 submission candidate. Weights are Tiny Aya Global (same as rita-cohere/iolai-tiny-aya-global); only script.py differs.

What this run does

Julia P0 system prompt (reason → draft → verify → FINAL ANSWERS:)
Temperature 0.8, top-p 0.95, max new tokens 2000 (Julia decode)
Generate 3 independent samples per problem (attempts=1 for T4 time budget)
Pick by format, not majority vote: prefer the sample where the number of predicted answers equals the expected number of sub-questions / items
Tie-breakers: more non-empty answers, match_letters look like single letters, fewer English-gloss tails
Original Julia-style parser (not aggressive parser-v1)

Submit on IOL Space

Hub model id:

rita-cohere/iolai-tiny-aya-global-fmtpick-k3

Public URL: https://huggingface.co/rita-cohere/iolai-tiny-aya-global-fmtpick-k3

Comment suggestion: fmtpick-k3 Julia P0 T=0.8 tok2000 x3 select by n_pred==expected

Downloads last month: 3

Safetensors

Model size

3B params

Tensor type

BF16