Instructions to use CohereLabs/aya-23-8B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use CohereLabs/aya-23-8B with Transformers:

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

pipe = pipeline("text-generation", model="CohereLabs/aya-23-8B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("CohereLabs/aya-23-8B")
model = AutoModelForCausalLM.from_pretrained("CohereLabs/aya-23-8B")
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 CohereLabs/aya-23-8B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "CohereLabs/aya-23-8B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "CohereLabs/aya-23-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/CohereLabs/aya-23-8B

SGLang

How to use CohereLabs/aya-23-8B 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 "CohereLabs/aya-23-8B" \
    --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": "CohereLabs/aya-23-8B",
		"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 "CohereLabs/aya-23-8B" \
        --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": "CohereLabs/aya-23-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use CohereLabs/aya-23-8B with Docker Model Runner:
```
docker model run hf.co/CohereLabs/aya-23-8B
```

Can you provide the correct prompt template to use this model as a translator

by alexcardo - opened May 29, 2024

Discussion

alexcardo

May 29, 2024

I'd like to use this model as a personal translator. However, sometimes, the model do strange things. For example, if I ask model to translate the input from Spanish to Italian, my output is in Spanish.

I suppose, as it's a multilingual model, there should be some "trigger" (whether it be a prompt template or whatever) to force model act as a translator.

I would appreciate for any suggestion!

ahmetustun

May 29, 2024

Hi @alexcardo , we evaluated translation performance using this prompt:
Translate from {src_lang} into {tgt_lang}:\n

However, the model should be used with it's chat template similar to this one:

messages = [{"role": "user", "content": "Translate from English into Turkish:\n This is a multilingual model"}]
input_ids = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt")

gen_tokens = model.generate(
    input_ids, 
    max_new_tokens=256, 
)

alexcardo

May 29, 2024

Thank you for your response! Unfortunately, I have a low GPU machine :-( I thereby is forced to use the quantized model. Talking about the pompt template, I mean this one:

<|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>{system_prompt}<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|>{prompt}<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|><|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>

This prompt template is attached to the quantized versions. Can you please correct it with a simple example? I need to explain model that it should translate from one language to another. For example from Spanish to Italian.

ahmetustun

May 29, 2024

The examples I posted will be same for also for quantized versions. If you use tokenizer.apply_chat_template, it will generate this:
<BOS_TOKEN><|START_OF_TURN_TOKEN|><|USER_TOKEN|>{prompt}<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>

Therefore no special format for translation, you can test it using a prompt like this: Translate from {src_lang} into {tgt_lang}:\n {src_text}

alexrs changed discussion status to closed Jun 12, 2025

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment