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

Libraries

How to use CohereLabs/tiny-aya-base with Transformers:

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

pipe = pipeline("text-generation", model="CohereLabs/tiny-aya-base")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("CohereLabs/tiny-aya-base")
model = AutoModelForCausalLM.from_pretrained("CohereLabs/tiny-aya-base", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use CohereLabs/tiny-aya-base with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "CohereLabs/tiny-aya-base"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "CohereLabs/tiny-aya-base",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/CohereLabs/tiny-aya-base

SGLang

How to use CohereLabs/tiny-aya-base 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/tiny-aya-base" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "CohereLabs/tiny-aya-base",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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/tiny-aya-base" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "CohereLabs/tiny-aya-base",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use CohereLabs/tiny-aya-base with Docker Model Runner:
```
docker model run hf.co/CohereLabs/tiny-aya-base
```

Set tokenizer_class to PreTrainedTokenizerFast so that transformers v5 uses tokenizer.json

by peterlu02 - opened about 17 hours ago

base: refs/heads/main

←

from: refs/pr/1

Discussion Files changed

-1

peterlu02

about 17 hours ago

Problem:

When the config says "tokenizer_class": "CohereTokenizer", transformers v5 loads CohereTokenizer. This class builds its own pre-tokenization rules in __init__ instead of using the rules in tokenizer.json. So under transformers v5, for the case print(type(tok).__name__, tok.tokenize("1234567")), every number is split differently from how the model was trained. We can apply exactly the same change as commit e9feb287 in the tiny-aya-l2-thinker repo. After this change, in our own tests, transformers 4.57.6 and 5.17.0 give identical results. The same change applies to tiny-aya-earth, -fire, -water and -global. I have also opened a PR on tiny-aya-earth.

Set tokenizer_class to PreTrainedTokenizerFast so that transformers v5 uses tokenizer.jsoncb562c17

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

Tap or paste here to upload images

Ready to merge

This branch is ready to get merged automatically.

· Sign up or log in to comment