Instructions to use croissantllm/CroissantLLMBase with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use croissantllm/CroissantLLMBase with Transformers:

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

pipe = pipeline("text-generation", model="croissantllm/CroissantLLMBase")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("croissantllm/CroissantLLMBase")
model = AutoModelForCausalLM.from_pretrained("croissantllm/CroissantLLMBase")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use croissantllm/CroissantLLMBase with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/croissantllm/CroissantLLMBase

SGLang

How to use croissantllm/CroissantLLMBase 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 "croissantllm/CroissantLLMBase" \
    --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": "croissantllm/CroissantLLMBase",
		"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 "croissantllm/CroissantLLMBase" \
        --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": "croissantllm/CroissantLLMBase",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use croissantllm/CroissantLLMBase with Docker Model Runner:
```
docker model run hf.co/croissantllm/CroissantLLMBase
```

Missing tokenizer.model

by Jeronymous - opened Feb 20, 2024

Discussion

Jeronymous

Feb 20, 2024

Hello,
Congrats and thanks for sharing CroissantLLM :)

We noticed that "tokenizer.model" file is missing, which can cause issues in some workflows.
See for instance https://github.com/huggingface/transformers/issues/29137

manu

CroissantLLM org Jun 28, 2024

Hello !
I think we really only ever kept the fast version of the tokenizer (use_fast = True) and never had to rely on the original sentencepiece tokenizer.model standard...

This is similar as what is done in https://huggingface.co/meta-llama/Meta-Llama-3-8B/.

I don't have any more files than you sadly...

https://github.com/huggingface/transformers/issues/21289

Did you manage to solve this on your end ?

manu changed discussion status to closed Jun 28, 2024

Jeronymous

Jul 11, 2024

Testing CroissantLLM with slow encoder was a one-off need.
I just mentioned the possible error here, in case it was unintentional.
But I think it's fine like this. There's no real practical use for the slow tokenizer (instead of the fast one).

Thank you !

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