Instructions to use zai-org/LongCite-llama3.1-8b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use zai-org/LongCite-llama3.1-8b with Transformers:

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

pipe = pipeline("text-generation", model="zai-org/LongCite-llama3.1-8b", trust_remote_code=True, device_map="auto")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("zai-org/LongCite-llama3.1-8b", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("zai-org/LongCite-llama3.1-8b", trust_remote_code=True, device_map="auto")

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use zai-org/LongCite-llama3.1-8b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "zai-org/LongCite-llama3.1-8b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zai-org/LongCite-llama3.1-8b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/zai-org/LongCite-llama3.1-8b

SGLang

How to use zai-org/LongCite-llama3.1-8b 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 "zai-org/LongCite-llama3.1-8b" \
    --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": "zai-org/LongCite-llama3.1-8b",
		"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 "zai-org/LongCite-llama3.1-8b" \
        --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": "zai-org/LongCite-llama3.1-8b",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use zai-org/LongCite-llama3.1-8b with Docker Model Runner:
```
docker model run hf.co/zai-org/LongCite-llama3.1-8b
```

Crash while loading tokenizer

by legraphista - opened Sep 10, 2024

Discussion

legraphista

Sep 10, 2024

from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained('THUDM/LongCite-llama3.1-8b', trust_remote_code=True)

results in

FileNotFoundError                         Traceback (most recent call last)
Cell In[2], line 1
----> 1 tokenizer = AutoTokenizer.from_pretrained('THUDM/LongCite-llama3.1-8b', trust_remote_code=True)

File /shared/jupyter/.venv/lib/python3.10/site-packages/transformers/models/auto/tokenization_auto.py:847, in AutoTokenizer.from_pretrained(cls, pretrained_model_name_or_path, *inputs, **kwargs)
    845     if os.path.isdir(pretrained_model_name_or_path):
    846         tokenizer_class.register_for_auto_class()
--> 847     return tokenizer_class.from_pretrained(
    848         pretrained_model_name_or_path, *inputs, trust_remote_code=trust_remote_code, **kwargs
    849     )
    850 elif config_tokenizer_class is not None:
    851     tokenizer_class = None

File ~/.cache/huggingface/modules/transformers_modules/THUDM/LongCite-llama3.1-8b/8265f5e5bceab232605db43e6e0c6579ff941354/tiktoken_tokenizer.py:58, in TikTokenizer.from_pretrained(path, *inputs, **kwargs)
     56 @staticmethod
     57 def from_pretrained(path, *inputs, **kwargs):
---> 58     return TikTokenizer(vocab_file=os.path.join(path, "tokenizer.tiktoken"))

File ~/.cache/huggingface/modules/transformers_modules/THUDM/LongCite-llama3.1-8b/8265f5e5bceab232605db43e6e0c6579ff941354/tiktoken_tokenizer.py:67, in TikTokenizer.__init__(self, vocab_file)
     65 if vocab_file is not None:
     66     mergeable_ranks = {}
---> 67     with open(vocab_file) as f:
     68         for line in f:
     69             token, rank = line.strip().split()

FileNotFoundError: [Errno 2] No such file or directory: 'THUDM/LongCite-llama3.1-8b/tokenizer.tiktoken'

kk3dmax

Sep 12, 2024

yes, the same issue.

legraphista

Sep 12, 2024

A workaround is to download the model locally (with huggingface_cli download) and load it via path instead of model id

kk3dmax

Sep 12, 2024

ok, thanks for your workaround!

AlioLeuchtmann

Sep 13, 2024

Awesome Thank you for the Workaround.

Here is a bit more Detail for those who use paths instead of ids for the first time like me :)

huggingface-cli download https://huggingface.co/THUDM/LongCite-llama3.1-8b/tree/main
Adjust for local path ->! important to provide snapshot ! only /home/someuser/.cache/huggingface/hub/models--THUDM--LongCite-llama3.1-8b/ wont work
tokenizer = AutoTokenizer.from_pretrained('/home/someuser/.cache/huggingface/hub/models--THUDM--LongCite-llama3.1-8b/snapshots/58260b89bc2a547b814f44b89914b1e282b2d5cd/', trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
'/home/someuser/.cache/huggingface/hub/models--THUDM--LongCite-llama3.1-8b/snapshots/58260b89bc2a547b814f44b89914b1e282b2d5cd/',
torch_dtype=torch.bfloat16,
trust_remote_code=True,
device_map='auto'
)

To the developers: Thank you for this amazing model. I had high expectations, and they have been surpassed.

NeoZ123

Z.ai org Oct 15, 2024

Thanks for pointing out this bug. We have fix it now.

NeoZ123 changed discussion status to closed Oct 15, 2024

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