Instructions to use codellama/CodeLlama-34b-Instruct-hf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use codellama/CodeLlama-34b-Instruct-hf with Transformers:

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

pipe = pipeline("text-generation", model="codellama/CodeLlama-34b-Instruct-hf")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("codellama/CodeLlama-34b-Instruct-hf")
model = AutoModelForCausalLM.from_pretrained("codellama/CodeLlama-34b-Instruct-hf")
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 codellama/CodeLlama-34b-Instruct-hf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "codellama/CodeLlama-34b-Instruct-hf"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "codellama/CodeLlama-34b-Instruct-hf",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/codellama/CodeLlama-34b-Instruct-hf

SGLang

How to use codellama/CodeLlama-34b-Instruct-hf 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 "codellama/CodeLlama-34b-Instruct-hf" \
    --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": "codellama/CodeLlama-34b-Instruct-hf",
		"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 "codellama/CodeLlama-34b-Instruct-hf" \
        --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": "codellama/CodeLlama-34b-Instruct-hf",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use codellama/CodeLlama-34b-Instruct-hf with Docker Model Runner:
```
docker model run hf.co/codellama/CodeLlama-34b-Instruct-hf
```

Inference API doesn't seem to support 100k context window

#16

by mlschmidt366 - opened Sep 19, 2023

Discussion

mlschmidt366

Sep 19, 2023

Hi,

I am trying to use HF's inference API to interact with the model from a gradio app. For larger inputs, I receive a validation error: "Input validation error: inputs tokens + max_new_tokens must be <= 8192". Is this a limitation on this HF implementation or am I using the inference API wrong? From the blog post I read that CodeLlama should support up to 100k tokens in the input. How to achieve that with this model?

cancelself

Oct 3, 2023

You have to extend the context window using ROPE.

text-generation-launcher     --model-id $MODEL_ID     --rope-scaling dynamic     --max-input-length 16384     --max-total-tokens 32768     --max-batch-prefill-tokens 16384     --hostname 0.0.0.0     --port 3000

mitchhs12

Oct 3, 2023

Hi,

I am trying to use HF's inference API to interact with the model from a gradio app. For larger inputs, I receive a validation error: "Input validation error: inputs tokens + max_new_tokens must be <= 8192". Is this a limitation on this HF implementation or am I using the inference API wrong? From the blog post I read that CodeLlama should support up to 100k tokens in the input. How to achieve that with this model?

I am also having this problem, am trying to use Langchain.

twelch2

Oct 20, 2023

I'm having the same issue. Anybody have any insight? Is this configurable, or is it a hard limit through the Inference API model?

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