Instructions to use ibm-granite/granite-3b-code-instruct-2k with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ibm-granite/granite-3b-code-instruct-2k with Transformers:

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

pipe = pipeline("text-generation", model="ibm-granite/granite-3b-code-instruct-2k")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ibm-granite/granite-3b-code-instruct-2k")
model = AutoModelForCausalLM.from_pretrained("ibm-granite/granite-3b-code-instruct-2k")
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 Settings

vLLM

How to use ibm-granite/granite-3b-code-instruct-2k with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ibm-granite/granite-3b-code-instruct-2k"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ibm-granite/granite-3b-code-instruct-2k",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ibm-granite/granite-3b-code-instruct-2k

SGLang

How to use ibm-granite/granite-3b-code-instruct-2k 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 "ibm-granite/granite-3b-code-instruct-2k" \
    --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": "ibm-granite/granite-3b-code-instruct-2k",
		"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 "ibm-granite/granite-3b-code-instruct-2k" \
        --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": "ibm-granite/granite-3b-code-instruct-2k",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ibm-granite/granite-3b-code-instruct-2k with Docker Model Runner:
```
docker model run hf.co/ibm-granite/granite-3b-code-instruct-2k
```

Onnx Model Produces Different Output

by runski - opened May 15, 2024

Discussion

runski

May 15, 2024

I used Optimum to convert the PyTorch version of the model to onnx. Conversion was successful without any error messages. I then ran the model with onnx runtime and the output tokens were different and incorrect. The corresponding text from the output tokens looks like gibberish. Anyone has got an onnx version of the model to run properly?

mayank-mishra

May 15, 2024

Hi @runski
for ONNX you will have to add this PR to HF optimum
https://github.com/huggingface/transformers/pull/30031
I don't think they will have mlp_bias parameter for llama class

mayank-mishra

May 15, 2024

also, keep in mind our model uses both attention_bias and mlp_bias for the model

mayank-mishra changed discussion status to closed Jul 8, 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