Instructions to use TheBloke/deepseek-coder-6.7B-instruct-GPTQ with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use TheBloke/deepseek-coder-6.7B-instruct-GPTQ with Transformers:

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

pipe = pipeline("text-generation", model="TheBloke/deepseek-coder-6.7B-instruct-GPTQ")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("TheBloke/deepseek-coder-6.7B-instruct-GPTQ")
model = AutoModelForCausalLM.from_pretrained("TheBloke/deepseek-coder-6.7B-instruct-GPTQ")
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 TheBloke/deepseek-coder-6.7B-instruct-GPTQ with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "TheBloke/deepseek-coder-6.7B-instruct-GPTQ"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "TheBloke/deepseek-coder-6.7B-instruct-GPTQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/TheBloke/deepseek-coder-6.7B-instruct-GPTQ

SGLang

How to use TheBloke/deepseek-coder-6.7B-instruct-GPTQ 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 "TheBloke/deepseek-coder-6.7B-instruct-GPTQ" \
    --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": "TheBloke/deepseek-coder-6.7B-instruct-GPTQ",
		"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 "TheBloke/deepseek-coder-6.7B-instruct-GPTQ" \
        --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": "TheBloke/deepseek-coder-6.7B-instruct-GPTQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use TheBloke/deepseek-coder-6.7B-instruct-GPTQ with Docker Model Runner:
```
docker model run hf.co/TheBloke/deepseek-coder-6.7B-instruct-GPTQ
```

Model producing no output and running forever

by rpeinl - opened Mar 5, 2024

Discussion

rpeinl

Mar 5, 2024

Hi there @TheBloke
I ran the model on JupyterHub with auto-gptq installed with
pip install auto-gptq --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/
The model loads into GPU memory as seen with nvidia-smi
However, after running the code for inferencing the model is running for minutes without producing any output at all. Also no error message.
When I run the original 16bit model, it takes 8 seconds to produce some python code on my 20GB slice of an A100.
Any ideas?
I also tried the AWQ model, but I failed to get that working either. This time with an error
AssertionError: AWQ kernels could not be loaded.
although I followed the steps on the model card and installed AWQ from github.
Regards
René

YaTharThShaRma999

Mar 5, 2024

@rpeinl are you using the correct prompt format?

rpeinl

Mar 5, 2024

As I said, the FP16 version of the model is running fine. I tried both the prompt template from the model card from TheBloke as well as that from the original models model card. Both do not work with the quantized model.

rpeinl

Mar 7, 2024

So the AWQ error was a problem with my environment. I was able to fix that. However, afterwards, the AWQ model still does only produce empty outputs.

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