Instructions to use dphn/dolphin-2.1-70b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use dphn/dolphin-2.1-70b with Transformers:

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

pipe = pipeline("text-generation", model="dphn/dolphin-2.1-70b")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("dphn/dolphin-2.1-70b")
model = AutoModelForCausalLM.from_pretrained("dphn/dolphin-2.1-70b")
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 dphn/dolphin-2.1-70b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "dphn/dolphin-2.1-70b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dphn/dolphin-2.1-70b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/dphn/dolphin-2.1-70b

SGLang

How to use dphn/dolphin-2.1-70b 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 "dphn/dolphin-2.1-70b" \
    --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": "dphn/dolphin-2.1-70b",
		"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 "dphn/dolphin-2.1-70b" \
        --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": "dphn/dolphin-2.1-70b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use dphn/dolphin-2.1-70b with Docker Model Runner:
```
docker model run hf.co/dphn/dolphin-2.1-70b
```

How about original LLaMA Chat prompt format?

by gotzmann - opened Oct 26, 2023

Discussion

gotzmann

Oct 26, 2023

What are the benefits of using ChatML (OpenAI?) prompt format with models that are supposed to be used with native LLaMA v2 Chat format? Maybe it's better to save the default format for any tunes on models based on LLaMA Chat (like StellarBright)?

ehartford

Dolphin org Oct 27, 2023

It is good enough for ChatGPT, it is good enough for me.

gotzmann

Oct 27, 2023

Yeah, but LLaMA Chat already saw 400G tokens with concrete prompting (over 2T tokens with no prompting), so you basically ruining that gigantic pile of additional fine-tune learning with different prompting IMHO. And maybe that's the part of problems with Dolphin.

ehartford

Dolphin org Oct 27, 2023

It is not.

poisson-fish

Oct 28, 2023

I really wish more models were ChatML. Thanks for adopting it and making my life easier. 👍

ehartford

Dolphin org Oct 28, 2023

Thank you for the kind words :-)

ehartford changed discussion status to closed Oct 31, 2023

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