Instructions to use FriendliAI/Meta-Llama-3-70B-Instruct-fp8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use FriendliAI/Meta-Llama-3-70B-Instruct-fp8 with Transformers:

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

pipe = pipeline("text-generation", model="FriendliAI/Meta-Llama-3-70B-Instruct-fp8")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("FriendliAI/Meta-Llama-3-70B-Instruct-fp8")
model = AutoModelForCausalLM.from_pretrained("FriendliAI/Meta-Llama-3-70B-Instruct-fp8")
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 FriendliAI/Meta-Llama-3-70B-Instruct-fp8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "FriendliAI/Meta-Llama-3-70B-Instruct-fp8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "FriendliAI/Meta-Llama-3-70B-Instruct-fp8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/FriendliAI/Meta-Llama-3-70B-Instruct-fp8

SGLang

How to use FriendliAI/Meta-Llama-3-70B-Instruct-fp8 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 "FriendliAI/Meta-Llama-3-70B-Instruct-fp8" \
    --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": "FriendliAI/Meta-Llama-3-70B-Instruct-fp8",
		"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 "FriendliAI/Meta-Llama-3-70B-Instruct-fp8" \
        --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": "FriendliAI/Meta-Llama-3-70B-Instruct-fp8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use FriendliAI/Meta-Llama-3-70B-Instruct-fp8 with Docker Model Runner:
```
docker model run hf.co/FriendliAI/Meta-Llama-3-70B-Instruct-fp8
```

New friendli container not working

by Papersnake - opened Jun 2, 2024

Discussion

Papersnake

Jun 2, 2024

docker run \
  -p 8000:8000 \
  --gpus all \
  -v ~/.cache/huggingface:/root/.cache/huggingface \
  -e FRIENDLI_CONTAINER_SECRET=$FRIENDLI_CONTAINER_SECRET \
  $FRIENDLI_CONTAINER_IMAGE \
    --hf-model-name FriendliAI/Meta-Llama-3-8B-Instruct-fp8

However, error occurs:

[2024-06-02 09:09:48.002] [info] [::] gRPC Server listening on 0.0.0.0:0, selected_port=40565
[2024-06-02 09:09:48.003] [info] [::] Worker-0 is launched.
[2024-06-02 09:09:48.003] [info] [::] gRPC Server listening on 0.0.0.0:0, selected_port=43699
[2024-06-02 09:09:48.003] [info] [::] Master is launched.
[2024-06-02 09:09:48.009] [info] [::] Worker-0 at 0.0.0.0:43699 is joined.
[2024-06-02 09:09:48.009] [info] [::] All workers joined.
[2024-06-02 09:09:48.010] [info] [::] Received a worker start message.
[2024-06-02 09:09:48.087] [info] [::] Use a default algorithm policy as no algo policy file provided.
[2024-06-02 09:09:48.087] [info] [::] Worker-0 starts model checkpoint loading.
[2024-06-02 09:09:48.338] [info] [::] Enabled /v1/chat/completions API with chat template: "{% set loop_messages = messages %}{% for message in loop_messages %}{% set content = '<|start_header_id|>' + message['role'] + '<|end_header_id|>\n\n'+ message['content'] | trim + '<|eot_id|>' %}{% if loop.index0 == 0 %}{% set content = bos_token + content %}{% endif %}{{ content }}{% endfor %}{{ '<|start_header_id|>assistant<|end_header_id|>\n\n' }}"
[2024-06-02 09:09:48.772] [info] [::] Try to find sessions from session-uris.
[2024-06-02 09:09:48.773] [warning] [::] Tokenizer's vocab size is 128000, while we got a vocab size of 128256 for the model; make sure that you really want this.
[2024-06-02 09:09:48.773] [info] [::] Found a session at 0.0.0.0:40565, now wait for this session to be pre-ready.
[2024-06-02 09:10:15.015] [info] [::] Worker-0 finished model checkpoint loading.
[2024-06-02 09:10:15.015] [info] [::] Wait for workers pre-ready.
terminate called after throwing an instance of 'pfdnn::NotSupportedException'
  what():  No backend supports this parameter

I'm using Friendli Engine v1.5.27-54b7800d+ , and I'm sure that previous version(about 15 days ago?) works.

Papersnake changed discussion status to closed Jun 2, 2024

Papersnake

Jun 2, 2024

the gpu I'm using doesn't support fp8

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