Instructions to use Nexusflow/Athene-70B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Nexusflow/Athene-70B with Transformers:

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

pipe = pipeline("text-generation", model="Nexusflow/Athene-70B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Nexusflow/Athene-70B")
model = AutoModelForCausalLM.from_pretrained("Nexusflow/Athene-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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Nexusflow/Athene-70B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Nexusflow/Athene-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": "Nexusflow/Athene-70B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Nexusflow/Athene-70B

SGLang

How to use Nexusflow/Athene-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 "Nexusflow/Athene-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": "Nexusflow/Athene-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 "Nexusflow/Athene-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": "Nexusflow/Athene-70B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Nexusflow/Athene-70B with Docker Model Runner:
```
docker model run hf.co/Nexusflow/Athene-70B
```

Fails addition that llama-3.0 can solve

by SaisExperiments - opened Jul 26, 2024

Discussion

SaisExperiments

Jul 26, 2024

•

edited Jul 26, 2024

Out of curiosity I decided to sum up the total cost of all the games on my steam wishlist.
I got to a final equation of

28.95+67.95+14.75+99.95+7.39+42.99+14.75+35.99+26.29+89.95+14.75+7.39+71.49+59.95+49.95+89.95+99.99+54.95+36.99+69.95+34.95+29.95+39.95+49.95+89.95+28.99+104.95+89.95+56.95+99.95+17.99+34.99+17.99+7.39+21.99+35.99+99.99+43.59+61.99

Using lmsys arena I found athene got this question wrong every time I asked the model whereas llama-3.0-70B-Instruct gets it right everytime.

Athene-70B

Here's the correct total:

28.95 + 67.95 + 14.75 + 99.95 + 7.39 + 42.99 + 14.75 + 35.99 + 26.29 + 89.95 + 14.75 + 7.39 + 71.49 + 59.95 + 49.95 + 89.95 + 99.99 + 54.95 + 36.99 + 69.95 + 34.95 + 29.95 + 39.95 + 49.95 + 89.95 + 28.99 + 104.95 + 89.95 + 56.95 + 99.95 + 17.99 + 34.99 + 17.99 + 7.39 + 21.99 + 35.99 + 99.99 + 43.59 + 61.99 =

1,532.91

Let me know if you need any further assistance!

Llama-3.0-70B-Instruct:

Let me add up those numbers for you:

28.95 + 67.95 = 96.90
96.90 + 14.75 = 111.65
111.65 + 99.95 = 211.60
211.60 + 7.39 = 218.99
218.99 + 42.99 = 261.98
261.98 + 14.75 = 276.73
276.73 + 35.99 = 312.72
312.72 + 26.29 = 339.01
339.01 + 89.95 = 428.96
428.96 + 14.75 = 443.71
443.71 + 7.39 = 451.10
451.10 + 71.49 = 522.59
522.59 + 59.95 = 582.54
582.54 + 49.95 = 632.49
632.49 + 89.95 = 722.44
722.44 + 99.99 = 822.43
822.43 + 54.95 = 877.38
877.38 + 36.99 = 914.37
914.37 + 69.95 = 984.32
984.32 + 34.95 = 1019.27
1019.27 + 29.95 = 1049.22
1049.22 + 39.95 = 1089.17
1089.17 + 49.95 = 1139.12
1139.12 + 89.95 = 1229.07
1229.07 + 28.99 = 1258.06
1258.06 + 104.95 = 1363.01
1363.01 + 89.95 = 1452.96
1452.96 + 56.95 = 1509.91
1509.91 + 99.95 = 1609.86
1609.86 + 17.99 = 1627.85
1627.85 + 34.99 = 1662.84
1662.84 + 17.99 = 1680.83
1680.83 + 7.39 = 1688.22
1688.22 + 21.99 = 1710.21
1710.21 + 35.99 = 1746.20
1746.20 + 99.99 = 1846.19
1846.19 + 43.59 = 1889.78
1889.78 + 61.99 = 1951.77

The grand total is: 1951.77

Llama-3.0-8B-Instruct

Let me calculate the sum for you:

28.95 + 67.95 = 96.9
96.9 + 14.75 = 111.65
111.65 + 99.95 = 211.6
211.6 + 7.39 = 218.99
218.99 + 42.99 = 261.98
261.98 + 14.75 = 276.73
276.73 + 35.99 = 312.72
312.72 + 26.29 = 339.01
339.01 + 89.95 = 428.96
428.96 + 14.75 = 443.71
443.71 + 7.39 = 451.1
451.1 + 71.49 = 522.59
522.59 + 59.95 = 582.54
582.54 + 49.95 = 632.49
632.49 + 89.95 = 722.44
722.44 + 99.99 = 822.43
822.43 + 54.95 = 877.38
877.38 + 36.99 = 914.37
914.37 + 69.95 = 984.32
984.32 + 34.95 = 1019.27
1019.27 + 29.95 = 1049.22
1049.22 + 39.95 = 1089.17
1089.17 + 49.95 = 1139.12
1139.12 + 89.95 = 1229.07
1229.07 + 28.99 = 1257.96
1257.96 + 104.95 = 1362.91
1362.91 + 89.95 = 1452.86
1452.86 + 56.95 = 1509.81
1509.81 + 99.95 = 1609.76
1609.76 + 17.99 = 1627.75
1627.75 + 34.99 = 1662.74
1662.74 + 17.99 = 1680.73
1680.73 + 7.39 = 1688.12
1688.12 + 21.99 = 1710.11
1710.11 + 35.99 = 1746.1
1746.1 + 99.99 = 1846.09
1846.09 + 43.59 = 1889.68
1889.68 + 61.99 = 1951.67

The final sum is: 1951.67

So llama-3.0-8B is only off by 10¢ while Athene is off by 418.86
Another answer from Athene is off by 702.56
It's answers are super random

Google's Gemini app thing solves it easily, but I'm pretty sure it's using tools of some sort as the api fails the same request

Edit - I forgot to specify the correct answer 1951.77

banghua

Nexusflow org Jul 26, 2024

Thx for the feedback. Apparently the model tries to calculate the sum directly which is almost impossible to get it correct. Also the current Athene-70B is indeed im-a-little-birdie on lmsys, and athene-70b on lmsys is indeed Athene-70B-preview on HF. We haven't got to change the model name on lmsys.

When I try prompting im-a-little-birdie with the following prompt it seems to get it correct with temperature 0. With temperature 0.7 (default on arena) it seems right in most cases but occasionally off by a little bit.

Calculate step by step. 28.95+67.95+14.75+99.95+7.39+42.99+14.75+35.99+26.29+89.95+14.75+7.39+71.49+59.95+49.95+89.95+99.99+54.95+36.99+69.95+34.95+29.95+39.95+49.95+89.95+28.99+104.95+89.95+56.95+99.95+17.99+34.99+17.99+7.39+21.99+35.99+99.99+43.59+61.99

banghua changed discussion status to closed Jul 26, 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