Instructions to use radm/Athene-V2-Chat-AWQ with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use radm/Athene-V2-Chat-AWQ with Transformers:

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

pipe = pipeline("text-generation", model="radm/Athene-V2-Chat-AWQ")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("radm/Athene-V2-Chat-AWQ")
model = AutoModelForCausalLM.from_pretrained("radm/Athene-V2-Chat-AWQ")
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 radm/Athene-V2-Chat-AWQ with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "radm/Athene-V2-Chat-AWQ"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "radm/Athene-V2-Chat-AWQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/radm/Athene-V2-Chat-AWQ

SGLang

How to use radm/Athene-V2-Chat-AWQ 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 "radm/Athene-V2-Chat-AWQ" \
    --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": "radm/Athene-V2-Chat-AWQ",
		"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 "radm/Athene-V2-Chat-AWQ" \
        --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": "radm/Athene-V2-Chat-AWQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use radm/Athene-V2-Chat-AWQ with Docker Model Runner:
```
docker model run hf.co/radm/Athene-V2-Chat-AWQ
```

Athene-V2-Chat-72B: Rivaling GPT-4o across Benchmarks

AWQ 4bit version of Nexusflow/Athene-V2-Chat
Quantization code
This model only fits to 1 gpu. Use kosbu/Athene-V2-Chat-AWQ for multi-gpu support

Eval AWQ version

Evaluation results on ZebraLogic

│              Model               │  Mode  │  N_Mode  │  N_Size  │  Puzzle Acc  │  Easy Puzzle Acc  │  Hard Puzzle Acc  │  Cell Acc  │  No answer  │  Total Puzzles  │  Reason Lens  │
│      o1-preview-2024-09-12       │ greedy │  single  │    1     │     71.4     │       98.57       │       60.83       │   75.14    │     0.3     │      1000       │    1565.88    │
│    claude-3-5-sonnet-20241022    │ greedy │  single  │    1     │     36.2     │       91.07       │       14.86       │   54.27    │      0      │      1000       │    861.18     │
│ Llama-3.1-405B-Inst-fp8@together │ greedy │  single  │    1     │     32.6     │       87.14       │       11.39       │    45.8    │    12.5     │      1000       │    314.66     │
│        Athene-V2-Chat-AWQ        │ greedy │  single  │    1     │     27.8     │       77.14       │       8.61        │   45.83    │     6.4     │      1000       │    1785.7     │
│       Qwen2.5-72B-Instruct       │ greedy │  single  │    1     │     26.6     │       76.43       │       7.22        │   40.92    │    11.9     │      1000       │    1795.9     │
│       Qwen2.5-32B-Instruct       │ greedy │  single  │    1     │     26.1     │       77.5        │       6.11        │   43.39    │     6.3     │      1000       │    1333.07    │
│            Athene-70B            │ greedy │  single  │    1     │     16.7     │       52.5        │       2.78        │   32.98    │    21.1     │      1000       │    391.19     │

Downloads last month: 10

Safetensors

Model size

73B params

Tensor type

I32

F16

Model tree for radm/Athene-V2-Chat-AWQ

Base model

Qwen/Qwen2.5-72B

Finetuned

Qwen/Qwen2.5-72B-Instruct

Quantized

(83)

this model