Instructions to use TheBloke/Mixtral-8x7B-Instruct-v0.1-AWQ with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use TheBloke/Mixtral-8x7B-Instruct-v0.1-AWQ with Transformers:

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

pipe = pipeline("text-generation", model="TheBloke/Mixtral-8x7B-Instruct-v0.1-AWQ")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("TheBloke/Mixtral-8x7B-Instruct-v0.1-AWQ")
model = AutoModelForCausalLM.from_pretrained("TheBloke/Mixtral-8x7B-Instruct-v0.1-AWQ", device_map="auto")
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/Mixtral-8x7B-Instruct-v0.1-AWQ with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "TheBloke/Mixtral-8x7B-Instruct-v0.1-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": "TheBloke/Mixtral-8x7B-Instruct-v0.1-AWQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/TheBloke/Mixtral-8x7B-Instruct-v0.1-AWQ

SGLang

How to use TheBloke/Mixtral-8x7B-Instruct-v0.1-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 "TheBloke/Mixtral-8x7B-Instruct-v0.1-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": "TheBloke/Mixtral-8x7B-Instruct-v0.1-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 "TheBloke/Mixtral-8x7B-Instruct-v0.1-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": "TheBloke/Mixtral-8x7B-Instruct-v0.1-AWQ",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use TheBloke/Mixtral-8x7B-Instruct-v0.1-AWQ with Docker Model Runner:
```
docker model run hf.co/TheBloke/Mixtral-8x7B-Instruct-v0.1-AWQ
```

always getting 0 in output

by xubuild - opened Jan 10, 2024

Discussion

xubuild

Jan 10, 2024

the model always respond with token id 0 for any input, while same prompt can get correct response from https://huggingface.co/casperhansen/mixtral-instruct-awq

tested with vllm
llm = LLM(model=model_path, quantization="awq", trust_remote_code=True, dtype="auto", enforce_eager=True, max_model_len=12288)

sd3ntato

Jan 11, 2024

same for me!

hassantrixly

Jan 12, 2024

Is there any fix for that?

hnhlester

Jan 12, 2024

Also happening for me. Seems to be an issue with this model. Other 8x7b awq models work perfectly fine, for example dolphin-2.7-8x7b.

bd2lcco

Jan 15, 2024

@TheBloke Any update about this issue? Always get an empty output as well...

FYI,

python3 -m vllm.entrypoints.openai.api_server --model "$MODEL_NAME" --host 0.0.0.0 --port 8181 --quantization awq --dtype auto

    "choices": [
        {
            "index": 0,
            "text": "",
            "logprobs": null,
            "finish_reason": "length"
        }
    ],

Bourhano

Jan 15, 2024

•

edited Jan 15, 2024

same, maybe a prompt template issue?

Fouldon

Jan 15, 2024

Any update ?

dafen

Jan 17, 2024

俺也一样

hiranya911

Jan 17, 2024

Also seeing this in testing. Our vLLM setup:

sampling_params = SamplingParams(temperature=0.1, top_p=0.95)
model = 'TheBloke/Mixtral-8x7B-Instruct-v0.1-AWQ'

llm = LLM(
  model=model, 
  gpu_memory_utilization=0.7, 
  max_model_len=2048,
)