Instructions to use Qwen/Qwen1.5-14B-Chat with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Qwen/Qwen1.5-14B-Chat with Transformers:

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

pipe = pipeline("text-generation", model="Qwen/Qwen1.5-14B-Chat")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen1.5-14B-Chat")
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen1.5-14B-Chat", 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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Qwen/Qwen1.5-14B-Chat with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Qwen/Qwen1.5-14B-Chat"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Qwen/Qwen1.5-14B-Chat",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Qwen/Qwen1.5-14B-Chat

SGLang

How to use Qwen/Qwen1.5-14B-Chat 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 "Qwen/Qwen1.5-14B-Chat" \
    --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": "Qwen/Qwen1.5-14B-Chat",
		"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 "Qwen/Qwen1.5-14B-Chat" \
        --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": "Qwen/Qwen1.5-14B-Chat",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Qwen/Qwen1.5-14B-Chat with Docker Model Runner:
```
docker model run hf.co/Qwen/Qwen1.5-14B-Chat
```

support marlin kernel

#11

by xun - opened Apr 11, 2024

Discussion

xun

Apr 11, 2024

Repacking weights to be compatible with Marlin kernel...:   5%|███▉                                                                         | 15/296 [00:00<00:02, 120.10it/s]
Traceback (most recent call last):
  File "/data/vllm/to_marlin.py", line 5, in <module>
    marlin_model = AutoGPTQForCausalLM.from_quantized(
  File "/data/test/AutoGPTQ/auto_gptq/modeling/auto.py", line 142, in from_quantized
    return quant_func(
  File "/data/test/AutoGPTQ/auto_gptq/modeling/_base.py", line 1100, in from_quantized
    model, model_save_name = prepare_model_for_marlin_load(
  File "/data/test/AutoGPTQ/auto_gptq/utils/marlin_utils.py", line 63, in prepare_model_for_marlin_load
    model = convert_to_marlin(model, quant_linear_class, quantize_config, repack=True)
  File "/data/anaconda3/envs/qwen-q/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
    return func(*args, **kwargs)
  File "/data/test/AutoGPTQ/auto_gptq/utils/marlin_utils.py", line 156, in convert_to_marlin
    new_module = MarlinQuantLinear(
  File "/data/test/AutoGPTQ/auto_gptq/nn_modules/qlinear/qlinear_marlin.py", line 98, in __init__
    raise ValueError(f"`infeatures:{infeatures}` must be divisible by 128 and `outfeatures:{outfeatures}` by 256.")
ValueError: `infeatures:2048` must be divisible by 128 and `outfeatures:5504` by 256.

xun

Apr 11, 2024

from transformers import AutoTokenizer
from auto_gptq import AutoGPTQForCausalLM

GPTQ_MODEL = "/data/test/Qwen-1_8B-Chat-Int4"
marlin_model = AutoGPTQForCausalLM.from_quantized(
      GPTQ_MODEL,
      use_marlin=True,
      device_map='auto',
      trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(
        "/data/test/Qwen-1_8B-Chat-Int4",
        trust_remote_code=True
    )
save_dir = "/data/test/Qwen-1_8B-Chat-Int4-marlin"
marlin_model.save_pretrained(save_dir)
tokenizer.save_pretrained(save_dir)
~

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