Add Qwen2-VL gpu example (#12135)
* qwen2-vl readme * add qwen2-vl example * fix * fix * fix * add link * Update regarding modules_to_not_convert and readme * Further fix * Small fix --------- Co-authored-by: Yuwen Hu <yuwen.hu@intel.com>
This commit is contained in:
Jinhe
GitHub
parent
310f18c8af
commit
f983f1a8f4
4 changed files with 253 additions and 0 deletions
|
|
@ -283,6 +283,7 @@ Over 50 models have been optimized/verified on `ipex-llm`, including *LLaMA/LLaM
|
|||
| Qwen2 | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/qwen2) | [link](python/llm/example/GPU/HuggingFace/LLM/qwen2) |
|
||||
| Qwen2.5 | | [link](python/llm/example/GPU/HuggingFace/LLM/qwen2.5) |
|
||||
| Qwen-VL | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/qwen-vl) | [link](python/llm/example/GPU/HuggingFace/Multimodal/qwen-vl) |
|
||||
| Qwen2-VL || [link](python/llm/example/GPU/PyTorch-Models/Model/qwen2-vl) |
|
||||
| Qwen2-Audio | | [link](python/llm/example/GPU/HuggingFace/Multimodal/qwen2-audio) |
|
||||
| Aquila | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/aquila) | [link](python/llm/example/GPU/HuggingFace/LLM/aquila) |
|
||||
| Aquila2 | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/aquila2) | [link](python/llm/example/GPU/HuggingFace/LLM/aquila2) |
|
||||
|
|
|
|||
|
|
@ -283,6 +283,7 @@ See the demo of running [*Text-Generation-WebUI*](https://ipex-llm.readthedocs.i
|
|||
| Qwen2 | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/qwen2) | [link](python/llm/example/GPU/HuggingFace/LLM/qwen2) |
|
||||
| Qwen2.5 | | [link](python/llm/example/GPU/HuggingFace/LLM/qwen2.5) |
|
||||
| Qwen-VL | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/qwen-vl) | [link](python/llm/example/GPU/HuggingFace/Multimodal/qwen-vl) |
|
||||
| Qwen2-VL || [link](python/llm/example/GPU/PyTorch-Models/Model/qwen2-vl) |
|
||||
| Aquila | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/aquila) | [link](python/llm/example/GPU/HuggingFace/LLM/aquila) |
|
||||
| Aquila2 | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/aquila2) | [link](python/llm/example/GPU/HuggingFace/LLM/aquila2) |
|
||||
| MOSS | [link](python/llm/example/CPU/HF-Transformers-AutoModels/Model/moss) | |
|
||||
|
|
|
|||
135
python/llm/example/GPU/PyTorch-Models/Model/qwen2-vl/README.md
Normal file
135
python/llm/example/GPU/PyTorch-Models/Model/qwen2-vl/README.md
Normal file
|
|
@ -0,0 +1,135 @@
|
|||
# Qwen2-VL
|
||||
In this directory, you will find examples on how you could use IPEX-LLM `optimize_model` API to accelerate Qwen2 models on [Intel GPUs](../../../README.md). For illustration purposes, we utilize the [Qwen/Qwen2-VL-7B-Instruct](https://huggingface.co/Qwen/Qwen2-VL-7B-Instruct) as a reference Qwen2-VL model.
|
||||
|
||||
## 0. Requirements
|
||||
To run these examples with IPEX-LLM on Intel GPUs, we have some recommended requirements for your machine, please refer to [here](../../../README.md#requirements) for more information.
|
||||
|
||||
## Example: Predict Tokens using `generate()` API
|
||||
In the example [generate.py](./generate.py), we show a basic use case for a Qwen2-VL model to predict the next N tokens using `generate()` API, with IPEX-LLM INT4 optimizations on Intel GPUs.
|
||||
### 1. Install
|
||||
#### 1.1 Installation on Linux
|
||||
We suggest using conda to manage environment:
|
||||
```bash
|
||||
conda create -n llm python=3.11
|
||||
conda activate llm
|
||||
# below command will install intel_extension_for_pytorch==2.1.10+xpu as default
|
||||
pip install --pre --upgrade ipex-llm[xpu] --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/
|
||||
|
||||
pip install transformers==4.45.0 # install transformers which supports Qwen2-VL
|
||||
pip install accelerate==0.33.0
|
||||
pip install qwen_vl_utils
|
||||
```
|
||||
|
||||
#### 1.2 Installation on Windows
|
||||
We suggest using conda to manage environment:
|
||||
```bash
|
||||
conda create -n llm python=3.11 libuv
|
||||
conda activate llm
|
||||
|
||||
# below command will install intel_extension_for_pytorch==2.1.10+xpu as default
|
||||
pip install --pre --upgrade ipex-llm[xpu] --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/
|
||||
|
||||
pip install transformers==4.45.0 # install transformers which supports Qwen2-VL
|
||||
pip install accelerate==0.33.0
|
||||
pip install qwen_vl_utils
|
||||
```
|
||||
|
||||
### 2. Configures OneAPI environment variables for Linux
|
||||
|
||||
> [!NOTE]
|
||||
> Skip this step if you are running on Windows.
|
||||
|
||||
This is a required step on Linux for APT or offline installed oneAPI. Skip this step for PIP-installed oneAPI.
|
||||
|
||||
```bash
|
||||
source /opt/intel/oneapi/setvars.sh
|
||||
```
|
||||
|
||||
### 3. Runtime Configurations
|
||||
For optimal performance, it is recommended to set several environment variables. Please check out the suggestions based on your device.
|
||||
#### 3.1 Configurations for Linux
|
||||
<details>
|
||||
|
||||
<summary>For Intel Arc™ A-Series Graphics and Intel Data Center GPU Flex Series</summary>
|
||||
|
||||
```bash
|
||||
export USE_XETLA=OFF
|
||||
export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
|
||||
export SYCL_CACHE_PERSISTENT=1
|
||||
```
|
||||
|
||||
</details>
|
||||
|
||||
<details>
|
||||
|
||||
<summary>For Intel Data Center GPU Max Series</summary>
|
||||
|
||||
```bash
|
||||
export LD_PRELOAD=${LD_PRELOAD}:${CONDA_PREFIX}/lib/libtcmalloc.so
|
||||
export SYCL_PI_LEVEL_ZERO_USE_IMMEDIATE_COMMANDLISTS=1
|
||||
export SYCL_CACHE_PERSISTENT=1
|
||||
export ENABLE_SDP_FUSION=1
|
||||
```
|
||||
> Note: Please note that `libtcmalloc.so` can be installed by `conda install -c conda-forge -y gperftools=2.10`.
|
||||
</details>
|
||||
|
||||
<details>
|
||||
|
||||
<summary>For Intel iGPU</summary>
|
||||
|
||||
```bash
|
||||
export SYCL_CACHE_PERSISTENT=1
|
||||
export BIGDL_LLM_XMX_DISABLED=1
|
||||
```
|
||||
|
||||
</details>
|
||||
|
||||
#### 3.2 Configurations for Windows
|
||||
<details>
|
||||
|
||||
<summary>For Intel iGPU</summary>
|
||||
|
||||
```cmd
|
||||
set SYCL_CACHE_PERSISTENT=1
|
||||
set BIGDL_LLM_XMX_DISABLED=1
|
||||
```
|
||||
|
||||
</details>
|
||||
|
||||
<details>
|
||||
|
||||
<summary>For Intel Arc™ A-Series Graphics</summary>
|
||||
|
||||
```cmd
|
||||
set SYCL_CACHE_PERSISTENT=1
|
||||
```
|
||||
|
||||
</details>
|
||||
|
||||
> [!NOTE]
|
||||
> For the first time that each model runs on Intel iGPU/Intel Arc™ A300-Series or Pro A60, it may take several minutes to compile.
|
||||
### 4. Running examples
|
||||
|
||||
```
|
||||
python ./generate.py --repo-id-or-model-path REPO_ID_OR_MODEL_PATH --prompt PROMPT --n-predict N_PREDICT
|
||||
```
|
||||
|
||||
Arguments info:
|
||||
- `--repo-id-or-model-path REPO_ID_OR_MODEL_PATH`: argument defining the huggingface repo id for the Qwen2-VL model (e.g. `Qwen/Qwen2-VL-7B-Instruct`) to be downloaded, or the path to the huggingface checkpoint folder. It is default to be `'Qwen/Qwen2-VL-7B-Instruct'`.
|
||||
- `--prompt PROMPT`: argument defining the prompt to be infered (with integrated prompt format for chat). It is default to be `'AI是什么？'`.
|
||||
- `--n-predict N_PREDICT`: argument defining the max number of tokens to predict. It is default to be `32`.
|
||||
|
||||
#### Sample Output
|
||||
##### [Qwen/Qwen2-VL-7B-Instruct](https://huggingface.co/Qwen/Qwen2-VL-7B-Instruct)
|
||||
```log
|
||||
-------------------- Input Image --------------------
|
||||
http://farm6.staticflickr.com/5268/5602445367_3504763978_z.jpg
|
||||
-------------------- Prompt --------------------
|
||||
Describe this image.
|
||||
-------------------- Output --------------------
|
||||
The image shows a young girl holdinging a white teddy bear in a pink dress in front of a stone wall.
|
||||
```
|
||||
|
||||
The sample input image is (which is fetched from [COCO dataset](https://cocodataset.org/#explore?id=264959)):
|
||||
|
||||
<a href="http://farm6.staticflickr.com/5268/5602445367_3504763978_z.jpg"><img width=400px src="http://farm6.staticflickr.com/5268/5602445367_3504763978_z.jpg" ></a>
|
||||
116
python/llm/example/GPU/PyTorch-Models/Model/qwen2-vl/generate.py
Normal file
116
python/llm/example/GPU/PyTorch-Models/Model/qwen2-vl/generate.py
Normal file
|
|
@ -0,0 +1,116 @@
|
|||
#
|
||||
# Copyright 2016 The BigDL Authors.
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
#
|
||||
|
||||
import torch
|
||||
import time
|
||||
import argparse
|
||||
import numpy as np
|
||||
|
||||
from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
|
||||
from qwen_vl_utils import process_vision_info
|
||||
from ipex_llm import optimize_model
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
parser = argparse.ArgumentParser(description='Predict Tokens using generate() API for Qwen2-VL-7B-Instruct model')
|
||||
parser.add_argument('--repo-id-or-model-path', type=str, default="Qwen/Qwen2-VL-7B-Instruct",
|
||||
help='The huggingface repo id for the Qwen2-VL model to be downloaded'
|
||||
', or the path to the huggingface checkpoint folder')
|
||||
parser.add_argument('--prompt', type=str, default="Describe this image.",
|
||||
help='Prompt to infer')
|
||||
parser.add_argument('--image-url-or-path', type=str,
|
||||
default='http://farm6.staticflickr.com/5268/5602445367_3504763978_z.jpg' ,
|
||||
help='The URL or path to the image to infer')
|
||||
|
||||
parser.add_argument('--n-predict', type=int, default=32,
|
||||
help='Max tokens to predict')
|
||||
|
||||
args = parser.parse_args()
|
||||
model_path = args.repo_id_or_model_path
|
||||
|
||||
model = Qwen2VLForConditionalGeneration.from_pretrained(model_path,
|
||||
trust_remote_code=True,
|
||||
torch_dtype='auto',
|
||||
low_cpu_mem_usage=True,
|
||||
use_cache=True,)
|
||||
|
||||
model = optimize_model(model, low_bit='sym_int4', modules_to_not_convert=["visual"])
|
||||
|
||||
# Use .float() for better output, and use .half() for better speed
|
||||
model = model.half().to("xpu")
|
||||
|
||||
# The following code for generation is adapted from https://huggingface.co/Qwen/Qwen2-VL-7B-Instruct#quickstart
|
||||
|
||||
# The default range for the number of visual tokens per image in the model is 4-16384. You can set min_pixels and max_pixels according to your needs, such as a token count range of 256-1280, to balance speed and memory usage.
|
||||
min_pixels = 256*28*28
|
||||
max_pixels = 1280*28*28
|
||||
processor = AutoProcessor.from_pretrained(model_path, min_pixels=min_pixels, max_pixels=max_pixels)
|
||||
|
||||
prompt = args.prompt
|
||||
image_path = args.image_url_or_path
|
||||
|
||||
messages = [
|
||||
{
|
||||
"role": "user",
|
||||
"content": [
|
||||
{
|
||||
"type": "image",
|
||||
"image": image_path,
|
||||
},
|
||||
{"type": "text", "text": prompt},
|
||||
],
|
||||
}
|
||||
]
|
||||
text = processor.apply_chat_template(
|
||||
messages, tokenize=False, add_generation_prompt=True
|
||||
)
|
||||
image_inputs, video_inputs = process_vision_info(messages)
|
||||
inputs = processor(
|
||||
text=[text],
|
||||
images=image_inputs,
|
||||
videos=video_inputs,
|
||||
padding=True,
|
||||
return_tensors="pt",
|
||||
)
|
||||
inputs = inputs.to('xpu')
|
||||
|
||||
with torch.inference_mode():
|
||||
# warmup
|
||||
generated_ids = model.generate(
|
||||
**inputs,
|
||||
max_new_tokens=args.n_predict
|
||||
)
|
||||
|
||||
st = time.time()
|
||||
generated_ids = model.generate(
|
||||
**inputs,
|
||||
max_new_tokens=args.n_predict
|
||||
)
|
||||
torch.xpu.synchronize()
|
||||
end = time.time()
|
||||
generated_ids = generated_ids.cpu()
|
||||
generated_ids = [
|
||||
output_ids[len(input_ids):] for input_ids, output_ids in zip(inputs.input_ids, generated_ids)
|
||||
]
|
||||
|
||||
response = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
|
||||
print(f'Inference time: {end-st} s')
|
||||
print('-'*20, 'Input Image', '-'*20)
|
||||
print(image_path)
|
||||
print('-'*20, 'Prompt', '-'*20)
|
||||
print(prompt)
|
||||
print('-'*20, 'Output', '-'*20)
|
||||
print(response)
|
||||
Loading…
Reference in a new issue