Can not install genai_server_deps vllm

[Lyra-stellAI]

🔎 Search before asking

• I have searched the PaddleOCR Docs and found no similar bug report.
• I have searched the PaddleOCR Issues and found no similar bug report.
• I have searched the PaddleOCR Discussions and found no similar bug report.

🐛 Bug (问题描述)

When I tried to install !paddleocr install_genai_server_deps vllm, it shows me
Traceback (most recent call last):
File "/usr/local/bin/paddleocr", line 5, in
from paddleocr.main import console_entry
File "/usr/local/lib/python3.12/dist-packages/paddleocr/init.py", line 15, in
from paddlex.inference.utils.benchmark import benchmark
File "/usr/local/lib/python3.12/dist-packages/paddlex/init.py", line 49, in
from .inference import create_pipeline, create_predictor
File "/usr/local/lib/python3.12/dist-packages/paddlex/inference/init.py", line 16, in
from .models import create_predictor
File "/usr/local/lib/python3.12/dist-packages/paddlex/inference/models/init.py", line 22, in
from ..utils.official_models import official_models
File "/usr/local/lib/python3.12/dist-packages/paddlex/inference/utils/official_models.py", line 25, in
import modelscope
File "/usr/local/lib/python3.12/dist-packages/modelscope/init.py", line 5, in
from modelscope.utils.import_utils import (LazyImportModule,
File "/usr/local/lib/python3.12/dist-packages/modelscope/utils/import_utils.py", line 18, in
from modelscope.utils.ast_utils import (INDEX_KEY, MODULE_KEY, REQUIREMENT_KEY,
File "/usr/local/lib/python3.12/dist-packages/modelscope/utils/ast_utils.py", line 24, in
from modelscope.utils.registry import default_group
File "/usr/local/lib/python3.12/dist-packages/modelscope/utils/registry.py", line 11, in
logger = get_logger()
^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/modelscope/utils/logger.py", line 49, in get_logger
from modelscope.utils.torch_utils import is_dist, is_master
File "/usr/local/lib/python3.12/dist-packages/modelscope/utils/torch_utils.py", line 14, in
import torch
File "/usr/local/lib/python3.12/dist-packages/torch/init.py", line 416, in
from torch._C import * # noqa: F403
^^^^^^^^^^^^^^^^^^^^^^
ImportError: /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_cuda.so: undefined symbol: ncclCommWindowRegister

Then I uninstalled nvidia-nccl and reinstalled
nvidia-cudnn-cu12==9.10.2.21
nvidia-cusparselt-cu12==0.7.1
nvidia-nccl-cu12==2.27.3

it then showed me
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
paddlepaddle-gpu 3.2.1 requires nvidia-cudnn-cu12==9.5.1.17; platform_system == "Linux" and platform_machine == "x86_64", but you have nvidia-cudnn-cu12 9.10.2.21 which is incompatible.
paddlepaddle-gpu 3.2.1 requires nvidia-cusparselt-cu12==0.6.3; platform_system == "Linux" and platform_machine == "x86_64", but you have nvidia-cusparselt-cu12 0.7.1 which is incompatible.
paddlepaddle-gpu 3.2.1 requires nvidia-nccl-cu12==2.25.1; platform_system == "Linux" and platform_machine == "x86_64", but you have nvidia-nccl-cu12 2.27.3 which is incompatible.

🏃‍♂️ Environment (运行环境)

Google Colab (Linux)
2.8.0+cu126

🌰 Minimal Reproducible Example (最小可复现问题的Demo)

ImportError: /usr/local/lib/python3.12/dist-packages/torch/lib/libtorch_cuda.so: undefined symbol: ncclCommWindowRegister

[scyyh11]

Since inference acceleration frameworks may have dependency conflicts with the PaddlePaddle framework, it is recommended to install them in a virtual environment.

# If a virtual environment is currently active, deactivate it first
deactivate

# Create a new virtual environment
python -m venv .venv_vlm

# Activate the environment
source .venv_vlm/bin/activate

# Install PaddleOCR
python -m pip install "paddleocr[doc-parser]"

# Install the inference acceleration service dependencies
paddleocr install_genai_server_deps vllm

[zhang-prog]

What specific instructions do you execute? It would be great if you could ensure that the dependencies are installed in the order specified in the documentation.

Additionally, PaddleOCR provides a Docker image for quickly launching the vLLM inference service, and you may refer to here for further guidance.

[TingquanGao]

The issue has no response for a long time and will be closed. You can reopen or new another issue if are still confused.

---

From Bot