fix(backend): disable cublaslt for cu102 (#947)

* fix(backend): disable cublaslt for cu102

* fix

* fix(backend): update

* fix(tensorrt/util.py): add find cuda version

* fix

* fix(CI): first use cmd to get cuda version

* docs(tensorrt/utils.py): update docstring

docs/en/01-how-to-build/linux-x86_64.md

@@ -323,6 +323,7 @@ pip install -e .
 - Some dependencies are optional. Simply running `pip install -e .` will only install the minimum runtime requirements.
   To use optional dependencies, install them manually with `pip install -r requirements/optional.txt` or specify desired extras when calling `pip` (e.g. `pip install -e .[optional]`).
   Valid keys for the extras field are: `all`, `tests`, `build`, `optional`.
+- It is recommended to [install patch for cuda10](https://developer.nvidia.com/cuda-10.2-download-archive?target_os=Linux&target_arch=x86_64&target_distro=Ubuntu&target_version=1804&target_type=runfilelocal), otherwise GEMM related errors may occur when model runs
 
 ### Build SDK and Demo
 

docs/zh_cn/01-how-to-build/linux-x86_64.md

@@ -319,6 +319,7 @@ pip install -e .
 
 - 有些依赖项是可选的。运行 `pip install -e .` 将进行最小化依赖安装。 如果需安装其他可选依赖项，请执行`pip install -r requirements/optional.txt`，
   或者 `pip install -e .[optional]`。其中，`[optional]`可以替换为：`all`、`tests`、`build` 或 `optional`。
+- cuda10 建议安装[补丁包](https://developer.nvidia.com/cuda-10.2-download-archive?target_os=Linux&target_arch=x86_64&target_distro=Ubuntu&target_version=1804&target_type=runfilelocal)，否则模型运行可能出现 GEMM 相关错误
 
 #### 编译 SDK 和 Demos
 

mmdeploy/backend/tensorrt/utils.py

@@ -1,5 +1,8 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 import logging
+import os
+import re
+import sys
 from typing import Dict, Optional, Sequence, Union
 
 import onnx
@@ -38,6 +41,55 @@ def load(path: str) -> trt.ICudaEngine:
         return engine
 
 
+def search_cuda_version() -> str:
+    """try cmd to get cuda version, then try `torch.cuda`
+
+    Returns:
+        str: cuda version, for example 10.2
+    """
+
+    version = None
+
+    pattern = re.compile(r'[0-9]+\.[0-9]+')
+    platform = sys.platform.lower()
+
+    def cmd_result(txt: str):
+        cmd = os.popen(txt)
+        return cmd.read().rstrip().lstrip()
+
+    if platform == 'linux' or platform == 'darwin' or platform == 'freebsd':  # noqa E501
+        version = cmd_result(
+            " nvcc --version | grep  release | awk '{print $5}' | awk -F , '{print $1}' "  # noqa E501
+        )
+        if version is None or pattern.match(version) is None:
+            version = cmd_result(
+                " nvidia-smi  | grep CUDA | awk '{print $9}' ")
+
+    elif platform == 'win32' or platform == 'cygwin':
+        # nvcc_release = "Cuda compilation tools, release 10.2, V10.2.89"
+        nvcc_release = cmd_result(' nvcc --version | find "release" ')
+        if nvcc_release is not None:
+            result = pattern.findall(nvcc_release)
+            if len(result) > 0:
+                version = result[0]
+
+        if version is None or pattern.match(version) is None:
+            # nvidia_smi = "| NVIDIA-SMI 440.33.01    Driver Version: 440.33.01    CUDA Version: 10.2     |" # noqa E501
+            nvidia_smi = cmd_result(' nvidia-smi | find "CUDA Version" ')
+            result = pattern.findall(nvidia_smi)
+            if len(result) > 2:
+                version = result[2]
+
+    if version is None or pattern.match(version) is None:
+        try:
+            import torch
+            version = torch.version.cuda
+        except Exception:
+            pass
+
+    return version
+
+
 def from_onnx(onnx_model: Union[str, onnx.ModelProto],
               output_file_prefix: str,
               input_shapes: Dict[str, Sequence[int]],
@@ -118,6 +170,16 @@ def from_onnx(onnx_model: Union[str, onnx.ModelProto],
 
     config = builder.create_builder_config()
     config.max_workspace_size = max_workspace_size
+
+    cuda_version = search_cuda_version()
+    if cuda_version is not None:
+        version_major = int(cuda_version.split('.')[0])
+        if version_major < 11:
+            # cu11 support cublasLt, so cudnn heuristic tactic should disable CUBLAS_LT # noqa E501
+            tactic_source = config.get_tactic_sources() - (
+                1 << int(trt.TacticSource.CUBLAS_LT))
+            config.set_tactic_sources(tactic_source)
+
     profile = builder.create_optimization_profile()
 
     for input_name, param in input_shapes.items():