Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add GPU model support policy #70

Closed
wants to merge 1 commit into from
Closed

Add GPU model support policy #70

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 4 additions & 4 deletions unidock/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,11 +17,11 @@ check_language(CUDA)
if(NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
# https://en.wikipedia.org/wiki/CUDA#GPUs_supported
set(CMAKE_CUDA_ARCHITECTURES
60 # P100
70 # V100
75 # T4
80 # A100
# 89 # RTX 40
75 # RTX 20, T4
80 # A100, A800
86 # RTX 30
89 # RTX 40, L40
90 # H100
)
endif()
Expand Down
37 changes: 20 additions & 17 deletions unidock/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,26 +19,26 @@ Developed by [DP Technology](https://dp.tech/en), [Hermite®](https://dp.tech/en

Uni-Dock is now available on the new-generation drug computing design platform [Hermite®](https://dp.tech/en/product/hermite) for ultralarge virtual screening.

For further cooperations on developing Uni-Dock and trying out Hermite®, please contact us at [email protected] .
For further cooperations on developing Uni-Dock and trying out Hermite®, please contact us at <[email protected]> .

## Installation

Uni-Dock supports NVIDIA GPUs on Linux platform.
[CUDA toolkit](https://developer.nvidia.com/cuda-downloads) is required.
Uni-Dock officially supports NVIDIA GPUs with [compute capability](https://en.wikipedia.org/wiki/CUDA#GPUs_supported) >= 7.0 on Linux platform.
The performance is not guaranteed on legacy GPU models. To build Uni-Dock with a customized compute capability, please change the `CMAKE_CUDA_ARCHITECTURES` variable in `CMakeLists.txt` to the corresponding value.

### Building from source

1. Install dependencies

- CUDA toolkit: Please refer to the [installation tutorial](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html) provided by nvidia.
- CMake >= 3.16
- A C++ compiler (should be compatible with NVCC; `g++` works for most cases)
- A C++ compiler (should be [compatible with NVCC](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#host-compiler-support-policy); `g++` works for most cases)
- Boost >= 1.72

```bash
sudo apt install libboost-system-dev libboost-thread-dev libboost-serialization-dev libboost-filesystem-dev libboost-program-options-dev libboost-timer-dev
```

Alternatively, install from [boost source codes](https://www.boost.org/users/download/) or from [anaconda](https://anaconda.org/anaconda/libboost) with `conda install -c anaconda libboost`

2. Clone repo or retrieve source code from [release page](https://github.com/dptech-corp/Uni-Dock/releases):
Expand All @@ -47,8 +47,8 @@ Uni-Dock supports NVIDIA GPUs on Linux platform.
git clone https://github.com/dptech-corp/Uni-Dock.git
```

4. Build Uni-Dock using CMake:
3. Build Uni-Dock using CMake:

```bash
cd Uni-Dock/unidock
cmake -B build
Expand All @@ -57,12 +57,13 @@ Uni-Dock supports NVIDIA GPUs on Linux platform.
# Otherwise, prepend the building directory to your `PATH` environment variable.
```


code foramt

```shell
cd ./build/
make clang-format
```

### Using binary

Please download the latest binary of Uni-Dock at the assets tab of [the Release page](https://github.com/dptech-corp/Uni-Dock/releases).
Expand All @@ -78,7 +79,7 @@ To launch a Uni-Dock job, the most important parameters are as follows:

- `--receptor`: filepath of the receptor (PDBQT)
- `--gpu_batch`: filepath of the ligands to dock with GPU (PDBQT), enter multiple at a time, separated by spaces (" ")
- `--search_mode`: computational complexity, choice in [*`fast`*, *`balance`*, and *`detail`*].
- `--search_mode`: computational complexity, choose in \[*`fast`*, *`balance`*, and *`detail`*].

***Advanced options***
`--search_mode` is the recommended setting of `--exhaustiveness` and `--max_step`, with three combinations called `fast`, `balance`, and `detail`.
Expand Down Expand Up @@ -205,7 +206,7 @@ If you used Uni-Dock in your work, please cite:
Yu, Y., Cai, C., Wang, J., Bo, Z., Zhu, Z., & Zheng, H. (2023).
Uni-Dock: GPU-Accelerated Docking Enables Ultralarge Virtual Screening.
Journal of Chemical Theory and Computation.
https://doi.org/10.1021/acs.jctc.2c01145
<https://doi.org/10.1021/acs.jctc.2c01145>

Tang, S., Chen, R., Lin, M., Lin, Q., Zhu, Y., Ding, J., ... & Wu, J. (2022).
Accelerating autodock vina with gpus. Molecules, 27(9), 3041.
Expand All @@ -225,7 +226,7 @@ DOI 10.1002/jcc.21334
## FAQ

1. The GPU encounters out-of-memory error.

Uni-Dock estimates the number of ligands put into GPU memory in one pass based on the available GPU memory size. If it fails, please use `--max_gpu_memory` to limit the usage of GPU memory size by Uni-Dock.

2. I want to put all my ligands in `--gpu_batch`, but it exceeds the maximum command line length that linux can accept.
Expand All @@ -235,13 +236,14 @@ DOI 10.1002/jcc.21334
3. Uni-Dock computes slowly for few (<10) ligands.

The optimal application of Uni-Dock occurs in scenarios where one binding pocket interacts with numerous (in an order of 1000) ligands. As the number of ligands within a single computational batch increases, the average processing speed improves. In instances where only a few ligands are present for one binding pocket, the overhead proportion becomes considerably large, leading to slower computational performance.
### Addendum to "FAQ 3 - Uni-Dock computes slowly for few (<10) ligands.":

### Addendum to "FAQ 3 - Uni-Dock computes slowly for few (<10) ligands."

The `paired_batch` mode provides a mechanism to accelerate simultaneous 1:1 docking in batches using Vina scoring, using CUDA streams. To run docking using this mode, invoke unidock with the parameter `--paired_batch_size` value >0, with the protein:ligand configurations passed in JSON form using `--ligand_index`. The JSON file should use schema defined in paired_batching.schema.json.

A sample input data.json is as below, complying to the schema:
```

```json
{
"7LCU": {
"protein": "molecular_docking/PoseBuster/7LCU/7LCU_receptor.pdbqt",
Expand All @@ -258,5 +260,6 @@ A sample input data.json is as below, complying to the schema:

A typical usage using paired_batch mode is as shown below, with batch size of 10.

`build/unidock --paired_batch_size 10 --ligand_index data_pb1.json --size_x 25 --size_y 25 --size_z 25 --dir test/prof_25_1024_80 --exhaustiveness 1024 --max_step 80 --seed 5`

```bash
unidock --paired_batch_size 10 --ligand_index data_pb1.json --size_x 25 --size_y 25 --size_z 25 --dir test/prof_25_1024_80 --exhaustiveness 1024 --max_step 80 --seed 5
```
Loading