CUDA Runtime Error when using ad4 maps

[NiBoyang]

Describe the bug

The CUDA runtime error happened just after the first batch was completed.

Batch 1 running time: 46419ms
Small
Max num_atoms: 40 Max num_torsions: 12 Max num_rigids: 5 Max num_lig_pairs: 467
Group size: 1309
Small Batch 2 size: 502
Performing docking (random seed: 1512773227) ... atom_type::AD=1
CUDA error at /home/s2263780/Uni-Dock/unidock/src/cuda/monte_carlo.cu:2854 code=700(cudaErrorIllegalAddress) "cudaDeviceSynchronize()"


An error occurred: CUDA Runtime Error.

Expected behavior

No response

To Reproduce

The command I used was

unidock \
--maps /home/s2263780/unidock_test/ad4map/receptor \
--gpu_batch $(find /home/s2263780/unidock_test/lib/ -name "*.pdbqt" -print) \
--search_mode balance \
--scoring ad4 \
--center_x 9.297 \
--center_y 12.077 \
--center_z -21.830 \
--size_x 26 \
--size_y 28 \
--size_z 24 \
--num_modes 1 \
--dir /home/s2263780/unidock_test/results/

unidock_test.tar.gz

Environment

• CUDA 12.4
• GPU RTX4080
• Uni-Dock v1.1.1

Additional Context

No response

[ysyecust]

Hi @NiBoyang can you show me your CUDA version? Uni-Dock need your CUDA version >=11.8.

[NiBoyang]

Hi @ysyecust , thanks for you reply. Please find the attached screenshot which contains the error (first batch completed but the second one crashed) and the CUDA version 12.4.

[NiBoyang]

Also if I use the ligand_index parameter (file_paths.txt in the compressed), it shows 'segmentation fault' as #112

[NiBoyang]

Any update on this issue?

Many thanks,
Boyang