yaledeus
commented
9 months ago For the environment
Sorry for the inconvenience caused by the incompatible environment files environment.yml. We have updated README and list external python packages that need to be downloaded.
For dataset
The bash file works well in my server with the following outputs:
PDB_DIR: ***
download sabdab from summary file ../summaries/sabdab_summary_all.tsv
using local PDB files: **/all_structures/imgt
PDB file already renumbered with scheme imgt
downloading raw files
8CAF not found in **/all_structures/imgt, try fetching from remote server
7XY4 not found in **/all_structures/imgt, try fetching from remote server
7XY3 not found in **/all_structures/imgt, try fetching from remote server
7Y0N not found in **/all_structures/imgt, try fetching from remote server
0%| | 0/7709 [00:00<?, ?it/s]It should be noted that the downloaded items are loaded in memory until the downloading process finishes. We recommend a more efficient way mentioned here. First download the zip file from the official website of SAbDab as shown below:
Then you can put the unzipped directory under your PDB_DIR and run download_and_split.sh.
For training and evaluation
From your description, I'm not quite sure what your specific problem is during the training phase. Similarly, we cannot reproduce the error you reported during the inference phase. Part of the outputs of running the inference code on our machine are as follows:
python test.py --dataset DB5 --gpu 0 --ckpt ./checkpoints/db5_best.ckpt
Model type: ElliDock 100%████████████████████████████████| 25/25 [00:55<00:00, 2.22s/it] total runtime: 81.07781410217285 CRMSD(aligned) median: 12.995396614074707 mean: 14.41262475013733 std: 6.779665588791658 IRMSD(aligned) median: 11.134011268615723 mean: 12.47975996017456 std: 4.965853350631719
Please confirm whether your environment contains incompatible packages or other related problems.
ELMIAR-0642
Had to manually build the conda enviroinment and install the necessary libraries and packages, because clearly the python version mentioned in line 130@enviroinment.yml makes a conflict while installing so either creation freezes or fails.
Tried Checking out every step mentioned in the README.md
Can't Download and Split SabDab Data as mentioned here
Training code doesn't work. Gave the GPU id as 0 after verifying through the following code snippet,
still it fails.
Moved on and tried to run inference code, and it gave
Segmentation fault (core dumped)error. Tried to trace it back and added try-exception block above every function calls, and print statements, and found the linemodel = torch.load(args.ckpt, map_location='cpu'), intest.pydoes not load the model file, for some reason.Output after adding Try-Except blocks can be seen here:
Can you please Address the issue as early as possible?
Segmentation fault (core dumped)