chore: clean up before opening PR

Author: amorehead

configs/scripts/build_inference_script.yaml

@@ -29,5 +29,5 @@ methods_to_sweep: [
   ] # the methods to sweep
 vina_binding_site_methods_to_sweep: ["p2rank"] # the Vina binding site prediction methods to sweep
 ensemble_ranking_methods_to_sweep: ["consensus"] # the ensemble ranking methods to sweep - NOTE: must be one of (`consensus`, `ff`)
-datasets_to_sweep: ["posebusters_benchmark", "astex_diverse", "dockgen"] # the datasets to sweep
+datasets_to_sweep: ["posebusters_benchmark", "astex_diverse", "dockgen", "casp15"] # the datasets to sweep
 num_sweep_repeats: 3 # the number of repeats to run for each method-dataset sweep (if the method is a generative method)

configs/scripts/build_interaction_analysis_script.yaml

@@ -18,4 +18,4 @@ methods_to_sweep: [
     "alphafold3_ss",
     "alphafold3",
   ] # the methods to sweep
-datasets_to_sweep: ["astex_diverse", "dockgen", "posebusters_benchmark"] # the datasets to sweep
+datasets_to_sweep: ["astex_diverse", "dockgen", "posebusters_benchmark", "casp15"] # the datasets to sweep

notebooks/posebusters_method_interaction_analysis_plotting.ipynb

@@ -142,7 +142,7 @@
     "    \"alphafold3\": \"AF3\",\n",
     "}\n",
     "\n",
-    "MAX_POSEBUSTERS_BENCHMARK_ANALYSIS_PROTEIN_SEQUENCE_LENGTH = 2000  # Only PoseBusters Benchmark targets with protein sequences below this threshold can be analyzed"
+    "MAX_POSEBUSTERS_BENCHMARK_ANALYSIS_PROTEIN_SEQUENCE_LENGTH = 700  # Only PoseBusters Benchmark targets with protein sequences below this threshold can be analyzed"
    ]
   },
   {

notebooks/posebusters_method_interaction_analysis_plotting.py

@@ -107,7 +107,7 @@
     "alphafold3": "AF3",
 }
 
-MAX_POSEBUSTERS_BENCHMARK_ANALYSIS_PROTEIN_SEQUENCE_LENGTH = 2000  # Only PoseBusters Benchmark targets with protein sequences below this threshold can be analyzed
+MAX_POSEBUSTERS_BENCHMARK_ANALYSIS_PROTEIN_SEQUENCE_LENGTH = 700  # Only PoseBusters Benchmark targets with protein sequences below this threshold can be analyzed
 
 # %% [markdown]
 # #### Define utility functions

posebench/models/ensemble_generation.py

@@ -144,27 +144,29 @@ def insert_hpc_headers(
         string.
     :return: Batch headers string for SLURM job scheduling.
     """
-    return """######################### Batch Headers #########################
-#SBATCH --qos=shared                                          # use specified partition for job
-#SBATCH --image=registry.nersc.gov/m5008/acmwhb/posebench:0.0.1 # use specified container image
-#SBATCH --account=m5008                                       # use specified account for billing (e.g., `m5008` for AI4Science projects)
-#SBATCH --nodes=1                                             # NOTE: this needs to match Lightning's `Trainer(num_nodes=...)`
-#SBATCH --ntasks-per-node=1                                   # NOTE: this needs to be `1` on SLURM clusters when using Lightning's `ddp_spawn` strategy`; otherwise, set to match Lightning's quantity of `Trainer(devices=...)`
-#SBATCH --time=00-05:00:00                                    # time limit for the job (up to 2 days: `02-00:00:00`)
-#SBATCH --job-name=inference_analysis_sweep                         # job name
-#SBATCH --output=scripts/perlmutter/regular/logs/inference_analysis_sweep%j.out  # output log file
-#SBATCH --error=scripts/perlmutter/regular/logs/inference_analysis_sweep%j.err   # error log file
-
-# Wait for 5-10 seconds randomly to avoid race condition
-sleep $((RANDOM % 6 + 5))
-
-# Determine location of the project's directory
-# PROJECT_ID="m5008"
-# PROJECT_DIR="/global/cfs/cdirs/$PROJECT_ID/$USER/Repositories/posebench"            # long term storage community drive
-PROJECT_DIR="/pscratch/sd/a/$USER/Repositories/posebench"                   # high-performance storage scratch drive with an 8-week purge policy
-cd "$PROJECT_DIR" || exit
-
-"""
+    return f"""######################### Batch Headers #########################
+#SBATCH --partition {gpu_partition} # use reserved partition `chengji-lab-gpu`
+#SBATCH --account {gpu_account}  # NOTE: this must be specified to use the reserved partition above
+#SBATCH --nodes=1              # NOTE: this needs to match Lightning's `Trainer(num_nodes=...)`
+#SBATCH --gres gpu:{f'{gpu_type}:' if gpu_type else ''}1      # request {gpu_type} GPU resource(s)
+#SBATCH --ntasks-per-node=1    # NOTE: this needs to be `1` on SLURM clusters when using Lightning's `ddp_spawn` strategy`; otherwise, set to match Lightning's quantity of `Trainer(devices=...)`
+#SBATCH --mem={cpu_memory_in_gb}G              # NOTE: use `--mem=0` to request all memory "available" on the assigned node
+#SBATCH -t {time_limit}          # time limit for the job (up to two days: `2-00:00:00`)
+#SBATCH -J posebench_{method}_ensembling # job name
+#SBATCH --output=R-%x.%j.out   # output log file
+#SBATCH --error=R-%x.%j.err    # error log file
+module purge
+module load cuda/11.8.0_gcc_9.5.0
+# determine location of the project directory
+use_private_project_dir=false # NOTE: customize as needed
+if [ "$use_private_project_dir" = true ]; then
+    project_dir="/home/$USER/data/Repositories/Lab_Repositories/PoseBench"
+else
+    project_dir="/cluster/pixstor/chengji-lab/$USER/Repositories/Lab_Repositories/PoseBench"
+fi
+# shellcheck source=/dev/null
+source /home/$USER/mambaforge/etc/profile.d/conda.sh
+cd "$project_dir" || exit"""
 
 
 def create_diffdock_bash_script(