Merge pull request #5 from calico/tboard

Tboard

src/westminster/scripts/westminster_classify.py

@@ -35,11 +35,12 @@ def main():
         help="Take features absolute value [Default: %default]",
     )
     parser.add_option(
-        '-f',
-        dest='num_folds',
+        "-f",
+        dest="num_folds",
         default=8,
-        type='int',
-        help='Cross-validation folds [Default: %default]')
+        type="int",
+        help="Cross-validation folds [Default: %default]",
+    )
     parser.add_option(
         "-i",
         dest="iterations",

src/westminster/scripts/westminster_gtex_coef.py

@@ -39,11 +39,12 @@ def main():
         help="GTEx VCF directory",
     )
     parser.add_argument(
-        '-m',
-        '--min_variants',
+        "-m",
+        "--min_variants",
         default=32,
         type=int,
-        help='Minimum number of variants for tissue to be included')
+        help="Minimum number of variants for tissue to be included",
+    )
     parser.add_argument(
         "-p", "--plot", action="store_true", help="Generate tissue prediction plots"
     )
@@ -104,11 +105,15 @@ def main():
             gtex_scores_file = f"{args.gtex_dir}/{tissue}_pos/scores.h5"
             try:
                 variant_scores = read_scores(
-                    gtex_scores_file, keyword, eqtl_df, args.snp_stat, verbose=args.verbose
+                    gtex_scores_file,
+                    keyword,
+                    eqtl_df,
+                    args.snp_stat,
+                    verbose=args.verbose,
                 )
                 variant_scores = variant_scores[eqtl_df.consistent]
             except TypeError:
-                print(f'Tracks matching {tissue} are missing', file=sys.stderr)
+                print(f"Tracks matching {tissue} are missing", file=sys.stderr)
                 continue
 
             # compute sign AUROCs

src/westminster/scripts/westminster_train_composer.py

@@ -60,6 +60,12 @@ def main():
         default="train_out",
         help="Training output directory [Default: %default]",
     )
+    train_options.add_option(
+        "-l",
+        dest="log_dir",
+        default="log_out",
+        help="Tensorboard log directory [Default: %default]",
+    )
     train_options.add_option(
         "--restore",
         dest="restore",
@@ -184,6 +190,7 @@ def main():
         print("Output directory %s exists. Please remove." % options.out_dir)
         exit(1)
     os.makedirs(options.out_dir, exist_ok=True)
+    os.makedirs(options.log_dir, exist_ok=True)
 
     # read model parameters
     with open(params_file) as params_open:
@@ -229,6 +236,7 @@ def main():
     for ci in range(options.crosses):
         for fi in range(num_folds):
             rep_dir = "%s/f%dc%d" % (options.out_dir, fi, ci)
+            rep_log_dir = "%s/f%dc%d" % (options.log_dir, fi, ci)
 
             train_dir = "%s/train" % rep_dir
             if options.restart and not options.checkpoint and os.path.isdir(train_dir):
@@ -242,7 +250,7 @@ def main():
 
                 # train command
                 cmd = "python3 -m baskerville.scripts.hound_train"
-                cmd += " %s" % options_string(options, train_options, rep_dir)
+                cmd += " %s" % options_string(options, train_options, rep_dir, rep_log_dir)
                 cmd += " %s %s" % (params_file, " ".join(rep_data_dirs))
                 train_jobs.append(cmd)
 
@@ -459,7 +467,7 @@ def make_rep_data(data_dir, rep_data_dir, fi, ci):
         ti += 1
 
 
-def options_string(options, train_options, rep_dir):
+def options_string(options, train_options, rep_dir, rep_log_dir):
     options_str = ""
 
     for opt in train_options.option_list:
@@ -485,6 +493,10 @@ def options_string(options, train_options, rep_dir):
         elif opt.dest == "out_dir":
             opt_value = "%s/train" % rep_dir
 
+        # modify
+        elif opt.dest == "log_dir":
+            opt_value = "%s/log" % rep_log_dir
+
         # find matching restore
         elif opt.dest == "restore":
             fold_dir_mid = rep_dir.split("/")[-1]