Hide several less-used CLI options

Let's be real, what are the chances a Vamb user is going to experiment with tweaking
alpha or the latent space size?
It's nice to have these flags when testing, but we don't need to display them
in the help menu. All they do is confuse users and drown out the actually
important parameters users need to set.

vamb/__main__.py

@@ -1696,7 +1696,7 @@ def add_input_output_arguments(subparser):
 
     # RPKM arguments
     rpkmos = subparser.add_argument_group(
-        title="RPKM input (either BAMs or .npz required)"
+        title="RPKM input (exactly one input required)"
     )
     # Note: This argument is deprecated, but we'll keep supporting it for now.
     # Instead, use --bamdir.
@@ -1753,7 +1753,7 @@ def add_input_output_arguments(subparser):
         metavar="",
         type=float,
         default=None,
-        help="ignore reads with nucleotide identity below this [0.0]",
+        help=argparse.SUPPRESS,
     )
     inputos.add_argument(
         "-p",
@@ -1762,7 +1762,7 @@ def add_input_output_arguments(subparser):
         type=int,
         default=DEFAULT_THREADS,
         help=(
-            "number of threads to use " "[min(" + str(DEFAULT_THREADS) + ", nbamfiles)]"
+            "number of threads to use when parsing BAM" "[min(" + str(DEFAULT_THREADS) + ", nbamfiles)]"
         ),
     )
     inputos.add_argument(
@@ -1805,53 +1805,53 @@ def add_vae_arguments(subparser):
         type=int,
         nargs="+",
         default=None,
-        help="hidden neurons [Auto]",
+        help=argparse.SUPPRESS,
     )
     vaeos.add_argument(
         "-l",
         dest="nlatent",
         metavar="",
         type=int,
         default=32,
-        help="latent neurons [32]",
+        help=argparse.SUPPRESS,
     )
     vaeos.add_argument(
         "-a",
         dest="alpha",
         metavar="",
         type=float,
         default=None,
-        help="alpha, weight of TNF versus depth loss [Auto]",
+        help=argparse.SUPPRESS,
     )
     vaeos.add_argument(
         "-b",
         dest="beta",
         metavar="",
         type=float,
         default=200.0,
-        help="beta, capacity to learn [200.0]",
+        help=argparse.SUPPRESS,
     )
     vaeos.add_argument(
         "-d",
         dest="dropout",
         metavar="",
         type=float,
         default=None,
-        help="dropout [Auto]",
+        help=argparse.SUPPRESS,
     )
 
     trainos = subparser.add_argument_group(title="Training options", description=None)
 
     trainos.add_argument(
-        "-e", dest="nepochs", metavar="", type=int, default=300, help="epochs [300]"
+        "-e", dest="nepochs", metavar="", type=int, default=300, help=argparse.SUPPRESS
     )
     trainos.add_argument(
         "-t",
         dest="batchsize",
         metavar="",
         type=int,
         default=256,
-        help="starting batch size [256]",
+        help=argparse.SUPPRESS,
     )
     trainos.add_argument(
         "-q",
@@ -1860,7 +1860,7 @@ def add_vae_arguments(subparser):
         type=int,
         nargs="*",
         default=[25, 75, 150, 225],
-        help="double batch size at epochs [25 75 150 225]",
+        help=argparse.SUPPRESS,
     )
     trainos.add_argument(
         "-r",
@@ -1885,31 +1885,31 @@ def add_predictor_arguments(subparser):
         metavar="",
         type=int,
         default=100,
-        help="taxonomy predictor epochs [100]",
+        help=argparse.SUPPRESS,
     )
     pred_trainos.add_argument(
         "-pt",
         dest="pred_batchsize",
         metavar="",
         type=int,
         default=1024,
-        help="batch size for the taxonomy predictor [1024]",
+        help=argparse.SUPPRESS,
     )
     pred_trainos.add_argument(
         "-pthr",
         dest="pred_softmax_threshold",
         metavar="",
         type=float,
         default=0.5,
-        help="conditional probability threshold for accepting the taxonomic prediction [0.5]",
+        help=argparse.SUPPRESS,
     )
     pred_trainos.add_argument(
         "-ploss",
         dest="ploss",
         metavar="",
         type=str,
         default="flat_softmax",
-        help='Hierarchical loss (one of flat_softmax, cond_softmax, soft_margin) ["flat_softmax"]',
+        help=argparse.SUPPRESS,
     )
     return subparser
 
@@ -1923,15 +1923,15 @@ def add_clustering_arguments(subparser):
         metavar="",
         type=int,
         default=300,
-        help="size of window to count successes [300]",
+        help=argparse.SUPPRESS,
     )
     clusto.add_argument(
         "-u",
         dest="min_successes",
         metavar="",
         type=int,
         default=15,
-        help="minimum success in window [15]",
+        help=argparse.SUPPRESS,
     )
     clusto.add_argument(
         "-c",
@@ -1951,7 +1951,7 @@ def add_clustering_arguments(subparser):
         default=None,
         const="",
         nargs="?",
-        help="binsplit separator [C] (pass empty string to disable)",
+        help="binsplit separator [C if present] (pass empty string to disable)",
     )
     return subparser
 
@@ -1989,47 +1989,47 @@ def add_aae_arguments(subparser):
         metavar="",
         type=int,
         default=547,
-        help="hidden neurons AAE [547]",
+        help=argparse.SUPPRESS,
     )
     aaeos.add_argument(
         "--z_aae",
         dest="nlatent_aae_z",
         metavar="",
         type=int,
         default=283,
-        help="latent neurons AAE continuous latent space  [283]",
+        help=argparse.SUPPRESS,
     )
     aaeos.add_argument(
         "--y_aae",
         dest="nlatent_aae_y",
         metavar="",
         type=int,
         default=700,
-        help="latent neurons AAE categorical latent space [700]",
+        help=argparse.SUPPRESS,
     )
     aaeos.add_argument(
         "--sl_aae",
         dest="sl",
         metavar="",
         type=float,
         default=0.00964,
-        help="loss scale between reconstruction loss and adversarial losses [0.00964]",
+        help=argparse.SUPPRESS,
     )
     aaeos.add_argument(
         "--slr_aae",
         dest="slr",
         metavar="",
         type=float,
         default=0.5,
-        help="loss scale between reconstruction adversarial losses [0.5] ",
+        help=argparse.SUPPRESS,
     )
     aaeos.add_argument(
         "--aae_temp",
         dest="temp",
         metavar="",
         type=float,
         default=0.1596,
-        help=" Temperature of the softcategorical prior [0.1596]",
+        help=argparse.SUPPRESS,
     )
 
     aaetrainos = subparser.add_argument_group(
@@ -2042,15 +2042,15 @@ def add_aae_arguments(subparser):
         metavar="",
         type=int,
         default=70,
-        help="epochs AAE [70]",
+        help=argparse.SUPPRESS,
     )
     aaetrainos.add_argument(
         "--t_aae",
         dest="batchsize_aae",
         metavar="",
         type=int,
         default=256,
-        help="starting batch size AAE [256]",
+        help=argparse.SUPPRESS,
     )
     aaetrainos.add_argument(
         "--q_aae",
@@ -2059,7 +2059,7 @@ def add_aae_arguments(subparser):
         type=int,
         nargs="*",
         default=[25, 50],
-        help="double batch size at epochs AAE [25,50]",
+        help=argparse.SUPPRESS,
     )
     return subparser