Initial prototype for wandb sweep

ivadomed/scripts/train_wandb_sweep_hyperparam.py

@@ -0,0 +1,40 @@
+#!/usr/bin/env python
+"""
+Launcher for training with the possibility to specify hyperparameters. This script is to be used by wandb-sweep.
+https://docs.wandb.ai/guides/sweeps
+
+Input: template JSON ivadomed config file, which will be edited by this script to insert the random parameters.
+
+Resumption of interrupted runs with
+    python train_wandb_sweep_hyperparam.py --resume
+
+"""
+
+import sys
+
+import wandb
+from wandb.keras import WandbCallback
+
+
+defaults = dict(
+    dropout=0.2,
+    hidden_layer_size=128,
+    layer_1_size=16,
+    layer_2_size=32,
+    learn_rate=0.01,
+    decay=1e-6,
+    momentum=0.9,
+    epochs=27,
+    )
+
+resume = sys.argv[-1] == "--resume"
+wandb.init(config=defaults, resume=resume)
+
+# This is where the parameters are read from the YAML file
+config = wandb.config
+
+# Build ivadomed's config file based on the generated parameters
+# TODO
+
+# Call to ivadomed
+ivadomed.main()