commit 310b01db6dde3aa1114d701894488fca0c9e2f0b
author Austin Schuh <austin.linux@gmail.com> Sat Oct 26 15:55:47 2024 -0700
committer Austin Schuh <austin.linux@gmail.com> Sat Oct 26 17:32:30 2024 -0700
tree adbf8d1f46218249e926f187d932e6c613583d44
parent 5687cdba59b44d7fdc77ed4f5ec09d465a4d842b

Rename --start_steps to --random_sample_steps for training

The previous name wasn't very descriptive.  Stop using it.

Change-Id: Ia022a4a8b43cd7e9285f1892e1b11c7f297d1c84
Signed-off-by: Austin Schuh <austin.linux@gmail.com>

frc971/control_loops/swerve/velocity_controller/train.py

diff --git a/frc971/control_loops/swerve/velocity_controller/train.py b/frc971/control_loops/swerve/velocity_controller/train.py
index ed0cdeb..2c3f56c 100644
--- a/frc971/control_loops/swerve/velocity_controller/train.py
+++ b/frc971/control_loops/swerve/velocity_controller/train.py
@@ -35,7 +35,7 @@
 )
 
 absl.flags.DEFINE_integer(
-    'start_steps',
+    'random_sample_steps',
     default=10000,
     help='Number of steps to randomly sample before using the policy',
 )
@@ -319,7 +319,7 @@
             return pi_action
 
         pi_action = jax.lax.cond(
-            step <= FLAGS.start_steps,
+            step <= FLAGS.random_sample_steps,
             true_fn,
             false_fn,
             i,
@@ -400,6 +400,7 @@
         'q_learning_rate': FLAGS.q_learning_rate,
         'pi_learning_rate': FLAGS.pi_learning_rate,
         'alpha_learning_rate': FLAGS.alpha_learning_rate,
+        'random_sample_steps': FLAGS.random_sample_steps,
         'batch_size': FLAGS.batch_size,
         'horizon': FLAGS.horizon,
         'warmup_steps': FLAGS.warmup_steps,