Skip to content

Commit 55db7e5

Browse files
committed
Fix output directory in LaunchLocalTrainingJob.ipynb
1 parent 0405248 commit 55db7e5

File tree

1 file changed

+9
-9
lines changed

1 file changed

+9
-9
lines changed

AirSimDistributedRL/LaunchLocalTrainingJob.ipynb

Lines changed: 9 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -41,7 +41,9 @@
4141
{
4242
"cell_type": "code",
4343
"execution_count": 4,
44-
"metadata": {},
44+
"metadata": {
45+
"collapsed": true
46+
},
4547
"outputs": [],
4648
"source": [
4749
"batch_update_frequency = 300\n",
@@ -60,18 +62,16 @@
6062
"cell_type": "markdown",
6163
"metadata": {},
6264
"source": [
63-
"We will now generate a training batch file. Run this file from an activated python environment to train."
65+
"We will now generate a training batch file. The file will be written to Share\\scripts_downpour\\app. Run this file from an activated python environment in that directory to train."
6466
]
6567
},
6668
{
6769
"cell_type": "code",
68-
"execution_count": 5,
69-
"metadata": {
70-
"collapsed": true
71-
},
70+
"execution_count": 11,
71+
"metadata": {},
7272
"outputs": [],
7373
"source": [
74-
"train_cmd = 'python Share\\\\app\\\\scripts_downpour\\\\distributed_agent.py'\n",
74+
"train_cmd = 'python distributed_agent.py'\n",
7575
"train_cmd += ' batch_update_frequency={0}'.format(batch_update_frequency)\n",
7676
"train_cmd += ' max_epoch_runtime_sec={0}'.format(max_epoch_runtime_sec)\n",
7777
"train_cmd += ' per_iter_epsilon_reduction={0}'.format(per_iter_epsilon_reduction)\n",
@@ -83,15 +83,15 @@
8383
"train_cmd += ' airsim_path={0}'.format(airsim_path)\n",
8484
"train_cmd += ' data_dir={0}'.format(data_dir)\n",
8585
"\n",
86-
"with open('train.bat', 'w') as f:\n",
86+
"with open(os.path.join(os.getcwd(), 'Share/scripts_downpour/app/train.bat'), 'w') as f:\n",
8787
" f.write(train_cmd)"
8888
]
8989
},
9090
{
9191
"cell_type": "markdown",
9292
"metadata": {},
9393
"source": [
94-
"Note that training the m"
94+
"Note that training the model from scratch can take up to 5 days on a powerful GPU. Using pretrained weights, the model should begin to visibly converge after 3 hours of training on a GTX 1080."
9595
]
9696
}
9797
],

0 commit comments

Comments
 (0)