Fixed conflicts

speechbrain · Aug 14, 2023 · 067d94e · 067d94e
2 parents 327a3f5 + 75f4c66
commit 067d94e
Show file tree

Hide file tree

Showing 5 changed files with 20 additions and 8 deletions.
diff --git a/recipes/MOABB/README.md b/recipes/MOABB/README.md
@@ -130,7 +130,6 @@ For training the model using the `leave_one_subject_out` training approach, just
 
 ### Hyperparameter Tuning
 
-
 Efficient hyperparameter tuning is paramount when introducing novel models or experimenting with diverse datasets. Our benchmark establishes a standardized protocol for hyperparameter tuning, utilizing [Orion](https://orion.readthedocs.io/en/stable/) to ensure fair model comparisons.
 
 #### **Overview**
@@ -192,6 +191,19 @@ As evident from the example, you need to configure the hyperparameter file, spec
 When it comes to training the model utilizing the leave_one_subject_out approach, simply employ the `--train_mode leave-one-subject-out` flag.
 
 By default trainings are performed on gpu. However, in case you do not have any gpu available on your machine, you can train models on cpu by specifying the `--device cpu` flag.
+
+**Note:**
+- To monitor the status of the hyperparameter optimization, simply enter the following command: `orion status --all`. Ensure that you have added the necessary variables required by orion to your bash environment. You can achieve this by executing the following code within your terminal:
+
+```bash
+export ORION_DB_ADDRESS=results/MotorImagery/BNCI2014001/EEGNet/hopt/EEGNet_BNCI2014001_hopt.pkl
+export ORION_DB_TYPE=pickleddb
+```
+
+Please note that the value of the `ORION_DB_ADDRESS` variable will vary depending on the experiment. Adjust it accordingly.
+
+- If needed, you can interrupt the code at any point, and it will resume from the last successfully completed experiment.
+
 #### **Output Structure**
 
 Results are organized within the specified output folder (`--output_folder`):
@@ -213,8 +225,6 @@ For further details on arguments and customization options, consult `./run_hpara
 
 - If you intend to perform multiple repetitions of the same hparam optimization, it is necessary to modify the `--exp_name`.
 
-- Feel free to interrupt and resume the hyperparameter optimization script at any point without encountering any complications. Orion is resumable and will restart from the last experiment completed.
-
 - This script is designed for a Linux-based system. In this context, we provide a bash script instead of a Python script due to its natural ability of orchestrating diverse training loops across various subjects and sessions.
 
 

diff --git a/recipes/MOABB/extra_requirements.txt b/recipes/MOABB/extra_requirements.txt
@@ -1,5 +1,5 @@
 mne
 moabb
 orion[profet]
-sklearn
+scikit-learn
 torchinfo
diff --git a/recipes/MOABB/hparams/orion/hparams_tpe.yaml b/recipes/MOABB/hparams/orion/hparams_tpe.yaml
@@ -1,5 +1,5 @@
 experiment:
-    algorithms:
+    algorithm:
         tpe:
             seed: 1986
             n_initial_points: 20

diff --git a/recipes/MOABB/run_experiments.sh b/recipes/MOABB/run_experiments.sh
@@ -175,6 +175,9 @@ if [ "$rnd_dir" = True ]; then
     output_folder="$output_folder/$rnd_dirname"
 fi
 
+# Make sure  the output_folder is created
+mkdir -p $output_folder
+
 # Print command line arguments and save to file
 {
     echo "hparams: $hparams"

diff --git a/recipes/MOABB/run_hparam_optimization.sh b/recipes/MOABB/run_hparam_optimization.sh
@@ -46,7 +46,6 @@
 # - Davide Borra (2023)
 ###########################################################
 
-
 # Initialize variables
 exp_name=""
 output_folder=""
@@ -65,8 +64,8 @@ mne_dir=""
 orion_db_address=""
 orion_db_type="PickledDB"
 exp_max_trials=50
-store_all=False
-compress_exp=False
+store_all=True
+compress_exp=True
 
 # Function to print argument descriptions and exit
 print_argument_descriptions() {