log df of train and test data

run.sh

@@ -5,7 +5,7 @@ python run_speech_recognition_ctc.py \
 	--train_split_name="train+validation,train" \
 	--eval_split_name="test,None" \
 	--output_dir="./" \
-	--overwrite_output_dir \
+	--preprocessing_only \
 	--num_train_epochs="3" \
 	--per_device_train_batch_size="32" \
 	--per_device_eval_batch_size="32" \

run_speech_recognition_ctc.py

@@ -750,6 +750,12 @@ def main():
     # If dataset_seed is set, shuffle train
     if data_args.dataset_seed is not None:
         vectorized_datasets["train"] = vectorized_datasets["train"].shuffle(seed=data_args.dataset_seed)
+    
+    # Log sample of datasets
+    pd_train = vectorized_datasets["train"].select(range(10)).to_pandas()
+    pd_eval = vectorized_datasets["eval"].select(range(10)).to_pandas()
+    wandb.log({"train_sample": pd_train})
+    wandb.log("eval_sample": pd_eval)
 
     # for large datasets it is advised to run the preprocessing on a
     # single machine first with ``args.preprocessing_only`` since there will mostly likely