test

run_flax_speech_recognition_seq2seq_streaming.py

@@ -754,13 +754,11 @@ def main():
             if filename.startswith("step_"):
                 with open(os.path.join(predictions_folder_name, filename), "r+") as f:
                     content = f.read()
-                    try:
-                        new_content = md_header + "\n" + content[content.index("| STEP| loss | wer"):]
-                        f.seek(0)
-                        f.write(new_content)
-                        f.truncate()
-                    except:
-                        print(f"Some strange stuff happened for file {filename}")
+                    new_content = md_header + "\n\n" + content[content.index("| STEP| loss | wer"):]
+                    f.seek(0)
+                    f.write(new_content)
+                    f.truncate()
+                   
                                 
         logger.info(f"Created {stats_file_name} and updated the headers of the other stats files")
     

run_streaming_pere.sh

@@ -1,5 +1,5 @@
 python run_flax_speech_recognition_seq2seq_streaming.py \
-            --model_name_or_path openai/whisper-tiny.en \
+            --model_name_or_path openai/whisper-small.en \
             --dataset_name mozilla-foundation/common_voice_11_0 \
             --dataset_config nn-NO \
             --text_column_name sentence \
@@ -14,8 +14,9 @@ python run_flax_speech_recognition_seq2seq_streaming.py \
             --num_train_steps 40 \
             --eval_steps 10 \
             --learning_rate=2e-4 \
-            --per_device_train_batch_size=2 \
+            --per_device_train_batch_size=184 \
             --per_device_eval_batch_size=2 \
             --predict_with_generate \
             --write_predictions \
-            --streaming=True
+            --streaming=True \
+            --push_to_hub