Training in progress, step 10

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5ee0ed3b21f8827c9c10e63faec07dede381d3b08adfa9dd9abceddfb0e8bc75
+oid sha256:67db960796c75f5af90d6ae38074eea1cad281bdbec28d10187feba4432720bf
 size 151098921

runs/Dec13_13-15-52_d7f040c448a8/1670937365.1814487/events.out.tfevents.1670937365.d7f040c448a8.37644.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe77ac37a6e84c93ba30d55178bdae96ab16203d158ae76c88604fd477f1fa10
+size 5883

runs/Dec13_13-15-52_d7f040c448a8/events.out.tfevents.1670937365.d7f040c448a8.37644.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:27f15a0f910e391d361ad1e8ab069c12eaa75435d50f8a8a4a0a905019c49e3d
+size 4889

src/run_debug.sh

@@ -7,12 +7,13 @@ python src/run_speech_recognition_seq2seq_streaming.py \
 	--eval_split_name="validation" \
 	--model_index_name="Whisper Tiny Belarusian" \
 	\
-	--max_steps="100" \
+	--max_steps="200" \
 	--max_eval_samples="64" \
 	--output_dir="./" \
 	--per_device_train_batch_size="32" \
 	--per_device_eval_batch_size="32" \
 	--logging_steps="10" \
+	--logging_first_step \
 	--learning_rate="1e-4" \
 	--warmup_steps="10" \
 	--evaluation_strategy="steps" \

train.log

@@ -167,3 +167,5 @@
   eval_samples_per_second =      3.798
   eval_steps_per_second   =      0.119
   eval_wer                =    60.0733
+{'loss': 2.7922, 'learning_rate': 0.0, 'epoch': 0.01}
+{'loss': 2.5366, 'learning_rate': 7e-05, 'epoch': 0.05}

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c8a86b0841271f87ec4240b29b5ca1620f243ad5ad67d99f9919cf45237c2add
+oid sha256:a7060ae76bbcda709483e8d25020c5fe10166a869e6cd092088018350b53a30f
 size 3643