Model save

config.json

@@ -17,7 +17,7 @@
   "decoder_layerdrop": 0.0,
   "decoder_layers": 6,
   "decoder_start_token_id": 50258,
-  "dropout": 0.3,
+  "dropout": 0.4,
   "encoder_attention_heads": 8,
   "encoder_ffn_dim": 2048,
   "encoder_layerdrop": 0.0,

run.sh

@@ -12,7 +12,7 @@ python run_speech_recognition_seq2seq_streaming.py \
 --per_device_eval_batch_size="32" \
 --gradient_accumulation_steps="1" \
 --logging_steps="10" \
---learning_rate="1e-5" \
+--learning_rate="1e-6" \
 --warmup_steps="30" \
 --evaluation_strategy="steps" \
 --eval_steps="100" \

run_speech_recognition_seq2seq_streaming.py

@@ -556,7 +556,7 @@ def main():
             elif isinstance(train_dataloader.dataset, IterableDataset):
                 train_dataloader.dataset.set_epoch(train_dataloader.dataset._epoch + 1)
 
-    model.config.dropout = 0.3
+    model.config.dropout = 0.4
     # Initialize Trainer
     trainer = Seq2SeqTrainer(
         model=model,

runs/Dec20_19-44-59_129-146-32-172/1671565514.8929353/events.out.tfevents.1671565514.129-146-32-172.144860.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f22e60461db94cd7e811ccbfb51c5ae03a2332f25e1b78df1e1fe69f8cebc15
+size 5871

runs/Dec20_19-44-59_129-146-32-172/events.out.tfevents.1671565514.129-146-32-172.144860.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57c0376246845e980e7bac514b13415fb28a3cee9165b2e20d4533d7e0033113
+size 4614

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:49fe45c6cb7f317e4bef576d0096f6218b8b88aac9a26faa50a5a451ace6c3fc
+oid sha256:82dc9b5c0b7437cb9e4f9be835037b5a9a619b3446fb5be7b09cc8bfeaff4ff4
 size 3579