Training in progress, step 100

config.json

@@ -17,7 +17,7 @@
   "decoder_layerdrop": 0.0,
   "decoder_layers": 4,
   "decoder_start_token_id": 50258,
-  "dropout": 0.0,
+  "dropout": 0.2,
   "encoder_attention_heads": 6,
   "encoder_ffn_dim": 1536,
   "encoder_layerdrop": 0.0,

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f3843686519777a4550909e8bd4961dcf7425e7183295f03d09a433a271f0887
+oid sha256:ff96921d8c3b6e5c120dc16cd09a558ba93bf720c4de39d0fc5dbe79b4468164
 size 151098921

run_1.sh

@@ -0,0 +1,37 @@
+python run_speech_recognition_seq2seq_streaming.py \
+--model_name_or_path="openai/whisper-tiny" \
+--dataset_name="google/fleurs" \
+--dataset_config_name="ps_af" \
+--language="pashto" \
+--train_split_name="train+validation" \
+--eval_split_name="test" \
+--model_index_name="Whisper Tiny Pashto" \
+--max_steps="200" \
+--output_dir="./" \
+--per_device_train_batch_size="64" \
+--per_device_eval_batch_size="32" \
+--gradient_accumulation_steps="1" \
+--logging_steps="10" \
+--learning_rate="3e-7" \
+--warmup_steps="40" \
+--evaluation_strategy="steps" \
+--eval_steps="100" \
+--save_strategy="steps" \
+--save_steps="100" \
+--generation_max_length="225" \
+--length_column_name="input_length" \
+--max_duration_in_seconds="30" \
+--text_column_name="transcription" \
+--freeze_feature_encoder="False" \
+--load_best_model_at_end \
+--report_to="tensorboard" \
+--gradient_checkpointing \
+--fp16 \
+--overwrite_output_dir="True" \
+--do_train \
+--do_eval \
+--predict_with_generate \
+--do_normalize_eval \
+--use_auth_token \
+--streaming="False" \
+--push_to_hub

run_speech_recognition_seq2seq_streaming.py

@@ -556,6 +556,7 @@ def main():
             elif isinstance(train_dataloader.dataset, IterableDataset):
                 train_dataloader.dataset.set_epoch(train_dataloader.dataset._epoch + 1)
 
+    model.config.dropout=0.2
     # Initialize Trainer
     trainer = Seq2SeqTrainer(
         model=model,

runs/Dec18_14-47-23_129-146-179-188/1671375054.7599893/events.out.tfevents.1671375054.129-146-179-188.128106.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ced6872824f9097402d177ba9acbfbb4db9fb3bcbc35fe09e556f36ddc6615f8
+size 5872

runs/Dec18_14-47-23_129-146-179-188/events.out.tfevents.1671375054.129-146-179-188.128106.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:920bc6c3d1c3287c26e8c137671e015172b1cb962f9fd089e820533dc58cf28d
+size 6111

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7ed9d485c4e39caa9b4a7d00e4304cc0361a5588c50e4b825534553db7003bde
+oid sha256:5cc4720d9aed8e3dfa8de40898a3a0253a9be1fbdae340a19e1743e520f212e6
 size 3579