Training in progress, step 1100

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:81ccdc32f560a2aebc2dbfb88eac81745809a56769c2244d2e253468e0afac1c
+oid sha256:ce652af1ecb6a681a0dfb56afdd0c2f5cd52a8f5fc0b282fe98b6fdb49c5fe2d
 size 3055754841

run_1.sh

@@ -6,14 +6,14 @@ python run_speech_recognition_seq2seq_streaming.py \
 --train_split_name="train+validation" \
 --eval_split_name="test" \
 --model_index_name="Whisper Medium Pashto" \
---max_steps="1000" \
+--max_steps="1200" \
 --output_dir="./" \
 --per_device_train_batch_size="32" \
   --per_device_eval_batch_size="16" \
   --gradient_accumulation_steps="2" \
 --logging_steps="25" \
 --learning_rate="3e-7" \
---warmup_steps="50" \
+--warmup_steps="10" \
 --evaluation_strategy="steps" \
 --eval_steps="100" \
 --save_strategy="steps" \
@@ -25,6 +25,8 @@ python run_speech_recognition_seq2seq_streaming.py \
 --freeze_feature_encoder="False" \
 --report_to="tensorboard" \
 --metric_for_best_model="wer" \
+--greater_is_better="False" \
+--load_best_model_at_end \
 --gradient_checkpointing \
 --fp16 \
 --overwrite_output_dir="False" \

runs/Dec10_09-37-22_132-145-129-157/1670665064.2132027/events.out.tfevents.1670665064.132-145-129-157.108751.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1e0857f7ddfcc5581c37d2395feb035e572467402b8b3b820bd965b32d037d7
+size 5871

runs/Dec10_09-37-22_132-145-129-157/events.out.tfevents.1670665064.132-145-129-157.108751.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f419ca8ebb5f934217de1095336ea06fb28d6e86d512e72cb534426a6668e986
+size 5214

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ad23432f1c5bf8f56438d4a86cff84e1ec6bb5d753feb5aaf092bada25f07048
+oid sha256:a9805d9bd109ed2a40594bfd2ab0a7e52952d96b67a7bca4857b01ac1f91b928
 size 3579