Training in progress, step 1000

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:066a4bd36d31aed8395dd3cd4af6a1b5224e7f6bdce73c3d445243e4ad5fc29d
+oid sha256:1de3a6d93fb07b3b0d98ebec9911cf2e3ca0bdb3518687fb8882b35497d8c94d
 size 3055754841

run.sh

@@ -5,13 +5,14 @@ python run_speech_recognition_seq2seq_streaming.py \
 --language="vietnamese" \
 --train_split_name="train+validation" \
 --eval_split_name="test" \
---model_index_name="Whisper Medium Vietnamese Vivos" \
+--model_index_name="Whisper Medium Vietnamese" \
 --max_steps="5000" \
 --output_dir="./" \
 --per_device_train_batch_size="32" \
---per_device_eval_batch_size="16" \
+--per_device_eval_batch_size="32" \
+--gradient_accumulation_steps="2" \
 --logging_steps="25" \
---learning_rate="1e-6" \
+--learning_rate="1e-5" \
 --warmup_steps="500" \
 --evaluation_strategy="steps" \
 --eval_steps="1000" \
@@ -25,7 +26,6 @@ python run_speech_recognition_seq2seq_streaming.py \
 --report_to="tensorboard" \
 --metric_for_best_model="wer" \
 --greater_is_better="False" \
---gradient_accumulation_steps="1" \
 --load_best_model_at_end \
 --gradient_checkpointing \
 --fp16 \
@@ -34,5 +34,6 @@ python run_speech_recognition_seq2seq_streaming.py \
 --do_eval \
 --predict_with_generate \
 --do_normalize_eval \
+--streaming \
 --use_auth_token \
---push_to_hub
+--push_to_hub

run_speech_recognition_seq2seq_streaming.py

@@ -415,7 +415,7 @@ def main():
     # 4. Load dataset
     dataset_names = ["mozilla-foundation/common_voice_11_0", "mozilla-foundation/common_voice_11_0", "vivos", "vivos"]
     dataset_config_names = ["vi", "vi", None, None]
-    text_column_names = ["sentence", "sentence", "sentence", "sentence"]
+    text_column_names = ["sentence", "sentence",  "sentence", "sentence"]
     splits = ['train', 'validation', 'train', 'test']
     
     raw_datasets = IterableDatasetDict()

runs/Dec16_14-32-26_129-213-131-48/1671201160.8968897/events.out.tfevents.1671201160.129-213-131-48.2969237.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06fbefe6b792f6d3495e9600d5939f5bf16679802daa37806cf0dc028923e44b
+size 5870

runs/Dec16_14-32-26_129-213-131-48/events.out.tfevents.1671201160.129-213-131-48.2969237.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e37f2986e473f5ed8943f7ef3ea97e844574456634125bdfc60091ac1ad6b846
+size 10899

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4c7d8508659c5d42ba410fbec08857ff731f35eecf4e825059dfa8cb5e29cd5d
+oid sha256:d4bd990ffd86596f3507ce3d35a1a36cc02e044793d691e47ee94fbf31352539
 size 3579