Training in progress, step 1000

README.md

@@ -7,8 +7,6 @@ tags:
 - generated_from_trainer
 datasets:
 - asierhv/composite_corpus_eu_v2.1
-language:
-- eu
 metrics:
 - wer
 model-index:

model-00001-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:07cfdf761c97ff20bb92439c8d2aa688992c5dc607decc2ce7fe7161d9db9e7f
+oid sha256:9f90393365a7608a490b562e2c4d4b86380fa95db974370b7eb6e5eee3d93508
 size 4993448880

model-00002-of-00002.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4ff0450e9cc19db3496689ac7771bc899dc40448f074095faa38b20458a782d3
+oid sha256:b221a4a91dbabd555a9fefa153d0f7778c7256b345fc674cad52b3ae7a222b78
 size 1180663192

run.sh

@@ -1,10 +1,10 @@
-WANDB_PROJECT=whisper-medium-eu \
+WANDB_PROJECT=whisper \
 	python run_speech_recognition_seq2seq_streaming.py \
 	--model_name_or_path="openai/whisper-large-v3" \
 	--dataset_name="asierhv/composite_corpus_eu_v2.1" \
 	--language="basque" \
 	--train_split_name="train" \
-	--eval_split_name="dev_parl+test_parl+test_cv+test_oslr" \
+	--eval_split_name="dev" \
 	--model_index_name="Whisper Large Basque" \
 	--max_steps="20000" \
 	--output_dir="./" \
@@ -30,7 +30,6 @@ WANDB_PROJECT=whisper-medium-eu \
 	--gradient_checkpointing \
 	--fp16 \
 	--overwrite_output_dir \
-	--resume_from_checkpoint="checkpoint-10000" \
 	--do_train \
 	--do_eval \
 	--predict_with_generate \
@@ -38,4 +37,4 @@ WANDB_PROJECT=whisper-medium-eu \
 	--streaming \
 	--push_to_hub \
 	--report_to "wandb" \
-	--run_name "whisper-large-eu-v3"
+	--run_name "whisper-large-eu-v3-25.02-r1"

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7976217da07f1f6b61d94b0655d73812ba2b4478e1d3f5d21cac23d2cd1cce18
+oid sha256:35b687ecb445b882e871a8e50bf8f8f0cf5da184c31d6cc7f22385f74addd658
 size 5496