Use return_tensors="np" instead of "tf" (#21266)

Return NP instead of TF tensors for our data loading pipeline
This commit is contained in:
Matt
2023-01-24 13:37:49 +00:00
committed by GitHub
parent f0fc791298
commit 071529bd54
6 changed files with 8 additions and 8 deletions

View File

@@ -499,7 +499,7 @@ def main():
# region TF Dataset preparation
num_replicas = training_args.strategy.num_replicas_in_sync
data_collator = DataCollatorForLanguageModeling(
tokenizer=tokenizer, mlm_probability=data_args.mlm_probability, return_tensors="tf"
tokenizer=tokenizer, mlm_probability=data_args.mlm_probability, return_tensors="np"
)
options = tf.data.Options()
options.experimental_distribute.auto_shard_policy = tf.data.experimental.AutoShardPolicy.OFF