diff --git a/algorithmic_efficiency/data_utils.py b/algorithmic_efficiency/data_utils.py index 7d7053e7c..8879e6fe3 100644 --- a/algorithmic_efficiency/data_utils.py +++ b/algorithmic_efficiency/data_utils.py @@ -38,6 +38,7 @@ def shard_and_maybe_pad_np( remainder_size = current_batch_size % local_device_count # if pad_to_global_batch_size: if remainder_size != 0: + logging.info("PADDDINGGGGGGG") if global_batch_size is not None: pad_size = global_batch_size - current_batch_size else: @@ -62,6 +63,7 @@ def _prepare(x): # Pad if remainder_size != 0 (should only be possible during evaluation). # if pad_to_global_batch_size: if remainder_size != 0: + logging.info("PADDDINGGGGGG in _prepare") x = pad(x, pad_size, padding_value=padding_value) # Reshape (global_batch_size, ...) to