Wei-Peng

training_logits, targets维度不匹配

2

cost = tf.contrib.seq2seq.sequence_loss(training_logits, targets, masks) tensorflow.python.framework.errors_impl.InvalidArgumentError: Incompatible shapes: [5632] vs. [6400] 就是我的target是（256，25） 可是输出得到的training_logits却是（256， 22， 358）358：词表数 我改了一下，这样就对了 ``` def pad_batch_sentence(batch, max_length, pad_id): # max_length = max([len(sentence) for sentence in batch]) return...

which GPU could I used if I want to do alpaca_finetuning_v1 on a single GPU?

2

see title，thanks

multi-gpu

1

"AssertionError: Loading a checkpoint for MP=1 but world size is 2" when I set --nproc_per_node to 2. How to run on 2 16G gpu? because it OOM when inference. Thanks

as title

Out of memory. 48G is not enough, either. What happend?

1

Out of memory. 48G is not enough, either. What happend?

When using the script marcov2_to_dureader.py to convert MARCOv2 to dureader, it failed because ValueError: Trailing data The command: sh run_marco2dureader_preprocess.sh ../Marco/train_v2.1.json ../Marco/train_v2.1_dureaderformat.json But it occurs an error -- ValueError: Trailing...