leon-cas
Results
2
comments of
leon-cas
> @coddinglxf I just solved that problem with `nn.NLLLoss(ignore_index=0)` which 0 is equal to pad_index. Even if we target the 0(unmasked_value), it doesn't affect to the loss of propagation why...
> 需要 4.49.0 @hiyouga 目前用的transformer就是4.49.0， 在相同的配置下，训练qwen2.5-vl相比qwen2-vl要慢5-6x； 不知该问题有解没？