xtuner

/

llava-llama-3-8b-pretrain

like 1

• "numpy.dtype",
• "mmengine.logging.history_buffer.HistoryBuffer",
• "numpy.core.multiarray.scalar",
• "numpy.core.multiarray._reconstruct",
• "numpy.ndarray",
• "torch.BFloat16Storage",
• "torch._utils._rebuild_tensor_v2",
• "collections.OrderedDict",
• "__builtin__.getattr",
• "_codecs.encode"

How to fix it?