TongZheng1999

/

alma-13b-sft-50-languages-bg-max-tokens-512

like 0

• "__builtin__.set",
• "collections.OrderedDict",
• "torch.HalfStorage",
• "torch.Size",
• "torch._utils._rebuild_tensor_v2"

How to fix it?

• "deepspeed.utils.tensor_fragment.fragment_address",
• "torch.FloatStorage",
• "collections.OrderedDict",
• "deepspeed.runtime.fp16.loss_scaler.DynamicLossScaler",
• "torch.float16",
• "torch._utils._rebuild_tensor_v2",
• "deepspeed.runtime.zero.config.ZeroStageEnum"

How to fix it?

• "deepspeed.utils.tensor_fragment.fragment_address",
• "torch.FloatStorage",
• "collections.OrderedDict",
• "deepspeed.runtime.fp16.loss_scaler.DynamicLossScaler",
• "torch.float16",
• "torch._utils._rebuild_tensor_v2",
• "deepspeed.runtime.zero.config.ZeroStageEnum"

How to fix it?

• "deepspeed.utils.tensor_fragment.fragment_address",
• "torch.float16",
• "deepspeed.runtime.fp16.loss_scaler.DynamicLossScaler",
• "torch.FloatStorage",
• "deepspeed.runtime.zero.config.ZeroStageEnum",
• "torch._utils._rebuild_tensor_v2",
• "collections.OrderedDict"

How to fix it?

• "deepspeed.utils.tensor_fragment.fragment_address",
• "torch.FloatStorage",
• "collections.OrderedDict",
• "deepspeed.runtime.fp16.loss_scaler.DynamicLossScaler",
• "torch.float16",
• "torch._utils._rebuild_tensor_v2",
• "deepspeed.runtime.zero.config.ZeroStageEnum"

How to fix it?