weight is on the meta device, we need a `value` to put it in 0 when convert Qwen2-0.5B

[austingg]

I have successfully SFTed Qwen2-0.5B using XTuner. When I convert the qlora checkpoint to hf, weight is on the meta device, we need a value to put it in 0 occurs.

convert command

CUDA_VISIBLE_DEVICES=0 xtuner convert pth_to_hf qwen2_config qwen2_checkpoint qwen2_res

Environment

• GPU: V100
• XTuner: 0.1.21
• CUDA: 11.8

[tttmck]

have you resolved it?

[GoatYoung]

I have the same issue.

[tttmck]

I have the same issue.

I have already solved this problem. It is strongly recommended that you check the configuration file parameters. In most cases, this problem is caused by a mismatch or absence of parameters of the model or dataset.