bash scripts/graph_constrained_decoding.sh get some issues.

[sarhangpour1986-ai]

Hi,
When I run bash scripts/graph_constrained_decoding.sh
I get some issue like:
view size is not compatible with input tensor's size and stride (at least one dimension spans across two contiguous subspaces). Use .reshape(...) instead.
None result for: WebQTest-0
0%| | 1/1628 [00:02<1:09:31, 2.56s/it]view size is not compatible with input tensor's size and stride (at least one dimension spans across two contiguous subspaces). Use .reshape(...) instead.
None result for: WebQTest-1
0%|▏ | 2/1628 [00:03<38:36, 1.42s/it]view size is not compatible with input tensor's size and stride (at least one dimension spans across two contiguous subspaces). Use .reshape(...) instead.

Can't get answer,I don't know why.The model I uesd is rmanluo/GCR-Qwen2-0.5B-Instruct.I hope you can help me clarify my doubts.
I have previously attempted to use the Qwen-0.5B model I trained myself for graph reasoning decoding, but encountered errors as well. It seems the issue is related to the vocabulary size.like that:

RuntimeError: Error(s) in loading state_dict for Qwen2ForCausalLM:
size mismatch for model.embed_tokens.weight: copying a param with shape torch.Size([151648, 896]) from checkpoint, the shape in current model is torch.Size([151936, 896]).
size mismatch for lm_head.weight: copying a param with shape torch.Size([151648, 896]) from checkpoint, the shape in current model is torch.Size([151936, 896]).